Datasets:

staghado

/

ArXiv-tables

like 2

1405.2377v1

\begin{table}[H] \centering \scriptsize \begin{tabular}{|p{3cm}|p{3cm}|p{4cm}|p{3cm}|} \hline \hline \multicolumn{4}{c}{Details of Experiments for the Employed Data Set}\\ \cline{1-4} \emph{Domain} & \emph{Raw Features} & \emph{Response} & \emph{Data Set Cardinality}\\ \hline Australian Credit Scoring & 16 & Desired credit approval of individuals based on characteristics & 690\\\hline \end{tabular} \caption{\small Data set descriptions for the experiments used to validate the efficacy of the proposed algorithm. We summarize here the domain of the application, the input features to the algorithm, the response variable we wish to predict and the number of examples provided in the data.} \end{table}

1405.2377v1

\begin{table}[H] \centering \scriptsize \begin{tabular}{|p{3cm}|p{2cm}|p{2cm}|p{2cm}|p{2cm}|} \hline \hline \multicolumn{5}{c}{Details of Experiments for the Variable Threshold Algorithm}\\ \cline{1-5} \emph{Statistic} & \emph{Average} & \emph{Minimum} & \emph{Maximum} & \emph{Standard Deviation}\\ \hline Predictive Accuracy of Random Forest & {\vspace{0mm}$85\%$} & {\vspace{0mm}$81\%$} & {\vspace{0mm}$90\%$} & {\vspace{0mm}$3.24\%$}\\\hline Convergence Time of Optimization Algorithm & {\vspace{0mm}$10$} & {\vspace{0mm}$7$} & {\vspace{0mm}$12$} & {\vspace{0mm}$2.2$}\\\hline \end{tabular} \caption{\small We present here some relevant statistics related to our experiments in parameter optimization. Notice that in the predictive accuracy criterion, larger values are preferable. By contrast, we have that convergence time is better for smaller values. We define as convergence time the number of iterations of the algorithm that are required to map out completely the known behavior of the accuracy function.} \end{table}

2207.05295v2

\begin{table}[t] \centering \resizebox{\columnwidth}{!}{ \begin{tabular}{c|c|c|c|c|c} \hline \multirow{3}{*}{Data}& \multirow{3}{*}{Number of rows}& \multicolumn{4}{c}{TabSynDex Score}\\ \cline{3-6} & & \multicolumn{4}{c}{\% of real data treated as synthetic}\\ \cline{3-6} & &10\%&25\%&50\%&100\%\\ \hline Concrete~\cite{concrete_data} &1030&0.768&0.869&0.914&0.894\\ \hline News Popularity~\cite{news_data} &39644&0.891&0.916&0.901&0.898\\ \hline Wine Quality~\cite{wine_data} &4898&0.867&0.911&0.925&0.938\\ \hline Power Plant~\cite{electrical_data} &9568&0.946&0.961&0.981&0.969\\ \hline \end{tabular} } \caption{Experiment for sanity check of the TabSynDex metric for tabular data synthesis evaluation. The real dataset is divided into different subsets to check the similarity between them using TabSynDex.~\textit{A higher TabSynDex score is better}} \label{tab:subset_similarity} \end{table}

1911.00623v2

\begin{table}[htbp] % \caption{Comparison of traditional machine learning algorithms. Notation: $m$-number of training samples; $n$-input dimension; $c$-number of classes.} % \centering % \begin{tabular}{ m{2.1cm}<{\centering}|m{1.8cm}<{\centering}| m{3.2cm}<{\centering}|m{2.8cm}<{\centering}|m{3.0cm}<{\centering}} % \toprule % Algorithm & Model size & Optimization & Training complexity & Inference complexity\\ % \noalign{ % \hrule height 2pt % } % Decision tree &$\mathcal{O}(m)$ &- &$\mathcal{O}(mnlog(m))$ & $\mathcal{O}(log(m))$\\ % \hline % Random forest & $\mathcal{O}(N_{tree}m)$ &- &$\mathcal{O}(N_{tree}mnlog(m))$ & $\mathcal{O}(N_{tree}log(m))$\\ % \hline % SVM & $\mathcal{O}(n)$ & gradient descent &$\mathcal{O}(m^2n)$ &$\mathcal{O}(m_{sv}n)$ \\ % \hline % Logistic regression & $\mathcal{O}(n)$ &Newton-Raphson &$\mathcal{O}(mn^2+n^3)$ & $\mathcal{O}(n)$ \\ % \hline % kNN &$\mathcal{O}(mn)$&- & - &$\mathcal{O}(mn)$ \\ % \hline % Naive Bayes &$\mathcal{O}(nc)$ &-&$\mathcal{O}(mn+nc)$ & $\mathcal{O}(nc)$ \\ % \hline % Linear regression &$\mathcal{O}(n)$ &matrix inversion &$\mathcal{O}(mn^2+n^3)$ &$\mathcal{O}(n)$ \\ % \noalign{ % \hrule height 2pt % } % k-Means &- &- &$\mathcal{O}(mnc)$ &- \\ % \hline % EM &-&-&$\mathcal{O}(mn^2+n^3)$ & -\\ % \noalign{ % \hrule height 2pt % } % PCA &- &eigen-decomposition &$\mathcal{O}(mn^2+n^3)$ & -\\ % \bottomrule % \end{tabular} % \label{table:MLcompare} % \end{table}

1911.00623v2

\begin{table}[htbp] % \caption{DNN resource requirements modeling. ASIC: Application-Specific Integrated Circuit. Matmul: matrix multiplication. RMSPE: root mean square percentage error.} % %\centering % %\begin{tabular}{ m{1.2cm}<{\centering}|m{1.6cm}<{\centering}| m{1.6cm}<{\centering}| m{1.8cm}<{\centering}|m{2.4cm}<{\centering}|m{2.2cm}<{\centering}|m{1.4cm}<{\centering}} % \begin{tabular}{ m{1.6cm}|m{1.6cm}| m{1.5cm}| m{1.4cm}|m{2.8cm}|m{1.4cm}|m{1.8cm}} % %\begin{tabularx}{1\linewidth}{X| X| X| X| X| X| X} % \toprule % Work & Platform & Framework & Metric & Measured features & Regression model & Relative ~~~~~~~~~~~~error \\ % \noalign{ % \hrule height 2pt % } % Augur\cite{lu2017modeling} & NVidia TK1, TX1 &Caffe &inference: memory, time &matrix dimensions in matmul, weights, activations&linear & memory: 28\% - 50\%; time: 6\% - 20\%\\ % \hline % Paleo\cite{qi2016paleo} & NVidia Titan X GPU cluster & TensorFlow&training \& inference: time&forward \& backward FLOPs, weights, activations, data, platform percent of peak&linear& 4\%-30\% \\ % \hline % Gianniti et al.\cite{giannitiperformance}&NVidia Quadro M6000 GPU & - &training: time & forward \& backward FLOPs of all types of layers & linear & < 23\% \\ % \hline % SyNERGY\cite{rodriguesfine} & Nvidia Jetson TX1 & Caffe & inference: energy & MACs & linear & < 17\% (w/o MobileNet)\\ % \hline % NeuralPower\cite{cai2017neuralpower}&Nvidia Titan X \& GTX 1070 &TensorFlow \& Caffe &inference: time, power, energy&layer configuration hyper-parameters, memory access, FLOPs, activations, batch size&polynomial&time: < 24\%; power: < 20\%; energy: < 5\%\\ % \hline % HyperPower\cite{stamoulis2018hyperpower}&Nvidia GTX1070 \& Tegra TX1&Caffe&inference: power, memory&layer configuration hyper-parameters&linear & RMSPE < 7\%\\ % \hline % Yang et al.\cite{yang2017designing} & ASIC Eyeriss\cite{chen2017eyeriss} & - & inference: energy & MACs, memory access&- & -\\ % \hline % DeLight\cite{rouhani2016delight} & Nvidia Tegra TK1& Theano & training\& inference: energy & layer configuration hyper-parameters & linear & -\\ % \bottomrule % \end{tabular} % \label{table:DNNresourcemodeling} % \end{table}

1911.00623v2

\begin{table}[htbp] \caption{Comparison of popular CNNs.} \centering \scriptsize \begin{tabular}{ m{2.4cm}<{\centering}|m{1.4cm}<{\centering}| m{1.4cm}<{\centering}| m{1.4cm}<{\centering}|m{1.4cm}<{\centering}|m{1.4cm}<{\centering}|m{1.4cm}<{\centering}} \toprule Metric & \makecell{ AlexNet \\ \cite{krizhevsky2012imagenet} } & \makecell{ VGG-16 \\ \cite{simonyan2014very}} & \makecell{ GoogLeNet \\ \cite{szegedy2015going} }& \makecell{ ResNet-18 \\ \cite{he2016deep}} & \makecell{ ResNet-50 \\ \cite{he2016deep}} & \makecell{ Inception\\ v3 \cite{szegedy2016rethinking} }\\ \noalign{ \hrule height 2pt } Top-1 acc. &57.2 &71.5 &69.8 &69.6 &76.0 &76.9 \\ \hline Top-5 acc. &80.2 &91.3 &90.0 &89.2 &93.0 &93.7 \\ \hline Input size &227$\times$227 &224$\times$224 & 224$\times$224&224$\times$224 &224$\times$224&299$\times$299 \\ \noalign{ \hrule height 2pt } $\#$ of stacked CONV layers &5 &13 & 21&17 &49 &16 \\ \hline Weights &2.3M &14.7M &6.0M & 9.5M&23.6M &22M \\ \hline Activations &0.94M &15.23M &6.8M &3.2M &11.5M &10.6M \\ \hline MACs &666M &15.3G &1.43G &1.8G &3.9G & 3.8G\\ \noalign{ \hrule height 2pt } $\#$ of FC layers &3 & 3&1 &1 & 1&1 \\ \hline Weights &58.7M &125M &1M &0.5M &2M & 2M\\ \hline Activations &9K &9K &2K &1.5K &3K &3K \\ \hline MACs &58.7M &125M&1M &0.5M &2M &2M \\ \noalign{ \hrule height 2pt } Total weights &61M &138M & 7M& 10M&25.6M &24M \\ \hline Total activations &0.95M &15.24M &6.8M &3.2M &11.5M &10.6M \\ \hline Total MACs &724M &15.5G &1.43G &1.8G &3.9G &3.8G \\ \bottomrule \end{tabular} \normalsize \label{table:bigCNNcompare} \end{table}

1911.00623v2

\begin{table}[tbp] \caption{Comparison of lightweight CNNs.} \centering \scriptsize \begin{tabular}{ m{2.3cm}<{\centering}|m{1.2cm}<{\centering}| m{1.2cm}<{\centering}| m{1.2cm}<{\centering}|m{1.2cm}<{\centering}|m{1.2cm}<{\centering}|m{1.2cm}<{\centering}|m{1.2cm}<{\centering}} \toprule Metric & MobileNet V1-1.0\cite{howard2017mobilenets}& MobileNet V2-1.0\cite{sandler2018mobilenetv2} & Squeeze-Net\cite{iandola2016squeezenet} & Squeeze-Next-1.0-23\cite{gholami2018squeezenext} & ShuffleNet $1\times g = 8$\cite{zhang1707shufflenet} & Condense-Net\cite{huang2018condensenet} & \makecell{ MnasNet \\ \cite{tan2018mnasnet} }\\ \noalign{ \hrule height 2pt } Top-1 acc. &70.9 &71.8 &57.5 &59.0 &67.6 &71.0 &74.0 \\ \hline Top-5 acc. &89.9 &91.0 &80.3 & 82.3&- &90.0 &91.8 \\ \hline Input size &224$\times$224 &224$\times$224 &224$\times$224 &227$\times$227 & 224$\times$224&224$\times$224 &224$\times$224 \\ \noalign{ \hrule height 2pt } $\#$ of stacked CONV layers &27 & 20&26 &22 &17 &37 & 18\\ \hline Weights &3.24M &2.17M &1.25M &0.62M &3.9M &2.8M &3.9M \\ \hline Activations &5.2M & 1.46M&4.8M &4.7M &3.2M & 1.1M&3.9M \\ \hline MACs &568M &299M &388M &282M &138M &274M &317M\\ \noalign{ \hrule height 2pt } $\#$ of FC layers & 1 &1 &0 &1 &1 &1 &1 \\ \hline Weights &1M & 1.3M&0 & 0.1M& 1.5M& 0.1M&0.3M \\ \hline Activations &2K &2.3K &0 &1.1K & 2.5K&1.1K &1.3K \\ \hline MACs &1M &1.3M & 0&0.1M &1.5M &0.1M & 0.3M\\ \noalign{ \hrule height 2pt } Total weights & 4.24M&3.47M &1.25M & 0.72M&5.4M &2.9M &4.2M \\ \hline Total activations &5.2M &1.46M&4.8M &4.7M &3.2M &1.1M &3.9M \\ \hline Total MACs & 569M&300M &388M &282M &140M &274M &317M \\ \bottomrule \end{tabular} \normalsize \label{table:smallCNNcompare} \end{table}

1911.00623v2

\begin{table} \begin{threeparttable}[tb] \centering \tabcolsep=0.04cm \caption{The chronology of the recent approaches which modifies the training algorithm to account for quantization error.} \label{quanttable} \begin{scriptsize} %\begin{sc} \begin{tabular}{c|c|c|ccc|cc} \noalign{ \hrule height 2pt } \multirow{2}{*}{Year} &\multirow{2}{*}{Approach} & \multirow{2}{*}{Keywords} & \multicolumn{3}{c}{Quantization\tnote{1}} & \multicolumn{2}{c}{Benchmark} \\ \cline{4-6} \cline{7-8} & & & Forward & Backward & \specialcell{Parameter\\ Update} & Data & Model \\ \noalign{ \hrule height 2pt } 2014 & EBP \cite{soudry2014expectation} & Expectation Back Propagation & 1 bit, FP & - & - & used in \cite{crammer2013adaptive} & Proprietary MLP \\ \noalign{ \hrule height 2pt } \multirow{4}{*} {2015} & \multirow{3}{*} {Gupta et. al \cite{gupta2015deep} } & \multirow{2}{*} {Stochastic Rounding} & 16 bits & 16 bits & 16 bits & MNIST & Proprietary MLP , LeNet-5 \\ \cline{4-8} & & & 20 bits & 20 bits & 20 bits & CIFAR-10& used in \cite{hinton2012improving}\\ \cline{2-8} & Binary Connect \cite{courbariaux2015binaryconnect} & Stochastic Binarization & 1 bit & 1 bit & Float 32 \tnote{2} & \specialcell{MNIST \\ CIFAR-10\\SVHN} & Proprietary MLP, CNN \\ \noalign{ \hrule height 2pt } \multirow{5}{*}{2016} & Lin et. al \cite{lin2015neural} & \specialcell{Stochastic Binarization \\No forward pass multiplication\\ Quantized back propagation} & 1 bit & 1 bit & Float 32 & \specialcell{MNIST \\ CIFAR-10\\SVHN} & \specialcell{Proprietary \\ MLP, CNN} \\ \cline{2-8} & Bitwise Net \cite{kim2016bitwise} & \specialcell{Weight Compression\\ Noisy back propagation} & 1 bit & 1 bit & \specialcell{1 bit \\ Float 32\tnote{3}} & MNIST & Proprietary MLP\\ \cline{2-8} & XNOR-Net \cite{rastegari2016xnor} & \specialcell{Binary convolution\\Binary dot-product\\ Scaling binary gradient} & 1 bit & 1 bit & \specialcell{1 bit \\ Float 32\tnote{4}} & ImageNet & \specialcell{AlexNet \\ ResNet-18 \\ GoogLenet} \\ \cline{2-8} & \multirow{2}{*}{DoReFa-Net \cite{zhou2016dorefa}} & \multirow{2}{*}{\specialcell{stochastic gradient quantization \\ arbitrary bit-width}} &\multirow{2}{*}{1-8 bit} & \multirow{2}{*}{1-8 bit} & \multirow{2}{*}{2-32 bit} & SVHN & proprietary CNN \\ \cline{7-8} && & & && ImageNet & AlexNet \\ \noalign{ \hrule height 2pt } \multirow{4}{*}{2017} & \multirow{4}{*}{QNN \cite{hubara2017quantized}} & \multirow{4}{*}{\specialcell{Deterministic binarization \\ Straight through estimators \\ to avoid saturation \\ Shift based Batch Normalization \\ Shift based AdaMAX}} & \multirow{3}{*}{1 bit}& \multirow{3}{*}{1 bit} & \multirow{3}{*}{1 bit \tnote{5}} & & \\ & & & & & & MNIST & proprietary MLP \\ \cline{7-8} & & & & & & \specialcell{CiFAR-10\\SVHN} & CNN from \cite{courbariaux2015binaryconnect} \\ \cline{7-8} & & & & & & \specialcell{ImageNet} & \specialcell{AlexNet\\GoogLenet} \\ \cline{4-8} & & & 4 bit & 4 bit & 4 bit \tnote{6} & \specialcell{Penn \\ Treebank} & \specialcell{proprietary RNN\\LSTM} \\ \noalign{ \hrule height 2pt } \multirow{6}{*}{2018} & \multirow{3}{*}{Wang et. al \cite{wang2018training} } & \multirow{3}{*}{\specialcell{novel floating point\\ chunk based accumulation \\ stochastic rounding}} & \multirow{3}{*}{8 bit} & \multirow{3}{*}{8 bit} & \multirow{3}{*}{8 bit \tnote{7}} & CIFAR-10 & \specialcell{proprietary CNN\\ResNET} \\\cline{7-8} && & & && BN50 \cite{van2017training}& proprietary MLP \\ \cline{7-8} &&&&&& ImageNet & \specialcell{AlexNet\\ResNET18\\ResNET50} \\ \cline{2-8} & \multirow{3}{*}{Jacob et. al \cite{jacob2018quantization}} & \multirow{3}{*}{\specialcell{training with simulated\\quantization}}& \multirow{3}{*}{8 bit} & \multirow{3}{*}{8 bit} & \multirow{3}{*}{8 bit \tnote{8}} & Imagenet & \specialcell{Resnet\\Inception v3\\MobileNet} \\ \cline{7-8} && & & && COCO & MobileNet SSD \\ \cline{7-8} && & & && Flickr \cite{howard2017mobilenets} & MobileNet SSD \\ \noalign{ \hrule height 2pt } 2019 & WAGEUBN \cite{yang2020training} & \specialcell{batch-norm layer quantization \\8-bit integer representation \\combination of direct, constant \\and shift quantization} &8 bit & 8 bit & 8 bit & ImageNet & ResNet18/34/50 \\ \noalign{ \hrule height 2pt } \multirow{12}{*}{2020} &\multirow{4}{*}{S2FP8 \cite{cambier2020shifted}} & \multirow{4}{*}{\specialcell{shifted and squeezed FP8 \\ representation of tensors \\ tensor distribution learning }} &\multirow{4}{*}{8 bit} & \multirow{4}{*}{8 bit} & \multirow{4}{*}{32 bit} & CIFAR-10 & ResNet20/34/50 \\ \cline{7-8} && & & && ImageNet & ResNet18/50 \\ \cline{7-8} && & & && English-Vietnamese & Transformer-Tiny \\ \cline{7-8} && & & && MovieLens & \specialcell{Neural Collaborative\\ Filtering (NCF)} \\ \cline{2-8} & \multirow{3}{*}{Wiedemann et. al \cite{wiedemann2020dithered}} & \multirow{3}{*}{\specialcell{stochastic gradient quantization \\ induce sparsity \\ non-subtractive dither}} &\multirow{3}{*}{8 bit} & \multirow{3}{*}{8 bit} & \multirow{3}{*}{32 bit} & MNIST & LeNet \\ \cline{7-8} && & & && CIFAR-10/100 & \specialcell{AlexNet\\ResNet18\\VGG11} \\ \cline{7-8} && & & && ImageNet &ResNet18 \\ \cline{2-8} & \multirow{3}{*}{Quant-Noise \cite{fan2020training} }& \multirow{3}{*}{\specialcell{training using\\quantization noise}} & \multirow{3}{*}{8 bit} & \multirow{3}{*}{8 bit} & \multirow{3}{*}{8 bit} & Wikitext-103 & RoBERT \\ && & & && MNLI & RoBERT\\ && & & && ImageNet & EfficientNet-B3\\ \\ \noalign{ \hrule height 2pt } \end{tabular} %\end{sc} \begin{tablenotes} \item[1] minimum quantization for best performing model reported. \item[2] all real valued vectors are reported as Float 32 by default. \item[3] involves tuning a separate set of parameters with floating point precision. \item[4] becomes Float 32 if gradient scaling is used. \item[5] except the first layer input of 8 bits. \item[6] contains results with 2 bit, 3 bit and floating point precision. \item[7] additional 16 bit for accumulation. \item[8] uses 7 bit precision for some Inception v3 experiments. \end{tablenotes} %\normalsize \end{scriptsize} \end{threeparttable} \end{table}

1508.05417v2

\begin{table}[!b]\scriptsize \centering \caption{Simulation Parameters} \begin{tabular}{ l | l } \hline \hline Size of active region ($W \times L$) & $0.1 \times 5$ ($\mu m$) \\ \hline Temperature ($T$) & $298$ ($K$) \\ \hline Relative permittivity of SiO$_2$ layer ($\epsilon_{ox}/\epsilon_0$) & $3.9$ \\ \hline Thickness of SiO$_2$ layer ($t_{ox}$) & $17.5$ ($nm$) \\ \hline Effective mobility ($\mu_{eff}$) & $16 \times 10^{-3}$ ($m^2 V^{-1} s^{-1}$) \\ \hline Drain-source voltage ($V_{DS}$) & $0.1$ ($V$) \\ \hline Relative permittivity of solvent ($\epsilon_R/\epsilon_0$) & $78$ \\ \hline Ionic concentration of medium ($c_{ion}$) & $70$ ($mM$) \\ \hline % physiological conditions Trap density ($N_t$) & $2.3 \times 10^{24}$ ($eV^{-1} m^{-3}$) \\ \hline Tunneling distance ($\lambda$) & $0.05$ ($nm$) \\ \hline Average net charge of ligands ($N_e$) & $4$ \\ \hline % protein, DNA Length of receptor ($L_R$) & $4$ ($nm$) \\ \hline % range of aptamer and antibody sizes. Binding rate ($k_+$) & $2 \times 10^{-18}$ ($m^3 s^{-1}$) \\ \hline Unbinding rate ($k_-$) & $10$ ($s^{-1}$) \\ \hline Ligand concentration in reception space ($c_i$) & 4$K_D$ \\ \hline Concentration of receptors on the surface ($c_R$) & $2 \times 10^{16}$ ($m^{-2}$) \\ \hline Molecular capacitance ($C_{mol,L}, C_{mol,R}$) & $2 \times 10^{-20}$ ($F$) \\ \hline Capacitance of dielectric layer ($C_{dl}$) & $5 \times 10^{-2}$ ($F/m^2$)\\ \hline Capacitance of silicon ($C_s$) & $2 \times 10^{-3}$ ($F/m^2$) \\ \hline \end{tabular} \label{table:parameters} \end{table}

2105.00030v1

\begin{table}[t] \caption{Summary of curatorial actions in annotation schema} \centering \begin{tabularx}{\columnwidth}{|l|X|} \hline \textbf{Curatorial Action} & \textbf{Examples} \\ \hline \textit{Initial review and planning} & Look at deposited files, determine curation work needed, compose processing plan, create processing history syntax \\ \hline \textit{Data transformation} & Locate identifiers, revise or add variable/value labels, designate or fix missing values, reorder/standardize/convert variables, create variable-level metadata, collapse categories for disclosure \\ \hline \textit{Metadata} & Draft or revise study description, copy metadata from deposit system, update collection dates based on dataset, create survey question text, describe variable level labels \\ \hline \textit{Documentation} & Create a codebook, document major changes or issues with the data, compile documentation archived by the data producer \\ \hline \textit{Quality checks} & Check all files and metadata for completeness, adherence to standards, alignment with JIRA request after all data and documentation curation is complete (Self QC, 1QC, 2QC) \\ \hline \textit{Communication} & Discuss study with project manager, consult supervisor on curation standards for study, check how to handle specific variables \\ \hline \textit{Other} & Compile folders for study, ambiguous or overly-general curation work \\ \hline \textit{Non-curation} & Staff meetings, timesheets, administrative work \\ \hline \end{tabularx} \label{table:definitions} \end{table}

2105.00030v1

\begin{table}[t] \centering \caption{Description of Jira ticket corpus of curation requests} \begin{tabularx}{\columnwidth}{|X|X|X|X|X|X|} \hline & & \textbf{Total \newline tickets \newline (n=669)} & \textbf{Total \newline studies \newline (n=566)} & \textbf{Average curation hours/study} \\ \hline \multirow{3}{*}{\textbf{Curation}} & Level 1 & 221 & 178 & 51 \\ & Level 2 & 229 & 210 & 79 \\ & Level 3 & 219 & 178 & 165 \\ \hline \multirow{3}{*}{\textbf{Archive}} & BJS & 131 & 124 & 78 \\ & ICPSR & 116 & 104 & 105 \\ & Other & 422 & 338 & 102 \\ \hline \multirow{3}{*}{\textbf{Year}} & 2017 & 133 & 119 & 107 \\ & 2018 & 305 & 276 & 99 \\ & 2019 & 231 & 171 & 88 \\ \hline \end{tabularx} \label{table:tickets} \end{table}

2105.00030v1

\begin{table}[t] \centering \caption{Studies recording curation actions and percent of hours logged across all studies} \begin{tabularx}{\columnwidth}{|l|X|X|} \hline \textbf{Action} & \textbf{Percent of studies containing action} & \textbf{Percent of total work log hours classified as action} \\ \hline \textit{Quality checks} & 90.1 & 31.6 \\ \hline \textit{Initial review and planning} & 70.0 & 14.0 \\ \hline \textit{Data transformation} & 67.6 & 29.9 \\ \hline \textit{Metadata} & 57.7 & 6.5 \\ \hline \textit{Documentation} & 56.2 & 7.5 \\ \hline \textit{Communication} & 54.6 & 7.9 \\ \hline \textit{Other} & 40.9 & 2.8 \\ \hline \end{tabularx} \label{table:actions} \end{table}

1704.06497v2

\begin{table}[t] \begin{center} \resizebox{0.9\columnwidth}{!}{ \begin{tabular}{ll|lll} \toprule \bf Domain &\bf Version &\bf Train &\bf Valid. &\bf Test\\ \midrule Europarl & v.5 & 1.6M & 2k & 2k\\ News Commentary & WMT07 & 40k & 1k & 2k\\ %nc-dev2007 nc-devtest2007 TED & TED2013 & 153k & 2k & 2k\\ \bottomrule \end{tabular} } \end{center} \caption{Number of parallel sentences for training, validation and test sets for French-to-English domain adaptation.} \label{tab:data} \end{table}

2312.01024v1

\begin{table}[ht] \caption{Hybrid Model Architecture and Parameters} \centering \begin{tabular}{|l|l|l|} \hline \textbf{Layer (type)} & \textbf{Output Shape} & \textbf{Param \#} \\ \hline Conv2d-1 & [-1, 64, 64, 64] & 9,408 \\ BatchNorm2d-2 & [-1, 64, 64, 64] & 128 \\ ReLU-3 & [-1, 64, 64, 64] & 0 \\ MaxPool2d-4 & [-1, 64, 32, 32] & 0 \\ Conv2d-5 & [-1, 128, 32, 32] & 8,192 \\ BatchNorm2d-6 & [-1, 128, 32, 32] & 256 \\ ReLU-7 & [-1, 128, 32, 32] & 0 \\ Conv2d-8 & [-1, 128, 32, 32] & 4,608 \\ BatchNorm2d-9 & [-1, 128, 32, 32] & 256 \\ ... & ... & ... \\ Linear-100 & [-1, 1] & 513 \\ TorchConnector-101 & [-1, 2] & 2 \\ \hline \textbf{Total params} & & 1,412,931 \\ \textbf{Trainable params} & & 1,412,931 \\ \textbf{Non-trainable params} & & 0 \\ \hline \end{tabular} \end{table}

1612.04858v1

\begin{table}[H] \begin{center} \begin{tabular}{ |>{\centering}m{1.7cm}|>{\centering}m{2.5cm}|>{\centering}m{2cm} |>{\centering}m{1.8cm}|>{\centering}m{2.0cm} | } \hline & SigOpt & Rnd. Search & Grid Search & \hspace{0.5mm} No Tuning \newline (Baseline) \tabularnewline \hline Best Found \newline ACC & \bf{0.8760} ({\color{ForestGreen}{+5.72\%}}) & 0.8673 & 0.8680 & 0.8286 \tabularnewline \hline \end{tabular} \vspace{4mm} \caption{Best found accuracy results averaged over 20 optimization runs, each run consisting of 60 function evaluations} \end{center} \end{table}

1612.04858v1

\begin{table}[H] \begin{center} \begin{tabular}{ |>{\centering}m{1.4cm}|>{\centering}m{2.5cm}|>{\centering}m{2.0cm} |>{\centering}m{1.9cm}|>{\centering}m{1.9cm} |>{\centering}m{1.9cm} | } \hline & \hspace{4mm} SigOpt \newline (xgboost + \newline Unsup. Feats) & Rnd Search \newline (xgboost + \newline Unsup. Feats) & SigOpt \newline (xgboost + \newline Raw Feats) & Rnd Search \newline (xgboost + \newline Raw Feats) & No Tuning \newline (sklearn RF + \newline Raw Feats) \tabularnewline \hline Hold out \newline ACC & \bf{0.8601} ({\color{ForestGreen}{+49.2\%}}) & 0.8190 & 0.7483 & 0.7386 & 0.5756 \tabularnewline \hline \end{tabular} \vspace{4mm} \caption{Comparison of model accuracy on held out (test) dataset after different tuning strategies} \end{center} \end{table}

1612.04858v1

\begin{table}[H] \begin{center} \begin{tabular}{ |>{\centering}m{1.25cm}|>{\centering}m{2.9cm}|>{\centering}m{3cm} |>{\centering}m{1.8cm}|>{\centering}m{2.9cm} | } \hline & \hspace{4mm} SigOpt \newline (TensorFlow CNN) & Random Search \newline (TensorFlow CNN) & No Tuning \newline (sklearn RF) & \hspace{4mm} No Tuning \newline (TensorFlow CNN) \tabularnewline \hline Hold out \newline ACC & \bf{0.8130} ({\color{ForestGreen}{+315.2\%}}) & 0.5690 & 0.5278 & 0.1958 \tabularnewline \hline \end{tabular} \vspace{4mm} \caption{Comparison of model accuracy on the held out (test) dataset after different tuning strategies} \end{center} \end{table}

1612.04858v1

\begin{table}[H] \begin{center} \begin{tabular}{ |>{\centering}m{1.5cm}|>{\centering}m{2.5cm} |>{\centering}m{2cm} |>{\centering}m{3.5cm}| } \hline & SigOpt & Random Search & \hspace{3mm} No Tuning \newline (Default MLlib ALS) \tabularnewline \hline Hold out \newline RMSE & \bf{0.7864} ({\color{ForestGreen}{-40.7\%}}) & 0.7901 & 1.3263 \tabularnewline \hline \end{tabular} \vspace{4mm} \caption{Comparison of RMSE on the hold out (test) ratings after tuning ALS algorithm} \end{center} \end{table}

2004.14107v1

\begin{table}[tb] \centering \begin{tabular}{c|c|c|c|c} Information / Setting & Random & SDR & SDRT & SDRTS \\ \hline \small{Starting/Dest. Areas} & $\checkmark$ & $\checkmark$ & $\checkmark$ & $\checkmark$ \\ \hline \small{Exact Starting/Dest. Positions} & $\times$ & $\checkmark$ & $\checkmark$ & $\checkmark$ \\ \hline \small{Trajectory Entry Timing} & $\times$ & $\times$ & $\checkmark$ & $\checkmark$ \\ \hline \small{Trajectory Average Speed} & $\times$ & $\times$ & $\times$ & $\checkmark$\\ \hline \end{tabular} \caption{Different simulation settings and the information provided.} \label{tab:simSettings} \vspace{-1em} \end{table}

2004.14107v1

\begin{table}[tb] \centering \begin{tabular}{c|c|c|c|c} \hline Metric/Simulations & SDR & SDRT & SDRTS & Ours\\ \hline DPD-Space & 0.4751 & 0.3813 & 0.4374 & {\bf0.2988} \\ \hline DPD-Time & 0.3545 & 0.0795 & 0.064 & {\bf0.0419} \\ \hline DPD-TS & 1.0 & 0.8879 & 1.0 & {\bf 0.4443} \\ \hline \hline \hline DPD-Space &0.2753 &0.2461 &0.2423 & {\bf 0.1173} \\ \hline DPD-Time & 0.0428 & 0.0319 & 0.0295 & {\bf 0.0213} \\ \hline DPD-TS &0.9970 &0.8157 &0.9724 & {\bf 0.5091} \\ \hline \end{tabular} \caption{Comparison on space flow P2 in Forum (Top) and space flow P1 in TrainStation (Bottom) based on DPD metrics, both shown in \figref{std_vis}. {\bf Lower} is better.} \label{tab:DPD_trainStation} \end{table}

2305.19573v1

\begin{table}[t] \centering \setlength{\tabcolsep}{11pt} \begin{tabular}{|m{0.7cm}|c|c|c|} \hline & Within ($d_1$) & Between ($d_2$)\\ \hline $d_J$ & $0.0784 \pm 0.0194$ & $0.1027 \pm 0.0232$\\ \hline $d_\text{H}$ & $0.3145 \pm 0.4864$ & $0.3623 \pm 0.4967$\\ \hline $d_{\rm basin}$ & $0.0309 \pm 0.0314$ & $0.0386 \pm 0.0325$\\ \hline $d_L$ & $0.2535 \pm 0.1619$ & $0.2921 \pm 0.1783$\\ \hline \end{tabular} \caption{Discrepancy between two energy landscapes estimated by the conventional likelihood maximization method applied to the HCP data. ``Within'' and ``Between'' in the table stand for within-participant and between-participant, respectively.} \label{table:HCP_results} \end{table}

2401.16220v1

\begin{table}[h] \centering \begin{tabular}{|c||c|c|c|}\hline $ t$&$L_t$&$P_t$&$A_t $ \\\hline 0& 107& 73& 214\\\hline 1& 33&86 &240 \\\hline 2& 67& 27& 267\\\hline \end{tabular} \caption{Population data for Section~\ref{ex:workedexample}} \label{table:workedexample} \end{table}

2401.16220v1

\begin{table}[h] \centering \begin{tabular}{|c||c|c|c|c|}\hline t &$R_{L_t}$&$\alpha_t$&$R_{A_t}$&$\beta_t$ \\\hline 1&$[-3.67,-2.67]$& $-3.17$&$[-1.48,-0.77]$& $-1.13$\\ \hline 2&$[-4.50,-3.28]$& $-3.89$&$[-1.65,-0.86] $&$-1.25$\\\hline \end{tabular} \caption{Expansion ranges and midpoints} \label{table:exampleranges} \end{table}

2401.16220v1

\begin{table}[h] \centering \begin{tabular}{|c||c|c||c|}\hline t & $\hat{T}_{L_t}$ & $\hat{T}_{A_t}$\\\hline 1 & $0.18 + 0.04\tau_1 + 0.02(\tau_1+ 3.17)^2$ & $0.69 + 0.32\tau_2 + 0.16(\tau_2 + 1.13)^2$\\\hline 2 &$0.10 + 0.02\tau_1 + 0.01(\tau_1 + 3.89)^2$ &$0.64 + 0.29\tau_2 + 0.14(\tau_2 + 1.25)^2$ \\\hline \end{tabular} \caption{Taylor polynomials in $\tau_1,\tau_2$} \label{table:examplepolynosubs} \end{table}

2401.16220v1

\begin{table} \centering \begin{tabular}{|c|c|c|c|}\hline $t$& $L_t$ &$P_t$&$A_t$\\\hline 3& 36 & 54&273\\\hline\hline $t$&$R_{L_t}$&-&$R_{A_t}$\\\hline 3&$[-4.16,-3.02]$&-&$[-1.69,-0.87]$\\\hline &$\alpha=-3.59$&-&$\beta=-1.28$\\\hline \end{tabular} \caption{ Prolongation data and results from Algorithm~\ref{alg:expansionrange}} \label{tab:continuedexampledata} \end{table}

1903.09030v1

\begin{table}[!t] \scriptsize \caption{A comparison with the ladder network. We represent error percentage.} \centering \begin{tabular}{c|ccc} \hline \textbf{Labeled Samples} & \textbf{10} & \textbf{100} & \textbf{1000} \\\hline \textbf{Baseline} & 58.88 & 28.39 & 7.25 \\ \textbf{Ladder Network} & 48.85 & 24.74 & 6.96 \\ \textbf{RBM DA} & \textbf{45.34 } & \textbf{ 18.66} &\textbf{ 5.60 } \end{tabular} \label{tab6} \end{table}

1411.4911v5

\begin{table}[htbp] \centering \caption{Factor coordinates of the variables obtained with \code{MFAmix}} {\small \label{coord_var} \begin{tabular}{l|r|r} \hline & dim 1 & dim 2\\ \hline farmers & -0.45 & -0.30\\ \hline tradesmen & -0.14 & 0.12\\ \hline \textbf{managers} & 0.31 &\textbf{ 0.55}\\ \hline workers & -0.13 & -0.04\\ \hline unemployed & 0.32 & -0.08\\ \hline \textbf{middleempl} & 0.24 & \textbf{0.60}\\ \hline retired & -0.03 & -0.44\\ \hline employrate & -0.33 & 0.55\\ \hline \textbf{income} & 0.13 & \textbf{0.60}\\ \hline \textbf{density} & \textbf{0.72} & -0.15\\ \hline primaryres & 0.03 & 0.36\\ \hline \textbf{owners} & \textbf{-0.69} & 0.41\\ \hline \textbf{building} & \textbf{0.72} & -0.21\\ \hline water & 0.19 & -0.20\\ \hline \textbf{vegetation} & 0.08 & 0.56\\ \hline \textbf{agricul }& \textbf{-0.54} & -0.47\\ \hline \end{tabular} } \end{table}

2007.06775v3

\begin{table}[!h] \small \centering \ra{0.95} \begin{tabular}{!{\VRule[1pt]}M{0.09\textwidth}!{\VRule[1pt]}m{0.3\textwidth}!{\VRule[1pt]}m{0.04\textwidth}!{\VRule[1pt]}} \specialrule{1.2pt}{0pt}{0pt} \rowcolor{white} %\begin{tabular}{@{}p{7.5cm}c@{}} % \toprule[1.2pt] % & \multicolumn{3}{c}{\% dataset cached}\\ %\% dataset cached & 8-GPU training & \multicolumn{2}{c}{8-job HP search}\\ % (Size : 146GB) & Cache Miss & Disk IO (GB) & Read amp \\ % \midrule \vheading{ Fetch Stalls (Remote)} & Is remote storage a bottleneck for training? & \sref{sec-fetch-remote}\\ \specialrule{0.5pt}{0pt}{0pt} \vheading{Fetch Stalls (Local)} & When does the local storage device (SSD/HDD) become a bottleneck for DNN training? & \sref{sec-fetch}\\ \specialrule{0.5pt}{0pt}{0pt} \vheading{Prep Stalls} & When does data prep at the CPU become a bottleneck for DNN training? & \sref{sec-prep} \\ \specialrule{0.5pt}{0pt}{0pt} \vheading{Generality} & Do fetch and prep stalls exist in other training platforms like TensorFlow? & \sref{sec-tf}\\ \specialrule{1.2pt}{0pt}{0pt} %\bottomrule[1.2pt] \end{tabular} %\vspace{-1em} % \mycaption{Data stalls in Tensorflow}{The fundamental problems that result in data stalls-inefficient caching and thrashing due to lack of coordination in HP search, exist in TF.} \label{tbl-analysis-q} \vspace{-1.5em} \end{table}

2003.04735v2

\begin{table}[http]{ \begin{tabular}{cc} \hline \multicolumn{2}{c}{Summary of Notations} \\ \hline \multicolumn{1}{c|}{$\mathcal{V}$, $v$, $\mathcal{B}_v$} & Set of Nodes, Node $v$, Set of Neighboring Nodes of Node $v$ \\ \multicolumn{1}{c|}{$\mathbf{w}_v$, $b_v$, $\mathbf{r}_v$} & Decision Variables at Node $v$ \\ \multicolumn{1}{c|}{$\mathbf{x}_{vn}$, $y_{vn}$} & $n$-th Data and Label at Node $v$ \\ \multicolumn{1}{c|}{$\mathbf{X}_v$, $\mathbf{Y}_v$} & Data Matrix and Label Matrix at Node $v$ \\ \multicolumn{1}{c|}{$\omega_{vu}$} & Consensus Variable between Node $v$ and Node $u$ \\ \multicolumn{1}{c|}{$\theta_v$} & Indicator Vector of Flipped Labels at Node $v$ \\ \multicolumn{1}{c|}{$\delta_{vn}$} & Vector of Data Poisoning on the $n$-th Data at Node $v$ \\ \hline \end{tabular}} \end{table}

2003.04735v2

\begin{table} \caption{Average equilibrium classification risks $(\%)$ of DSVM using Spambase dataset \cite{Spambase} in Network 1 and Network 2. } \label{tab:LabelNetwork12} \begin{center} \begin{small} \begin{sc} \begin{tabular}{|c|c|c|c|c|c|c|} \hline Net & 1 & 1L & 1D & 2 & 2L & 2D \\ \hline Risk & 11.6 & 32.3 & 42.2 & 10.6 & 29.4& 39.3 \\ \hline STD & 1.6 & 0.6 & 2.6 & 0.6 & 0.3 & 1.1 \\ \hline \end{tabular} \end{sc} \end{small} \end{center} \vskip -0.1in \end{table}

2003.04735v2

\begin{table} \caption{Average equilibrium classification risks $(\%)$ of DSVM using Spambase dataset \cite{Spambase} in Network 3 and Network 4. } \label{tab:LabelNetwork34} \begin{center} \begin{small} \begin{sc} \begin{tabular}{|c|c|c|c|c|c|c|c|c|} \hline Net & 3 & 3La & 3Lb & 3Da & 3Db & 4 & 4L & 4D \\ \hline Risk & 11.7 & 29.5 & 26.9 & 36.4 & 34.6 & 13.5 & 35.0 & 47.0\\ \hline STD & 1.5 & 0.6 & 1.2 & 0.9 & 0.8 & 1.8 & 0.9 & 2.5\\ \hline \end{tabular} \end{sc} \end{small} \end{center} \vskip -0.1in \end{table}

2210.00770v1

\begin{table}[H] \scriptsize \caption{Comparison Between Agents Trained With and Without PID Controller Coaching. Even though the PID controllers are less capable than the eventual RL agent, they are still useful and can accelerate the RL agent training. There two measures we used to gauge training acceleration. The first is five consecutive wins, and the second is the scoring average. The "win" is a predetermined benchmark. } \label{episode_compare} \centering \begin{tabular}{ cccccc } \rowcolor{airforceblue} Environment & Target & Measure & With PID & Without & Percentage\\ \rowcolor{airforceblue} Name & Score & & Coaching & Coaching & Increase \\ \hline Inverted & 800& Win Streak & 100 & 160& 37.5\% \\ Pendulum & &Average & 104 & 159& 34.6\%\\ \rowcolor{beaublue} Double & 5500& 5 Wins & 908 & 1335& 31.9\%\\ \rowcolor{beaublue} Pendulum & &Average & 935 & 1370& 29.9\%\\ Hopper & 800& 5 Wins & 2073 & 2851& 27.3\%\\ & &Average & 2155 & 2911& 25.9\%\\ \rowcolor{beaublue} Walker & 800& 5 Wins & 4784 & 5170& 7.5\%\\ \rowcolor{beaublue} & &Average & 5659 & 7135& 20.7\%\\ \end{tabular} \end{table}

2202.01947v1

\begin{table} \tbl{Response patterns and sample sizes for ADNI data.} {\begin{tabular}{c|c|cccc|c} \hline & \multicolumn{5}{c|}{Data source} & \\ \hline $k$ & MMSE & CSF & PET & MRI & GENE & Sample size \\ \hline 1 & $\ast$ & $\ast$ & $\ast$ & $\ast$ & $\ast$ & 409 \\ 2 & $\ast$ & $\ast$ & $\ast$ & $\ast$ & & 368 \\ 3 & $\ast$ & $\ast$ & $\ast$ & & $\ast$ & 40 \\ 4 & $\ast$ & & $\ast$ & $\ast$ & $\ast$ & 105 \\ 5 & $\ast$ & & $\ast$ & & $\ast$ & 86 \\ 6 & $\ast$ & & $\ast$ & $\ast$ & & 53 \\ 7 & $\ast$ & & & & $\ast$ & 53 \\ 8 & $\ast$ & & & $\ast$ & & 56 \\ \hline & & & & & Total & 1170 \\ \hline \end{tabular}} \tabnote{{$\ast$}the datum is available.} \label{table1}% \end{table}

2101.00523v2

\begin{table}[] \centering \begin{tabular}{|c|c|} \hline \textbf{Parameter} & \textbf{baseline value}\tabularnewline \hline $D$ & 0.00035 \\ \hline $\chi$ & 0.38 \\ \hline $\rho$ & 0.34\\ \hline $\beta$ & 0.05 \\ \hline $\gamma$ & 0.1 \\ \hline $\epsilon_1$ & 0.45 \\ \hline $\epsilon_2$ & 0.45 \\ \hline $k$ & 0.75 \\ \hline \end{tabular} \vspace{0.1cm} \textbf{S2 Table. Anderson-Chaplain model nondimensionalized parameters.} Baseline nondimensionalized mechanistic parameters for the Chaplain-Anderson Model from \cite{anderson_continuous_1998}. \label{tab:SItab2} \end{table}

2103.11233v4

\begin{table}[htbp] \caption{Signals' details and summary of parameters} \centering \scalebox{0.8}{\begin{tabular}{|| c | c | c | c | c ||} \hline Labels & Samples & $(L,a,b)$ & $x_0$ & $\mu_i$, $i=1,2,3,*$\\ \hline\hline Cusp & 33 & $(33,1,11)$ & zero vector & $\|\Phi_ix\|_\infty$\\ \hline Ramp & 33 & $(33,1,11)$ & zero vector & $\|\Phi_ix\|_\infty$\\ \hline Sing & 45 & $(45,1,9)$ & zero vector & $\|\Phi_ix\|_\infty$\\ \hline SI1899 & 22938 & $(20349,19,21)$ & $A^TAx$ & $10^{-1}\|\Phi_ix\|_\infty$\\ \hline SI1948 & 27680 & $(27531,19,23)$ & $A^TAx$ & $10^{-1}\|\Phi_ix\|_\infty$\\ \hline SI2141 & 42800 & $(41769,21,17)$ & $A^TAx$ & $10^{-1}\|\Phi_ix\|_\infty$\\ \hline SX5 & 24167 & $(23205,17,13)$ & $A^TAx$ & $10^{-1}\|\Phi_ix\|_\infty$\\ \hline SX224 & 25805 & $(24633,23,21)$ & $A^TAx$ & $10^{-1}\|\Phi_ix\|_\infty$\\ \hline SI1716 & 25908 & $(24633,23,21)$ & $A^TAx$ & $\|\Phi_ix\|_\infty$\\ \hline \end{tabular}} \label{sigdec} \end{table}

2401.06452v1

\begin{table}[] \caption{Hyperparameters of the BO-Auto-PU system, with their default values.}\label{tab2} \centering \small \begin{tabular}{|l|l|} \hline Hyperparameter & Value \\\hline \(It\_count\) & 50 \\\hline \(\#Configs\) & 101 \\\hline \(Surr\_model\) & Random Forest Regressor \\\hline \(Acquisition function\) & \(Surr\_model\) predicted value \\ \hline \end{tabular} \end{table}

2401.06452v1

\begin{table}[] \caption{Hyperparameters of the EBO-Auto-PU system, with their default values.}\label{tab3} \centering \small \begin{tabular}{|l|l|} \hline Hyperparameter & Value \\ \hline \(\#Configs\) & 101 \\ \hline \(It\_count\) & 50 \\ \hline \(Surr\_model\) & Random Forest Regressor \\ \hline Crossover probability & 0.9 \\ \hline Component crossover probability & 0.5 \\ \hline Mutation probability & 0.1 \\ \hline Tournament size & 2 \\ \hline \(k\) & 10 \\ \hline \end{tabular} \end{table}

2401.06452v1

\begin{table}[htbp] \caption{Linear (Pearson’s) correlation coefficient value between the F-measure and the percentage of positive examples in the original dataset (before hiding some positive examples in the unlabelled set) for each combination of an Auto-PU system (with base search space) or PU method and a \(\delta\) value}\label{tab6} \centering \small \begin{tabular}{|l|c|c|c|} \hline \begin{tabular}[c]{@{}l@{}}Auto-PU system \\ or PU method\end{tabular} & \(\delta\) = 20\% & \(\delta\) = 40\% & \(\delta\) = 60\% \\ \hline EBO-1 & 0.440 & 0.469 & 0.460 \\ \hline BO-1 & 0.398 & 0.360 & 0.498 \\ \hline GA-1 & 0.333 & 0.385 & 0.504 \\ \hline DF-PU & 0.381 & 0.504 & 0.445 \\ \hline S-EM & 0.690 & 0.631 & 0.686 \\ \hline \end{tabular} \end{table}

2401.06452v1

\begin{table}[htbp] \caption{Linear (Pearson’s) correlation coefficient value between the F-measure and the percentage of positive examples in the original dataset (before hiding some positive examples in the unlabelled set) for each combination of an Auto-PU system (with extended search space) or PU method and a \(\delta\) value}\label{tab11} \centering \small \begin{tabular}{|l|l|l|l|} \hline \multicolumn{1}{|c|}{Method} & \multicolumn{1}{c|}{\(\delta\) = 20\%} & \multicolumn{1}{c|}{\(\delta\) = 40\%} & \multicolumn{1}{c|}{\(\delta\) = 60\%} \\ \hline EBO-2 & 0.363 & 0.361 & 0.447 \\ \hline BO-2 & 0.339 & 0.348 & 0.225 \\ \hline GA-2 & 0.340 & 0.357 & 0.580 \\ \hline DF-PU & 0.381 & 0.504 & 0.445 \\ \hline S-EM & 0.690 & 0.631 & 0.686 \\ \hline \end{tabular} \end{table}

2401.06452v1

\begin{table}[] \caption{Hyperparameter values most frequently selected by BO-Auto-PU}\label{tab14} \centering \small \begin{tabular}{|l|l|l|ll|ll|ll|} \hline \textbf{Hyperparameter} & \textbf{\begin{tabular}[c]{@{}l@{}}Search \\ space\end{tabular}} & \textbf{\begin{tabular}[c]{@{}l@{}}Most selected \\ value\end{tabular}} & \multicolumn{2}{l|}{\textbf{\begin{tabular}[c]{@{}l@{}}Selection\\ Freq. (\%)\end{tabular}}} & \multicolumn{2}{l|}{\textbf{\begin{tabular}[c]{@{}l@{}}Baseline \\ Freq. (\%)\end{tabular}}} & \multicolumn{2}{l|}{\textbf{\begin{tabular}[c]{@{}l@{}}Diff.\\ (\%)\end{tabular}}} \\ \hline \multirow{2}{*}{Phase 1A Iteration Count} & base & 2 & \multicolumn{2}{l|}{19.00} & \multicolumn{2}{l|}{10.00} & \multicolumn{2}{l|}{9.00} \\ \cline{2-9} & extended & 2 & \multicolumn{2}{l|}{21.00} & \multicolumn{2}{l|}{10.00} & \multicolumn{2}{l|}{11.00} \\ \hline \multirow{2}{*}{Phase 1A RN Threshold} & base & 0.05 & \multicolumn{2}{l|}{14.33} & \multicolumn{2}{l|}{10.00} & \multicolumn{2}{l|}{4.33} \\ \cline{2-9} & extended & 0.25 & \multicolumn{2}{l|}{13.00} & \multicolumn{2}{l|}{10.00} & \multicolumn{2}{l|}{3.00} \\ \hline \multirow{2}{*}{Phase 1A Classifier} & base & Bernoulli NB & \multicolumn{2}{l|}{8.67} & \multicolumn{2}{l|}{5.56} & \multicolumn{2}{l|}{3.11} \\ \cline{2-9} & extended & Logistic reg. & \multicolumn{2}{l|}{8.67} & \multicolumn{2}{l|}{5.56} & \multicolumn{2}{l|}{3.11} \\ \hline \multirow{2}{*}{Phase 1B Flag} & base & True & \multicolumn{2}{l|}{52.67} & \multicolumn{2}{l|}{50.00} & \multicolumn{2}{l|}{2.67} \\ \cline{2-9} & extended & True & \multicolumn{2}{l|}{50.67} & \multicolumn{2}{l|}{50.00} & \multicolumn{2}{l|}{0.67} \\ \hline \multirow{2}{*}{Phase 1B RN Threshold} & base & 0.2 & \multicolumn{2}{l|}{14.00} & \multicolumn{2}{l|}{10.00} & \multicolumn{2}{l|}{4.00} \\ \cline{2-9} & extended & 0.2 & \multicolumn{2}{l|}{14.67} & \multicolumn{2}{l|}{10.00} & \multicolumn{2}{l|}{4.67} \\ \hline \multirow{2}{*}{Phase 1B Classifier} & base & HGBoost & \multicolumn{2}{l|}{8.00} & \multicolumn{2}{l|}{5.56} & \multicolumn{2}{l|}{2.44} \\ \cline{2-9} & extended & Bagging clas. & \multicolumn{2}{l|}{7.67} & \multicolumn{2}{l|}{5.56} & \multicolumn{2}{l|}{2.11} \\ \hline \multirow{2}{*}{Spy rate} & base & N/A & \multicolumn{2}{l|}{N/A} & \multicolumn{2}{l|}{N/A} & \multicolumn{2}{l|}{N/A} \\ \cline{2-9} & extended & 0.3 & \multicolumn{2}{l|}{18.00} & \multicolumn{2}{l|}{14.29} & \multicolumn{2}{l|}{3.71} \\ \hline \multirow{2}{*}{Spy tolerance} & base & N/A & \multicolumn{2}{l|}{N/A} & \multicolumn{2}{l|}{N/A} & \multicolumn{2}{l|}{N/A} \\ \cline{2-9} & extended & 0.08 & \multicolumn{2}{l|}{12.18} & \multicolumn{2}{l|}{9.09} & \multicolumn{2}{l|}{3.09} \\ \hline \multirow{2}{*}{Spy flag} & base & N/A & \multicolumn{2}{l|}{N/A} & \multicolumn{2}{l|}{N/A} & \multicolumn{2}{l|}{N/A} \\ \cline{2-9} & extended & False & \multicolumn{2}{l|}{74.00} & \multicolumn{2}{l|}{50.00} & \multicolumn{2}{l|}{24.00} \\ \hline \multirow{2}{*}{Phase 2 Classifier} & base & LDA & \multicolumn{2}{l|}{32.67} & \multicolumn{2}{l|}{5.56} & \multicolumn{2}{l|}{27.11} \\ \cline{2-9} & extended & LDA & \multicolumn{2}{l|}{51.67} & \multicolumn{2}{l|}{5.56} & \multicolumn{2}{l|}{46.11} \\ \hline \end{tabular} \end{table}

2401.06452v1

\begin{table}[] \centering \caption{Pearson’s correlation coefficient values for Phase 1A iteration count to class imbalance. }\label{tab16} \centering \small \begin{tabular}{|l|l|l|l|} \hline \multicolumn{1}{|c|}{\textbf{Method}} & \multicolumn{1}{c|}{\textbf{\(\delta\) = 20\%}} & \multicolumn{1}{c|}{\textbf{\(\delta\) = 40\%}} & \multicolumn{1}{c|}{\textbf{\(\delta\) = 60\%}} \\ \hline GA-1 & -0.646 & -0.655 & -0.689 \\ \hline GA-2 & -0.631 & -0.687 & -0.723 \\ \hline BO-1 & -0.677 & -0.700 & -0.700 \\ \hline BO-2 & -0.641 & -0.706 & -0.736 \\ \hline EBO-1 & -0.656 & -0.688 & -0.696 \\ \hline EBO-2 & -0.680 & -0.710 & -0.687 \\ \hline \end{tabular} \end{table}

1701.02440v1

\begin{table}[h] \centering \label{table_results} \begin{tabular}{|c|c|c|} \hline & Decay & Diff.\\ Gene & ($\lambda^a$) & ($D^a$) \\ \hline Hb & 0.1606 & 0.3669 \\ Kr & 0.0797 & 0.4490 \\ Gt & 0.1084 & 0.4543 \\ Kni & 0.0807 & 0.2683 \\ \hline \end{tabular} \caption{Inferred parameter values for the decay $\lambda^a$ and diffusion $D^a$ rates of protein $a$.} \end{table}

2008.10893v1

\begin{table}[!ht] \begin{center} \resizebox{\textwidth}{!}{ \begin{tabular}{ |c|ll|ll|ll|ll|} \hline & $\abs{y_\mathcal{N}-y^*_h}_1$ & $\abs{y_\mathcal{N}-y^*_h}_1$ & $\abs{y_\mathcal{N}-y^*}_1$ & $\abs{y_\mathcal{N}-y^*}_1$ & $\norm{y_\mathcal{N}-y^*_h}_0$ & $\norm{y_\mathcal{N}-y^*_h}_0$ &$\norm{y_\mathcal{N}-y^*}_0$ & $\norm{y_\mathcal{N}-y^*}_0$ \\ \hline & min & max & min & max & min & max & min & max \\ \hline 1-L & $0.2506 $ & $0.6532 $ & $ 0.2868 $ & $ 0.6713 $ & $0.0752 $ & $ 0.2422 $ & $0.0808 $ & $0.2435 $ \\ \hline 3-L & $ 0.2575 $ & $0.7537 $ & $ 0.2391 $ & $0.7777 $ & $ 0.0817 $ & $0.2524 $ & $0.0791 $ & $0.2565 $ \\ \hline 5-L & $0.2157 $ & $36.2640 $ & $ 0.2235 $ & $ 36.2731 $ & $0.0539 $ & $29.4926 $ & $ 0.0544 $ & $ 29.4936 $ \\ \hline & mean& deviation & mean & deviation & mean& deviation & mean& deviation \\ \hline 1-L & $0.4276 $ & $0.1099 $ & $ 0.4496 $ & $ 0.1075 $ & $0.1472 $ & $ 0.0484 $ & $0.1506 $ & $0.0485 $ \\ \hline 3-L & $ 0.3853 $ & $0.1350 $ & $ 0.4003 $ & $0.1687 $ & $ 0.1425 $ & $0.0462 $ & $0.1268 $ & $0.0482 $ \\ \hline 5-L & $3.0242 $ & $ 8.9087 $ & $3.0287 $ & $8.9103 $ & $ 2.1309 $ & $7.3143 $ & $ 2.1299 $ & $ 7.3149 $ \\ \hline \end{tabular}} {\small \caption{\label{tab:layer_comparison}Statistics on learning-informed PDEs with different layers in neural networks using small size training data, small DoF in $\Theta$, and 15 samples in total.}} \end{center} \end{table}

2310.14848v1

\begin{table}[htbp] \centering \caption{Efficiency} \label{E&F} \begin{tabular}{|c|c|c|} \hline \thead{Computation} & \thead{Scheme} & \thead{Efficiency\\(P/V/C)} \\ \hline \multirow{2}{*}{\thead{MatrixMult\\$O(n^3)$}} & \thead{SafetyNets} & \thead{$O(n^2)$ / $O(n^2)$/ $O(\log n)$} \\ \cline{2-3} % \multicolumn{1}{|c|}{} & \multicolumn{1}{c|}{} & & \thead{Mystique} & $\CIRCLE$ & $\CIRCLE$ & $\CIRCLE$ & \thead{extended\\privacy} & $O(n^2)$ \\ \cline{4-9} % \\(hybrid\\commitment) & \thead{Mystique,\\Fan,\\zkMLaaS} & $O(n^2)/O(n^2)/O(1)$ \\ \cline{1-3} % \\(threshold\\cryptosystem\\\& noise) \multirow{5}{*}{\thead{Convolution\\$O(n^2w^2)$}} & \thead{SafeTPU} & -- \\ \cline{2-3} & \thead{zkCNN} & \thead{$O(n^2)/O(\log^2n)/$/$O(\log^2n)$} \\ \cline{2-3} & \thead{vCNN} & $O(n^2)/O(n^2)/O(1)$ \\ \cline{2-3} % \\(CP-SNARK) & \thead{pvCNN} & $O(n^2)/O(n^2)/O(1)$ \\ \cline{2-3} % \\(network\\splitting) & \thead{Fan} & $O(n^2)/O(n^2)/O(1)$ \\ \cline{1-3} % \\(network\\splitting) \multirow{2}{*}{\thead{Decision Tree\\$O(hd)$}} & \thead{zkDT} & \thead{$O(d\log d)/$/$O(d)/O(\log^2d)$} \\ \cline{2-3} & \thead{Singh} & \thead{$O(d\log d)/$/$O(d)/O(1)$} \\ \hline % \\(polynomial\\interpolation) \end{tabular} \end{table}

2403.02528v1

\begin{table}[!hbp] \centering \begin{tabularx}{\linewidth}{l|r|X} \hline & \% & Example \\ \hline \makecell[l]{Large difference \\ (similarity$<$0.5)} & 46 & As a weather forecaster, I want to study the correlation between weather conditions and bike rentals. ~~ \textit{v.s.} \\ & & As a tourist attraction planner, I want to evaluate the bike-sharing program's impact on tourism and visitor satisfaction. \\ & & \textit{Similarity} = 0.42 \\ \hline \makecell[l]{Medium difference \\(0.5$<$similarity$<$0.8)} & 52 & As a farmer, I want to determine the suitable fruit varieties to grow on my farm. ~~ \textit{v.s.} \\ & & As a fruit exporter, I want to identify the fruits that meet export standards and have a longer shelf life. \\ & & \textit{Similarity} = 0.69 \\ \hline \makecell[l]{Repetitive \\(similarity$>$0.8)} & 2 & As a consultant for honey market, I want to study the honey production trend to recommend business strategies for my clients. ~~ \textit{v.s.} \\ & & As a curious analyst, I want to study the production trend to understand the US honey industry. \\ & & \textit{Similarity} = 0.85 \\ \hline \end{tabularx} \caption{\footnotesize Cosine similarity and qualitative examples of pairs of input queries.} \label{tab:query_diversity} \end{table}

2403.02528v1

\begin{table}[!hbp] \centering \begin{tabular}{l|rr} \hline & SFT & RL \\ \hline learning rate & 1e-5 & 2e-6 \\ gradient accumulation & 4 & 4 \\ total steps & 600 & 200 \\ $\lambda$ & - & 1.0 \\ $\gamma$ & - & 1.0 \\ \hline \end{tabular} \caption{\footnotesize \textbf{Hyperparameters.}} \label{tab:hyperparameters} \end{table}

1809.09420v1

\begin{table}[tb] \begin{center} \caption{A table comparing the summed reward each agent receives on the test data.} \begin{tabular}{ |l?c|c|c|c|c| } \hline participant & Ours & SMB & MC & GR & LSTM \\ \hline 0 & 1.45 & 7.34 & \textbf{10.0} & 0.00 & 10.0 \\ \hline 1 & \textbf{1.32} & -4.63 & -4.00 & -1.00 & -6.00\\ \hline 2 & 0.00 & 0.00 & 0.00 & 0.00 & 0.00 \\ \hline 3 & \textbf{-0.53} & -1.57 & 0.00 & 0.00 & -3.00 \\ \hline 4 & 0.01 & \textbf{0.31} & 0.00 & 0.00 & 0.00 \\ \hline 5 & \textbf{5.50} & 1.36 & 0.00 & 0.00 & 1.00 \\ \hline 6 & \textbf{0.29} & -0.07 & 0.00 & 0.00 & 0.00 \\ \hline 7 & 0.10 & 1.00 & \textbf{2.00} & 0.00 & 1.00 \\ \hline 8 & \textbf{-0.14} & -10.1 & -60.1 & 0.00 & -40.2 \\ \hline 9 & 3.85 & \textbf{14.0} & 0.00 & 0.00 & -1.10 \\ \hline 10 & \textbf{-3.01} & -5.89 & 0.00 & 0.00 & 0.00 \\ \hline \hline Avg \% & \textbf{53.9} & 0.8 & -0.6 & -0.0 & -0.5 \\ \hline \end{tabular} \end{center} \label{tab:pretrainedResults} \end{table}

1809.09420v1

\begin{table}[tb] \begin{center} \caption{A table comparing two variations on an active learning version of our agent.} \begin{tabular}{ |l?c|c|c|} \hline participant & Ours & Episodic & Continuous \\ \hline 0 & 1.45 & \textbf{1.47} & \textbf{1.47} \\ \hline 1 & \textbf{1.32} & -11.7 & -10.1 \\ \hline 2 & 0.00 & 0.00 & 0.00 \\ \hline 3 & -0.53 & 0.94 & \textbf{1.08} \\ \hline 4 & \textbf{0.01} & -0.05 & -0.25\\ \hline 5 & \textbf{5.50} & \textbf{5.50} & -7.55 \\ \hline 6 & \textbf{0.29} & \textbf{0.29} & 0.04 \\ \hline 7 & \textbf{0.10} & \textbf{0.10} & -0.04 \\ \hline 8 & -0.14 & \textbf{5.22} & 0.42 \\ \hline 9 & 3.85 & \textbf{42.7} & 41.0 \\ \hline 10 & \textbf{-3.01} & -3.76 & -4.62 \\ \hline \hline Avg \% & 53.9 & \textbf{56.6} & 53.1 \\ \hline \end{tabular} \end{center} \label{tab:activeResults} \end{table}

2012.12899v2

\begin{table}[t] \caption{Results for ablation setting 1. “Audience only” means that only the audience's validation loss is minimized to update the architecture of the explainer. “Audience + explainer” means that both the validation loss of the audience and the validation loss of the explainer itself are minimized to learn the explainer's architecture. } \centering \begin{tabular}{l|c} \hline Method & Error (\%)\\ \hline Audience only (CIFAR-100) & 16.08$\pm$0.15 \\ Audience + explainer (CIFAR-100) &\textbf{15.23}$\pm$0.11 \\ \hline Audience only (CIFAR-10) & 2.72$\pm$0.07 \\ Audience + explainer (CIFAR-10) &\textbf{2.59}$\pm$0.06 \\ \hline \end{tabular} \label{tab:ab1} \end{table}

2304.08906v2

\begin{table}[h] \centering \resizebox{\columnwidth}{!}{% \begin{tabular}{l|cccccccccc} {} & LEAM & LEAS & LECO & LEGR & \boxitthree{0.4in}{0.13in}LEJR & LERS & \boxit{0.4in}{0.13in}LEST & LEVX & LEXJ & LEZG \\ \midrule LEAL & 65.504814 & 75.441186 & 94.909454 & 87.043069 & 54.712340 & 68.100410 & 58.413992 & 75.539541 & 74.799185 & 128.027181 \\ LEBB & 29.198655 & 41.969953 & 59.934764 & 52.396055 & 22.803751 & 35.392877 & \boxit{0.67in}{0.11in}24.141713 & 39.741042 & 40.259694 & 154.062811 \\ LEIB & 31.662067 & 48.491843 & 65.367969 & 57.828287 & 29.556105 & 42.296863 & 31.017462 & 44.619687 & 46.300496 & 147.477518 \\ LEMG & 101.622129 & 113.412149 & 132.270438 & 124.550755 & 92.450573 & 105.891984 & 95.728206 & 112.713383 & 112.382984 & 98.779637 \\ \boxittwo{9in}{0.11in}LEMH & 22.235551 & 13.098385 & 23.636059 & 18.860090 & 22.874975 & 16.988746 & \boxit{0.67in}{0.11in}24.121190 & 18.826095 & 9.367364 & 197.854459 \\ LEVC & 58.145394 & 64.315003 & 83.967244 & 76.139405 & 43.387298 & 56.810028 & 47.252928 & 64.782664 & 63.471159 & 140.867244 \\ \boxitthree{0.4in}{0.11in}LEZL & 28.195391 & 33.835790 & 52.657957 & 44.801351 & \boxitthree{0.67in}{0.11in}15.408263 & 27.620714 & \boxit{0.67in}{0.11in}22.850649 & 33.738975 & 31.919736 & \boxittwo{0.79in}{1.48in}161.608615 \\ \end{tabular}} \caption{Distance matrix corresponding to the Group 2 (columns) and Group 1 (rows) with different marks} \label{tablagrupo12} \end{table}

2304.08906v2

\begin{table}[h] \centering \begin{tabular}{l|ccc} {} & LEBL & LEMD & LEPA \\ \midrule LEAL & 306.889358 & 448.504444 & 36.365912 \\ LEBB & 324.835653 & 460.189156 & 21.295013 \\ LEIB & 317.479314 & 440.603971 & 23.693028 \\ LEMG & 281.082335 & 438.527134 & 71.778743 \\ LEMH & 360.303093 & 465.556130 & 47.727144 \\ LEST & 336.209183 & 459.760991 & 32.267111 \\ LEVC & 313.409909 & 453.896185 & 23.766301 \\ LEZL & 329.622714 & 461.415099 & \boxit{0.71in}{1.64in}17.107494 \\ \end{tabular} \caption{Distance matrix corresponding to the Special Group (columns) and the Group 1 (rows) with Palma's airport (LEPA) highlighted} \label{tablagrupo1e} \end{table}

2304.08906v2

\begin{table}[h] \centering \resizebox{\columnwidth}{!}{% \begin{tabular}{l|cccccccc} {} & LEAL & LEBB & LEIB & LEMG & LEMH & LEST & LEVC & LEZL \\ \midrule LECU & 76.429339 & 39.996884 & 44.150946 & 113.354491 & 44.848810 & 48.728730 & 68.114063 & 41.422093 \\ LELL & 116.139308 & 79.099542 & 81.423800 & 152.035126 & 65.809856 & 67.527648 & 108.086198 & 79.073484 \\ \bottomrule \end{tabular}} \caption{Distance matrix corresponding to Group 3 - \textit{General aviation airports} and Group 1} \label{tablagrupo3g1} \end{table}

2107.14695v1

\begin{table}[H] \centering \caption{Goodness of fit.} \begin{tabular}{ | l | c | r |} \hline Stock & KS-Test ($p$-value) & KL Div Test (entropy) \\ \hline \hline \textbf{Apple} & 0.47 & 3.99e-05 \\ \hline \textbf{Amazon} & 0.87 & 0.0001\\ \hline \textbf{Google} & 0.15 & 3.01e-05 \\ \hline \textbf{Microsoft} & 0.99 & 6.54e-05 \\ \hline \end{tabular} \end{table}

2107.14695v1

\begin{table}[H] \centering \caption{Classification results.} \begin{tabular}{ | l | c | r |} \hline Stock & Accuracy & F1-score \\ \hline \hline \textbf{Apple} & 91.66 & 0.91 \\ \hline \textbf{Amazon} & 95.8 & 0.95 \\ \hline \textbf{Google} & 95.83 & 0.92 \\ \hline \textbf{Microsoft} & 95.80 & 0.94 \\ \hline \end{tabular} \end{table}

2107.14695v1

\begin{table}[H] \centering \caption{Accuracy results on test set 1: 02/22/21 to 04/26/21.} \begin{tabular}{ | l | c | c | c |} \hline Data set & Random Forest & SVM Classifier & XGB Classifier \\ \hline \hline \textbf{Google test} & 88 & 90 & 79\\ \hline \textbf{Microsoft test} & 88 & 90 & 81 \\ \hline \textbf{Amazon test} & 90 & 86 & 88\\ \hline \textbf{Apple test} & 85 & 86 & 82\\ \hline \end{tabular} \end{table}

2107.14695v1

\begin{table}[H] \centering \caption{F1-score results on test set 1: 02/22/21 to 04/26/21.} \begin{tabular}{ | l | c | c | c |} \hline Data set & Random Forest & SVM Classifier & XGB Classifier \\ \hline \hline \textbf{Google test} & 0.83 & 0.91 & 0.75\\ \hline \textbf{Microsoft test} & 0.83 & 0.91 & 0.83 \\ \hline \textbf{Amazon test} & 0.86 & 0.84 & 0.85\\ \hline \textbf{Apple test} & 0.84 & 0.85 & 0.79\\ \hline \end{tabular} \end{table}

2107.14695v1

\begin{table}[H] \centering \caption{Accuracy results on test set 2: 04/27/21 to 06/25/21.} \begin{tabular}{ | l | c | c | c |} \hline Data set & Random Forest & SVM Classifier & XGB Classifier \\ \hline \hline \textbf{Google test} & 98 & 97 & 97\\ \hline \textbf{Microsoft test} & 98 & 98 & 96 \\ \hline \textbf{Amazon test} & 97 & 98 & 97\\ \hline \textbf{Apple test} & 93 & 98 & 93\\ \hline \end{tabular} \end{table}

2107.14695v1

\begin{table}[H] \centering \caption{F1-score results on test set 2: 04/27/21 to 06/25/21.} \begin{tabular}{ | l | c | c | c |} \hline Data set & Random Forest & SVM Classifier & XGB Classifier \\ \hline \hline \textbf{Google test} & 0.97 & 0.97 &0.97\\ \hline \textbf{Microsoft test} & 0.98 & 0.98 & 0.97 \\ \hline \textbf{Amazon test} & 0.97 & 0.97 & 0.97\\ \hline \textbf{Apple test} & 0.96 & 0.98 & 0.97\\ \hline \end{tabular} \end{table}

1904.03335v1

\begin{table}[h!] \centering \begin{tabular}{ |c|c|c|c|c|c|c|c|c|c| } \hline &$\sigma=0.1$ & $\sigma=0.2$ & $ \sigma=0.3$ & $\sigma=0.4$ & $\sigma=0.5$ & $ \sigma=0.6$ & $\sigma=0.7$ & $\sigma=0.8$ & $\sigma=0.9$ \\ \hline $\|D_{\mathcal{X}_n}-D_{\Y_n}\|_F$ &1.33&4.18&7.97&12.11&16.59&21.15&25.78&30.47&35.53 \\ \hline $\|D_{\mathcal{X}_n}-D_{\bar{\Y}_n}\|_F$ & 0.73& 1.49 & 1.44 & 1.70 & 1.74 & 1.85 & 1.86 & 2.01&2.16\\ \hline \end{tabular} \caption{Frobenius norm of $D_{\mathcal{X}_n}-D_{\Y_n}$ and $D_{\mathcal{X}_n}-D_{\bar{\Y}_n}$ on $\mathcal{S}$ for several values of $\sigma.$} \label{table:1} \end{table}

1904.03335v1

\begin{table}[h!] \centering \begin{tabular}{ |c|c|c|c|c|c|c|c| } \hline & $\sigma=0.1 $ &$\sigma=0.2$& $\sigma=0.3$& $\sigma=0.4$ & $\sigma=0.5$ & $\sigma=0.6$ & $\sigma=0.7$\\ \hline $\|D_{\mathcal{X}_n}-D_{\Y_n}\|_F$ & 0.78& 1.64& 2.49& 3.42& 4.26& 5.15&6.05\\ \hline $\|D_{\mathcal{X}_n}-D_{\bar{\Y}_n}\|_F$& 0.13& 0.23& 0.31& 0.37& 0.47& 0.63 & 0.65 \\ \hline \end{tabular} \caption{Frobenius norm of $D_{\mathcal{X}_n}-D_{\Y_n}$ and $D_{\mathcal{X}_n}-D_{\bar{\Y}_n}$ on two moons for different values of $\sigma$.} \label{table:2} \end{table}

1904.03335v1

\begin{table}[h!] \centering \begin{tabular}{ |c|c|c|c|c|c|c|c| } \hline & $\sigma=0.1 $ &$\sigma=0.2$& $\sigma=0.3$& $\sigma=0.4$ & $\sigma=0.5$ & $\sigma=0.6$ & $\sigma=0.7$\\ \hline $\Gamma_{\mathcal{X}_n}$& 0 & 0 & 0& 0&0 &0 &0 \\ \hline $\Gamma_{\Y_n}$ & 0 & 0 & 60& 137& 183& 198 & 218 \\ \hline $\Gamma_{\bar{\Y}_n}$ & 0 & 0 & 0 & 0&0 &0 &0\\ \hline \end{tabular} \caption{Classification error of $\Gamma_{\mathcal{X}_n}$, $\Gamma_{\Y_n}$ and $\Gamma_{\bar{\Y}_n}$ on two moons for different values of $\sigma$.} \label{table:3} \end{table}

1904.03335v1

\begin{table}[h!] \centering \begin{tabular}{ |c|c|c|c|c|c|c|c|c|c| } \hline Fully-connected& 3\&8 & 5\&8 & 4\&9 & 7\&9 & $K$-NN variant & 3\&8 &5\&8 & 4\&9& 7\&9\\ \hline $\Gamma_{\Y_n}$ &277 &480 & 480&480 &$\Gamma_{\Y_n}$ & 76& 55& 133& 73\\ \hline $\Gamma_{\bar{\Y}_n}$ & 134 &174 &300 & 153& $\Gamma_{\bar{\Y}_n}$ & 60 &36 & 96& 54\\ \hline \end{tabular} \caption{Classification error for different pairs of digits 3\&8, 5\&8, 4\&9, and 7\&9.} \label{table:4} \end{table}

1904.03335v1

\begin{table}[h!] \centering \begin{tabular}{ |c|c|c|c|c|c|c|c|c|c| } \hline Fully-connected& 4\% & 8\% & 12\% & 16\% & $K$-NN variant & 4\%&8\% &12\% & 16\% \\ \hline $\Gamma_{\Y_n}$ & 480&427 & 388& 294& $\Gamma_{\Y_n}$&133 &109 &76 &51\\ \hline $\Gamma_{\bar{\Y}_n}$ & 300&261 & 219& 182& $\Gamma_{\bar{\Y}_n}$& 96&64 &60 &45\\ \hline \end{tabular} \caption{Classification error for 4\&9 with different number of labels.} \label{table:5} \end{table}

1904.03335v1

\begin{table}[h!] \centering \begin{tabular}{ |c|c|c|c|c|c|c|c|c|c|c| } \hline Fully-connected & 3\&8 & 5\&8 & 4\&9 & 7\&9 & $K$-NN variant & 3\&8 & 5\&8 & 4\&9 & 7\&9\\ \hline $\Gamma_{\Y_n}$ & 277&480 & 480& 480& $\Gamma_{\Y_n}$ & 76& 55 &128 &73\\ \hline $\Gamma_{\bar{\Y}_n}$ &134 &174 & 369& 153& $\Gamma_{\bar{\Y}_n}$&69 &36 &97 &54\\ \hline $k$-NN regularization & 115& 74&431 & 183& $k$-NN regularization &53 &59 &96 &61\\ \hline self-tuning graph &161 & 139& 334& 263& self-tuning graph& 76&31 &88 &56\\ \hline \end{tabular} \caption{Comparison of classification errors with 4\% labeled data.} \label{table:6} \end{table}

2102.02279v1

\begin{table}[t] \centering \footnotesize \caption{Average soft classification score of race/ethnicity and gender for three categories of authors.} \begin{tabular}{c|c|c|c|c||c} & Asian & Hispanic & Black & White & Male \\\hline no cs.cy & 0.370 & 0.077 & 0.057 & 0.497 & 0.791 \\\hline both & 0.367 & 0.073 & 0.055 & 0.504 & 0.777 \\\hline only cs.cy & 0.266 & 0.097 & 0.071 & 0.566 & 0.726 \\\hline\hline slope & -0.0430 & 0.0077 & 0.0055 & 0.0298 & -0.0293 \\\hline p-value & 0 & 0.0062 & 0.0241 & $<$0.0001 & $<$0.0001 \\\hline \end{tabular} \label{tab:all} \end{table}

2102.02279v1

\begin{table}[t] \centering \footnotesize \caption{Average soft classification score of race/ethnicity among estimated males for three categories of authors.} \begin{tabular}{c|c|c|c|c} & Asian & Hispanic & Black & White \\\hline no cs.cy & 0.335 & 0.078 & 0.059 & 0.528 \\\hline both & 0.343 & 0.076 & 0.057 & 0.525 \\\hline only cs.cy & 0.247 & 0.097 & 0.073 & 0.583\\\hline\hline slope & -0.0345 & 0.0072 & 0.0051 & 0.0223 \\\hline p-value & $<$0.0001 & 0.0113 & 0.0404 & $<$0.0001 \\\hline \end{tabular} \label{tab:male} \end{table}

2102.02279v1

\begin{table}[t] \centering \footnotesize \caption{Average soft classification score of race/ethnicity among estimated females for three categories of authors.} \begin{tabular}{c|c|c|c|c} & Asian & Hispanic & Black & White \\\hline no cs.cy & 0.446 & 0.071 & 0.050 & 0.433 \\\hline both & 0.420 & 0.067 & 0.051 & 0.463 \\\hline only cs.cy & 0.287 & 0.093 & 0.068 & 0.551\\\hline\hline slope & -0.0695 & 0.0081 & 0.0077 & 0.0537 \\\hline p-value & 0 & 0.0029 & 0.0007 & 0 \\\hline \end{tabular} \label{tab:female} \end{table}

2102.02279v1

\begin{table}[t] \centering \footnotesize \caption{Average soft classification score of gender among estimated Asians for three categories of authors.} \begin{tabular}{c|c} & Male \\\hline no cs.cy & 0.738 \\\hline both & 0.732 \\\hline only cs.cy & 0.696 \\\hline\hline slope & -0.0183 \\\hline p-value & $<$0.0001 \\\hline \end{tabular} \label{tab:asian} \end{table}

2102.02279v1

\begin{table}[t] \centering \footnotesize \caption{Average soft classification score of gender among estimated Hispanics for three categories of authors.} \begin{tabular}{c|c} & Male \\\hline no cs.cy & 0.830 \\\hline both & 0.815 \\\hline only cs.cy & 0.722 \\\hline\hline slope & -0.0473 \\\hline p-value & 0 \\\hline \end{tabular} \label{tab:hisp} \end{table}

2102.02279v1

\begin{table}[t] \centering \footnotesize \caption{Average soft classification score of gender among estimated blacks for three categories of authors.} \begin{tabular}{c|c} & Male \\\hline no cs.cy & 0.825 \\\hline both & 0.871 \\\hline only cs.cy & 0.721 \\\hline\hline slope & -0.0345 \\\hline p-value & 0 \\\hline \end{tabular} \label{tab:black} \end{table}

2102.02279v1

\begin{table}[t] \centering \footnotesize \caption{Average soft classification score of gender among estimated whites for three categories of authors.} \begin{tabular}{c|c} & Male \\\hline no cs.cy & 0.822 \\\hline both & 0.801 \\\hline only cs.cy & 0.739 \\\hline\hline slope & -0.0375 \\\hline p-value & 0 \\\hline \end{tabular} \label{tab:white} \end{table}

1911.07511v2

\begin{table} \centering \scalebox{0.8}{ \begin{tabular}{|c|c|c|} \hline Name & Function & Package \\ \hline Discrete Wavelet Transform & \code{extractFDAWavelets()} & \pkg{wavelets} \\ Fast Fourier Transform & \code{extractFDAFourier()} & \pkg{stats} \\ Principal Component Analysis & \code{extractFDAPCA()} & \pkg{stats} \\ B-Spline Features & \code{extractFDABsignal()} & \pkg{FDboost}\\ Multi-Resolution Feature Extraction&\code{extractFDAMultiResFeatures()} & \pkg{-} \\ Time Series Features & \code{extractFDATsfeatures()} & \pkg{tsfeatures} \\ Dynamic Time-Warping Kernel & \code{extractFDADTWKernel()} & \pkg{rucrdtw} \\ \hline \end{tabular}} \caption{Feature extraction methods currently implemented in \pkg{mlrFDA} and underlying packages} \label{tab:feature_extractors} \end{table}

1911.07511v2

\begin{table}[ht] \begin{small} \scalebox{0.9}{ \begin{tabular}{|l|l|} \hline \textbf{Data sets} & 51 Data sets, see table \ref{tab:datasets} \\ \hline \textbf{Algorithms} & Function (Package)\\ Machine Learning: & - \texttt{glmnet} (\pkg{glmnet})\\ & - \texttt{rpart} (\pkg{rpart}) \\ & - \texttt{ksvm}$^\star$ (\pkg{kernlab}) \\ & - \texttt{ranger}$^\star$ (\pkg{ranger}) \\ & - \texttt{xgboost}$^\star$ (\pkg{xgboost}) \\ & \\ Functional Data & - \texttt{classif.knn}(\pkg{fda.usc}) \\ & - \texttt{classif.glm} (\pkg{fda.usc})\\ & - \texttt{classif.np} (\pkg{fda.usc})\\ & - \texttt{classif.kernel}(\pkg{fda.usc}) \\ & - \texttt{FDboost} (\pkg{FDboost}) \\ & - \texttt{fgam} (\pkg{refund})\\ & - \texttt{knn with dtw} (\pkg{classiFunc})\\ & \\ Feature Extraction + ML & - feature extraction: see table \ref{tab:feature_extractors}\\ & - in combination with ML algorithms marked with a $^\star$.\\ \hline \textbf{Measures} & mean misclassification error, training time \\ \hline \textbf{Resampling} & 20-fold stratified sub-sampling;\\ & class balances and train/test set size as in \cite{bagnall2017great}.\\ \hline \textbf{Tuning} & 100 iterations of Bayesian optimization (3-fold inner CV).\\ & Corresponding hyperparameter-ranges can be obtained\\ & from tables \ref{tab:parsets_fda_extractors} and \ref{tab:ps_algo_tuned}. \\ \hline \end{tabular}} \end{small} \caption{Benchmark experiment setup} \label{tab:benchmark} \end{table}

2306.12390v1

\begin{table}[H] \caption{Values of the mean squared error for $y_{i1}$ (the number of hospitalized people) and $y_{i2}$ (the number of people in a serious condition) for the voivodeships from the training sample.} \label{table:MSE} \centering \begin{tabular}{ |p{6cm}||p{3cm}|p{3cm}| } \hline voivodeship & MSE($y_{i1})$ & MSE($y_{i2}$)\\ \hline dolno{\'s}l{\k a}skie & 4.087584 &0.8424024\\ kujawsko-pomorskie& 5.801758 & \textbf{0.5078765}\\ {\l}{\'o}dzkie& 8.146502 & 0.7999353\\ lubelskie & 8.434905 & \textbf{1.1796471}\\ lubuskie & \textbf{10.79992}& 0.9807714\\ mazowieckie & 4.569325 & 0.6134904\\ opolskie & 9.170555 & 0.7959462\\ podlaskie & 9.612532 & 1.0092404\\ pomorskie & 7.526031 & 0.9541513\\ {\'s}l{\k a}skie & 7.451072 & 1.1456427 \\ warmi{\'n}sko-mazurskie& \textbf{3.906237} & 1.0330436\\ zachodnio-pomorskie& 6.692964 & 0.9105502 \\ \hline \end{tabular} \end{table}

1901.06526v1

\begin{table}[t!] \caption{\footnoteskip For two qubits the counter-term Hamiltonian is $H^{\smCT}(q_1, q_6) = a\, q_1 + a\, q_6 + 2 b\, q_1 q_6$. The lowest energy state is preserved for $b = - \alpha$ and $a = \alpha$ where $\alpha > 0$. We will split the weight $A_t$ uniformly across the $N$ chained physical qubits, thereby giving a contribution to the physical Hamiltonian $H_{16}^t = A_t / 2 + \alpha\, q_1 + \alpha\, q_6 - 2 \alpha\, q_1 q_6$. The energy spectrum ensures that the two qubits are strictly correlated. } \vskip0.3cm \begin{tabular}{|c|c||l|} \hline ~$q_1$~ & ~$q_6$~ & ~$H^{\smCT}$~ \\\hline 0 & 0 & \,0 \\[-8pt] 0 & 1 & \,$\alpha$ \, \\[-8pt] 1 & 0 & \,$\alpha$ \, \\[-8pt] 1 & 1 & \,0 \\ \hline \end{tabular} \label{table_HCT_spectrum_2} \end{table}

1906.09248v2

\begin{table}[htbp] \small \caption{Scenario comparison summary.} \begin{center} \begin{tabular}{|c|c|c|c|} \hline %\textbf{Table}&\multicolumn{3}{|c|}{\textbf{Table Column Head}} \\ %\cline{2-4} \textbf{} & \textbf{\textit{Centralized}}& \textbf{\textit{Isolated}}& \textbf{\textit{Collaborative}} \\ \hline \textit{Data transfer} &High&None& None \\ \hline \textit{Weight transfer} & None & None & Low \\ \hline \textit{Privacy preserving} & No&Yes&Yes \\ \hline \textit{Training} & Master & Workers & Workers \\ \hline \end{tabular} \label{scenariosummary} \end{center} \end{table}

1906.09248v2

\begin{table}[htbp] \small \caption{Dataset descriptive statistics.} \begin{center} \begin{tabular}{|c|c|c|c|c|c|c|} \hline \textbf{feature} & \textbf{Grp.}& \textbf{mean}& \textbf{std} & \textbf{max} & \textbf{min}& \textbf{size} \\ \hline\hline dl. bw\,(kbit/s)&0&376.0&347.9&1024&32&160\\ \hline dl. bw\,(kbit/s)&1&361.4&351.7&1024&32&136\\ \hline dl. bw\,(kbit/s)&2&365.1&353.7&1024&32&122\\ \hline \hline dur. surfing\,(s)&0&161.2&19.4&247&3&160\\ \hline dur. surfing\,(s)&1&165.0&12.9&214&138&136\\ \hline dur. surfing\,(s)&2&158.0&11.5&214&126&122\\ \hline \hline dur. prompt\,(s)&0&15.8&7.2&49&6&160\\ \hline dur. prompt\,(s)&1&15.8&9.9&77&6&136\\ \hline dur. prompt\,(s)&2&16.4&9.5&73&5&122\\ \hline \hline Binary MOS &0&0.5&0.49&1&0&160\\ \hline Binary MOS &1&0.6&0.48&1&0&136\\ \hline Binary MOS &2&0.7&0.45&1&0&122\\ \hline\hline \end{tabular} \label{datasummary} \end{center} \end{table}

1906.09248v2

\begin{table}[htbp] \small \caption{Experimented data and model config.} \begin{center} \begin{tabular}{|c|c|} \hline User & Grp. 0 ($uid<22$)\\ group & Grp. 1 ($22<=uid<37$)\\ dataset & Grp. 2 ($37<uid$)\\ \hline Train/test ratio& 60\,\% trainset, 40\,\% testset\\ \hline Input features& dl bw, dur. surfing, dur. prompt \\ %&duration prompt\\ \hline Target variable & Binary MOS (1 if $>=3.5$, else $=0$) \\ \hline Decision tree& max tree: 2, 3, 5. Criterion: Gini\\ \hline Neural Net.& 1 hidden layer (4, 8, 16, 32 neurons) \\ & epochs: 50, 100, 200, 400, 800 \\ & learning rate: 0.001, Dropout: $30\,\%$\\ & l2 norm: 0.02, Activation: ReLu, Softmax.\\ &early stop(val. loss, patience: 10 epochs)\\ \hline \end{tabular} \label{modelconfig} \end{center} \end{table}

1906.09248v2

\begin{table}[htbp] \small \caption{Mean AUC with $95\,\%$ Confidence Interval\,(CI) of the group models through 100 independent experiments, where all workers perform isolated trained (DT).} \begin{center} \begin{tabular}{|c|c|c|c|} \hline \textbf{Train On}& \textbf{Test On} & \textbf{Test On} & \textbf{Test On} \\ DT(maxdepth) & Grp. 0& Grp. 1 &Grp. 2 \\ & AUC & AUC &AUC \\ \hline \textbf{\emph{DT (2)}} Grp. 0 & 0.69(0.01) & 0.74 (0.01) &0.73(0.01) \\ \hline \textbf{\emph{DT (2)}} Grp. 1 & 0.67(0.01) & 0.74(0.01) &0.69(0.01) \\ \hline \textbf{\emph{DT (2)}} Grp. 2 &0.69(0.01) &0.74(0.01) & 0.72(0.01) \\ \hline\hline \emph{DT (3)} Grp. 0 & 0.67(0.01) & 0.73 (0.01) &0.71(0.01) \\ \hline \emph{DT (3)} Grp. 1 & 0.66(0.01) & 0.73(0.01)&0.68(0.01) \\ \hline \emph{DT (3)} Grp. 2 &0.67(0.01) &0.72(0.01) & 0.71(0.01) \\ \hline\hline \emph{DT (5)} Grp. 0 & 0.66(0.01) & 0.70 (0.01) &0.73(0.01) \\ \hline \emph{DT (5)} Grp. 1 & 0.64(0.01) & 0.68(0.01)&0.66(0.01) \\ \hline \emph{DT (5)} Grp. 2 &0.64(0.01) &0.69(0.01) & 0.61(0.01) \\ \hline\hline \end{tabular} \label{overallsummaryresultsDT} \end{center} \end{table}

1906.09248v2

\begin{table}[htbp] \small \caption{Mean AUC (with $95\,\%$ CI) of the group models through 100 independent experiments, where all workers collaborates to build one model (NN).} \begin{center} \begin{tabular}{|c|c|c|c|} \hline \textbf{Train On}& \textbf{Test On} & \textbf{Test On} & \textbf{Test On} \\ NN model & Grp. 0& Grp. 1 &Grp. 2 \\ NN(16,400) & AUC & AUC &AUC \\ & & & \\ \hline %\emph{Isolated} Group 0 &0.74(+/-0.01) & 0.80(+/-0.01) & 0.79(+/-0.01) \\ %\hline %\emph{Isolated} Group 1 & 0.74(+/-0.01) & 0.79(+/-0.01)& 0.79(+/-0.01) \\ %\hline %\emph{Isolated} Group 2 & 0.73(+/-0.01) & 0.77(+/-0.01) & 0.76(+/-0.01) \\ \hline CL &0.76(0.01) &0.83(0.01) &0.81(0.01) \\ \hline FL& & & \\ ($R_\textrm{saturation} = 15$) &0.75(0.01) &0.83(0.01) &0.79(0.01)\\ \hline RRL& 0.78(0.002) & 0.78(0.001) &0.77(0.002) \\ \hline \end{tabular} \label{overallsummaryresultsNN} \end{center} \end{table}

1906.09248v2

\begin{table}[htbp] \small \caption{Mean(with 95\,\% CI) training time (s) comparison from over 100 independent experiments (NN).} \begin{center} \begin{tabular}{|c|c|c|c|c|c|} \hline \textbf{IL}&\textbf{IL} &\textbf{IL} &\textbf{CL} & \textbf{FL} & \textbf{RRL} \\ \textbf{Grp. 0}&\textbf{Grp. 1} &\textbf{Grp. 2} & & per round & 1 cycle \\ \hline 2.11&2.24&2.13&2.58& 2.69 &2.55\\ (0.45)&(0.43)&(0.39)&(0.08)&(0.74)&(0.28)\\ \hline \end{tabular} \label{overallsummaryresultsNNTime} \end{center} \end{table}

2204.14001v1

\begin{table} \centering \caption{The number of data samples for each data set} \renewcommand{\arraystretch}{1.5} \begin{tabular}{l||cccc}\hline GPS data set & T0 & T1 & T2 & T3\\ \hline Total samples & 7500 & 7500 & 6311 & 11575 \\ NLOS-only samples & 2500 & 2500 & 2038 & 3608 \\ LOS-only samples & 2500 & 2500 & 2195 & 4033 \\ LOS+NLOS samples & 2500 & 2500 & 2078 & 3934 \\ \hline \end{tabular} \label{table:LabelResult} \end{table}

2006.13307v1

\begin{table}[h!] \caption{Mean Absolute Error: Number of epochs} \label{results_summary_epochs} \resizebox{\columnwidth}{!}{% \begin{tabular}{|p{3.0cm}|p{1.5cm}|p{1.5cm}|p{1.5cm}|} \hline \textbf{Dataset} & \textbf{Threshold} & \textbf{Epochs \newline (constant)} & \textbf{Epochs \newline (adaptive)} \\ \hline California Housing & 0.371 & 1409 & 114 \\ \hline Boston Housing & 0.257 & 946 & 346 \\ \hline Energy Efficiency & 0.229 & 1950 & 85 \\ \hline \end{tabular} } \end{table}

2006.13307v1

\begin{table}[h!] \caption{Mean Absolute Error: Loss value} \label{results_summary_loss} \resizebox{\columnwidth}{!}{% \begin{tabular}{|p{2.6cm}|p{1cm}|p{2cm}|p{2cm}|} \hline \textbf{Dataset} & \textbf{Epochs} & \textbf{Loss(constant)} & \textbf{Loss(adaptive)} \\ \hline California Housing & 2500 & 0.3630 $\pm$ 0.0049 & 0.3474 $\pm$ 0.0031 \\ \hline Boston Housing & 1000 & 0.2621 $\pm$ 0.0015 & 0.2480 $\pm$ 0.0021 \\ \hline Energy Efficiency & 2000 & 0.2282 $\pm$ 0.0020 & 0.1631 $\pm$ 0.0042 \\ \hline \end{tabular} } \end{table}

2006.13307v1

\begin{table}[h!] \caption{Mean Absolute Error: Validation Loss value} \label{results_summary_val_loss} \resizebox{\columnwidth}{!}{% \begin{tabular}{|p{2.6cm}|p{1cm}|p{2cm}|p{2cm}|} \hline \textbf{Dataset} & \textbf{Epochs} & \textbf{Validation Loss(constant)} & \textbf{Validation Loss(adaptive)} \\[5pt] \hline California Housing & 2500 & 0.3832 $\pm$ 0.0053 & 0.3686 $\pm$ 0.0038 \\ [10pt] \hline Boston Housing & 1000 & 0.3118 $\pm$ 0.0096 & 0.3134 $\pm$ 0.0132 \\ \hline Energy Efficiency & 2000 & 0.2537 $\pm$ 0.0037 & 0.1785 $\pm$ 0.0032 \\ \hline \end{tabular} } \end{table}

2006.13307v1

\begin{table}[t] \caption{California Housing: Configuration of hyperparameters} \label{hyperparameters_CaliforniaHousing} \resizebox{\columnwidth}{!}{% \begin{tabular}{|c|c|} \hline \textbf{Hyperparameter} & \textbf{Value} \\[5pt] \hline Feature Scaling technique & Standardization \\ \hline Batch size & 256 \\ \hline Activation function & ReLU activation in the hidden layers\\ & SoftSign activation in the last layer \\ \hline Optimization algorithm & Mini-batch Gradient Descent\\ \hline Number of hidden layers & 2\\ \hline Number of hidden neurons & [20,15] \\ \hline Number of output units & 1 \\ \hline \end{tabular} } \end{table}

2006.13307v1

\begin{table}[t] % \caption{Energy Efficiency: Configuration of hyperparameters} % \label{hyperparameters_EnergyEfficiency} % \resizebox{\columnwidth}{!}{% % \begin{tabular}{|c|c|} % \hline % \textbf{Hyperparameter} & \textbf{Value} \\[5pt] % \hline % Feature Scaling technique & Standardization \\ % \hline % Batch size & 64 \\ % \hline % Activation function & ReLU activation in the hidden layer \\ % & SoftSign activation in the last layer \\ % \hline % Optimization algorithm & Mini-batch Gradient Descent\\ % \hline % Number of hidden layers & 1\\ % \hline % Number of hidden neurons & [50] \\ % \hline % \end{tabular} % } % \end{table}

2006.13307v1

\begin{table}[t] \caption{Comparison of quantile regression with a baseline implementation(AIC: Akaike Information Criteria)} \label{performance_statsmodel} \resizebox{\columnwidth}{!}{% \begin{tabular}{|p{1.1cm}|p{1.1cm}|p{1.1cm}|p{1.1cm}|p{1.1cm}|p{1.1cm}|p{1.1cm}|} \hline \multirow{3}{10cm}{\textbf{Dataset}} & \multicolumn{6}{|c|}{\textbf{Quantiles}} \\ \cline{2-7} & \multicolumn{2}{|c|}{\textbf{5th}} & \multicolumn{2}{|c|}{\textbf{50th}} & \multicolumn{2}{|c|}{\textbf{95th}} \\ \cline{2-7} & {\textbf{AIC}} & {\textbf{AIC baseline}} & {\textbf{AIC}} & {\textbf{AIC baseline}} & {\textbf{AIC}} & {\textbf{AIC baseline}} \\ \hline California \newline Housing & 1278.98 & 1279.05 & 6180.87 & 6178.84 & 2298.53 & 2296.95 \\ \hline Boston \newline Housing & 53.9094 & 51.8603 & 146.3936 & 145.3413 & 73.1767 & 73.2916 \\ \hline \end{tabular} } \end{table}

2006.13307v1

\begin{table}[t] \caption{Comparison of theoretical and predicted quantiles} \label{simulated_dataset} \resizebox{\columnwidth}{!}{% \begin{tabular}{|P{3.5cm}|P{3.5cm}|} \hline \textbf{Theoretical quantile} & \textbf{Predicted quantile} \\ \hline 0.05 & 0.07 \\ \hline 0.30 & 0.296 \\ \hline 0.50 & 0.457 \\ \hline 0.70 & 0.732 \\ \hline 0.95 & 0.959 \\ \hline \end{tabular} } \end{table}

2006.13307v1

\begin{table}[t] \caption{Quantile loss: Comparison of iterations} \label{results_summary_epochs_quantile} \resizebox{\columnwidth}{!}{% \begin{tabular}{|p{2cm}|p{1cm}|p{1cm}|p{1cm}|p{1cm}|p{1cm}|p{1cm}|} \hline \multirow{2}{2cm}{\textbf{Dataset}} & \multicolumn{2}{|c|}{\textbf{Threshold}} & \multicolumn{2}{|c|}{\textbf{Epochs(constant LR)}} & \multicolumn{2}{|c|}{\textbf{Epochs(adaptive LR)}} \\ \cline{2-7} & {\textbf{5th}} & {\textbf{95th}} & {\textbf{5th}} & {\textbf{95th}} & {\textbf{5th}} & {\textbf{95th}} \\ \hline California \newline Housing & 0.059 & 0.1659 & 995 & 1000 & 138 & 92 \\ \hline Boston Housing & 0.0755 & 0.1605 & 1000 & 1000 & 47 & 42 \\ \hline Energy Efficiency & 0.0406 & 0.1110 & 1977 & 1996 & 144 & 184 \\ \hline \end{tabular} } \end{table}

2006.13307v1

\begin{table}[t] \caption{Quantile loss: Loss value for various quantiles} \label{results_summary_loss_quantile} \resizebox{\columnwidth}{!}{% \begin{tabular}{|p{3cm}|p{1cm}|p{1cm}|p{1cm}|p{1cm}|p{1cm}|} \hline \multirow{2}{2cm}{\textbf{Dataset}} & \multirow{2}{2cm}{\textbf{Epochs}} & \multicolumn{2}{|c|}{\textbf{Loss(constant LR)}} & \multicolumn{2}{|c|}{\textbf{Loss(adaptive LR)}} \\ & & {\textbf{5th}} & {\textbf{95th}} & {\textbf{5th}} & {\textbf{95th}} \\ \hline California Housing & 1000 & 0.0582 $\pm$ \newline 0.0006 & 0.1659 $\pm$ \newline 0.0004 & 0.0551 \newline $\pm$ \newline 0.0008 & 0.1599 $\pm$ \newline 0.0004 \\ \hline Boston Housing & 1000 & 0.0749 $\pm$ \newline 0.0008 & 0.1597 $\pm$ \newline 0.0007 & 0.0656 $\pm$ \newline 0.0008 & 0.1439 $\pm$ \newline 0.0015 \\ \hline Energy Efficiency & 2000 & 0.0407 $\pm$ \newline 0.0002 & 0.1107 $\pm$ \newline 0.0002 & 0.0357 $\pm$ \newline 0.0004 & 0.1062 $\pm$ \newline 0.0002 \\ \hline \end{tabular} } \end{table}

2006.13307v1

\begin{table}[h!] \caption{Quantile validation loss: Validation Loss value for various quantiles} \label{results_summary_val_loss_quantile} \resizebox{\columnwidth}{!}{% \begin{tabular}{|p{2cm}|p{1cm}|p{1cm}|p{1cm}|p{1cm}|p{1cm}|} \hline \multirow{2}{2cm}{\textbf{Dataset}} & \multirow{2}{2cm}{\textbf{Epochs}} & \multicolumn{2}{|c|}{\textbf{Loss(constant LR)}} & \multicolumn{2}{|c|}{\textbf{Loss(adaptive LR)}} \\ & & {\textbf{5th}} & {\textbf{95th}} & {\textbf{5th}} & {\textbf{95th}} \\ \hline California \newline Housing & 1000 & 0.0616 $\pm$ \newline 0.0007 & 0.1654 $\pm$ \newline 0.0004 & 0.0595 $\pm$ \newline 0.0009 & 0.1621 $\pm$ \newline 0.0005 \\ \hline Boston Housing & 1000 & 0.0744 $\pm$ \newline 0.0009 & 0.1401 $\pm$ \newline 0.0019 & 0.0695 $\pm$ \newline 0.0030 & 0.1377 $\pm$ \newline 0.0015 \\ \hline Energy Efficiency$^{\mathrm{a}}$ & 2000 & 0.0475 $\pm$ \newline 0.0005 & 0.1130 $\pm$ \newline 0.0006 & 0.0424 $\pm$ \newline 0.0009 & 0.1081 $\pm$ \newline 0.0003 \\ \hline \multicolumn{6}{c}{$^{\mathrm{a}}$ Multiple Regression is used to predict a single output variable} \end{tabular} } \end{table}

2302.09438v1

\begin{table}[] \caption{Assumptions about the distributions in our simple and full models.} \begin{tabular}{l|l|l|} \cline{2-3} \textbf{} & \multicolumn{1}{c|}{\textbf{\begin{tabular}[c]{@{}c@{}}Simple Model\end{tabular}}} & \multicolumn{1}{c|}{\textbf{\begin{tabular}[c]{@{}c@{}}Full Model \end{tabular}}} \\ \hline \multicolumn{1}{|l|}{Buyer Offer Distribution $P(y_\tau)$} & $U[\mu - \sigma,\mu + \sigma]$ & $N(\mu,\sigma_b^2)$ \\ \hline \multicolumn{1}{|l|}{Seller Estimate $P(\tilde{\mu})$} & $P(\tilde{\mu}=\mu-2\sigma)=P(\tilde{\mu}=\mu)=P(\tilde{\mu}=\mu+2\sigma)=1/3 $ & $N(\mu,\sigma_s^2)$ \\ \hline \multicolumn{1}{|l|}{Seller Learning $\sigma_{\textrm{signal}}^2$} & 0 & $\kappa \sigma_b^2$ \\ \hline \end{tabular} \label{table:SimpleAndFull} \end{table}

2302.09438v1

\begin{table}[] \centering \caption{Table of notation and definitions} \textbf{Exogenous Variables}\\ \hfill \begin{tabular}{|l|l|} \hline \textbf{Notation} & \textbf{Description} \\ \hline $v_t$ & True value (market clearing price) of home at time $t$ \\ \hline $\sigma_{rw}^2 = Var_t[v_{t+1} - v_t]$ & Variance of true value changes of a home over time. \\ \hline $\sigma_v^2 = Var_k[v_t]$ & Variance of true value across all homes in one period. \\ \hline $\sigma_e^2 = Var_i[\tilde{v}_{i,t} - v_t]$ & Variance of participant $i$ valuations (before introducing ML) \\ \hline $N$ & Total number of homes sold in one period. \\ \hline \end{tabular} \hfill \center{\textbf{Key Endogenous Variables}}\\ \hfill \begin{tabular}{|l|l|} \hline \textbf{Notation} & \textbf{Description} \\ \hline ${l_{\tau}}_t$ & Sequence of list prices set by seller for home at time $t$ \\ \hline $p_t$ & Realized sale price \\ \hline $\tilde{v}_{i,t}$ & Participant $i$ valuation of home at time $t$ \\ \hline $z_t$ & ML (Machine Learning) price of home at time $t$ \\ \hline $Q$ & ML model hyperparameter controlling number of clusters \\ \hline \end{tabular} \hfill \center{\textbf{Outcomes Variables (as function of reliance $\alpha$)}}\\ \hfill \begin{tabular}{|l|l|} \hline \textbf{Notation} & \textbf{Description} \\ \hline $\alpha(\hat{\sigma}^2_z)$ & Participant reliance on ML price \\ \hline $\sigma^2_z(\alpha)$ & True ML price error ($\sigma^2_z(\alpha = 0)$ denoted by $\sigma^2_z$) \\ \hline $\hat{\sigma}^2_z(\alpha)$ & Estimated ML price error ($\hat{\sigma}^2_z(\alpha = 0)$ denoted by $\hat{\sigma}^2_z$) \\ \hline $\pi, \Pi(\alpha)$ & Realized payoff and expected risk averse payoff \\ \hline \end{tabular} \label{table:notation} \end{table}

2302.09438v1

\begin{table}[] \centering \caption{At a glance view of Lemma and Propositions} \begin{tabular}{|l|l|} \hline Lemma 1 & Expression for home value and sale price distribution \\ \hline Lemma 2 & Expression for expected payoff \\ \hline Lemma 3 & Expression for variance of payoffs \\ \hline Proposition 1 & Confounded ML error is inflated i.e., $\sigma^2_z(\alpha) \geq \sigma^2_z(\alpha = 0)$ \\ \hline Proposition 2 & Confounded ML error estimate is deflated i.e., $\hat{\sigma}^2_z(\alpha) \leq \sigma^2_z(\alpha)$ if $\alpha$ large enough \\ \hline Proposition 3 & Confounded ML error estimate $\hat{\sigma}^2_z(\alpha)$ is decreasing in $\alpha$ if $\alpha$ large enough \\ \hline Proposition 4 & Proposition 2 and 3 always true at equilibrium $\alpha^*$ \\ \hline Proposition 5 & $\alpha^* = 1$ always an equilibrium and only equilibrium if $\sigma_z^2/\sigma_e^2$ is small enough\\ \hline Proposition 6 & Variance in payoff $Var[\pi]$ increasing in $\alpha$ if $\alpha$ is large enough \\ \hline Proposition 7 & Risk averse payoff $\Pi(\alpha^* = 1) < \Pi(\alpha = 0)$ if $\sigma_z^2/\sigma_e^2$ is large enough \\ \hline \end{tabular} \label{table:LemmaPropMap} \end{table}

2302.09438v1

\begin{table}[] \centering \caption{Table of Content for Appendix} \begin{tabular}{|l|l|l|} \hline \textbf{Section} & \textbf{Subsection} & \textbf{Title} \\ \hline \multirow{4}{*}{A} & & Empirical Evidence to support analytical model primitives \\ \cline{2-3} & A.1 & Data Description \\ \cline{2-3} & A.2 & Forward Loop: Zestimate Impact on List and Sale Prices \\ \cline{2-3} & A.3 & Backward Loop: Zestimate Calculation from Sale Prices \\ \hline \multirow{4}{*}{B} & & Support for analytical model simplifications \\ \cline{2-3} & B.1 & Buyer-Seller Bargaining \\ \cline{2-3} & B.2 & Full Model \\ \hline \multirow{4}{*}{C} & & Proofs \\ \cline{2-3} & C.1 & Proofs for Lemma \\ \cline{2-3} & C.2 & Proofs for Propositions \\ \hline \end{tabular} \end{table}

2302.09438v1

\begin{table}[] \centering \caption{Sample values of the house features, location, listing information, and Zestimate} \begin{tabular}{|l|l|l|} \hline \textbf{Category} & \textbf{Variable} & \textbf{Sample Value} \\ \hline Location & Latitude & 29.7–42.3 degrees North \\ \cline{2-3} & Longitude & 71.0–95.3 degrees West \\ \cline{2-3} & Neighborhood & South Boston, Carrick, Brighton Heights, etc. \\ \cline{2-3} & Zip Code & 15210, 15212, 15232, etc. \\ \cline{2-3} & County & Suffolk, Allegheny, Travis \\ \hline Features & Floor Size & 100–10,000 sq. ft. \\ \cline{2-3} & Year Built & 1799–2019 \\ \cline{2-3} & Last Remodel Year & 1799–2019 \\ \cline{2-3} & Bathrooms & 0–15 \\ \cline{2-3} & Bedrooms & 0–15 \\ \cline{2-3} & Parking & 0–1000 sq. ft. \\ \cline{2-3} & Lot & 100–10,000 sq. ft. \\ \cline{2-3} & Stories & 0–50 \\ \cline{2-3} & Solar Potential & 0–100 \\ \cline{2-3} & Type & Single Family, Multi-Family, Condo, etc. \\ \cline{2-3} & Structure Type & Colonial, Victorian, Modern, etc. \\ \cline{2-3} & Roof Type & Composition, Shingle, Asphalt, etc. \\ \cline{2-3} & Flooring & Hardwood, Carpeted, Tile, etc. \\ \cline{2-3} & Patio & Porch, Deck, None, etc. \\ \cline{2-3} & Ex-Material & Brick, Wood, Cement, etc. \\ \hline Listing & List Price & $10,000–$10,000,000 \\ \cline{2-3} & Days Listed & 1–365 days \\ \cline{2-3} & Sale Price & $10,000–$10,000,000 \\ \hline ML Price & Z (Zestimate) & $10,000–$10,000,000 \\ \hline \end{tabular} \label{table:datadesc} \end{table}

2302.09438v1

\begin{table}[] \centering \caption{Descriptive Statistics for key house features (subset of $\vec{X}$), Zestimate, List and Sale Prices for model of Zestimate impact on List and Sale Prices} \begin{tabular}{|l|r|r|r|r|} \hline \textbf{Variable} & \multicolumn{1}{l|}{\textbf{Mean}} & \multicolumn{1}{l|}{\textbf{Std. Dev}} & \multicolumn{1}{l|}{\textbf{Min}} & \multicolumn{1}{l|}{\textbf{Max}} \\ \hline Z (pre algorithm update) & 433737.6 & 393379.4 & 25065 & 8244491 \\ \hline Z (post algorithm update) & 433321.7 & 382211.1 & 21226 & 7911868 \\ \hline Z Error \% & 0.109 & 0.09 & -20 & 20 \\ \hline Z Confidence Interval \% & 17.3 & 10.34 & 10 & 130 \\ \hline Sale Price & 387602.5 & 376852.3 & 13000 & 4950000 \\ \hline List Price & 438729.8 & 390021 & 4800 & 6675000 \\ \hline Markup \% & 1.08 & 0.12 & -9.55 & 22.5 \\ \hline Time to Sale & 27.8 & 47.94 & 1 & 350 \\ \hline Floor Size & 1768.6 & 831.3 & 293 & 6000 \\ \hline Year Built & 1961 & 37.56 & 1799 & 2019 \\ \hline Bathrooms & 2.3 & 0.96 & 1 & 14 \\ \hline Bedrooms & 3.1 & 1.2 & 1 & 12 \\ \hline Parking & 156.6 & 220.55 & 0 & 995 \\ \hline Lot & 4396.4 & 5113.88 & 293 & 9000 \\ \hline Stories & 1.9 & 2.39 & 0 & 25 \\ \hline Last Remodel Year & 1973.2 & 39.91 & 0 & 2019 \\ \hline Solar Potential & 73 & 25.47 & 0 & 95.66 \\ \hline \end{tabular} \label{Table: PostPSMDescriptive} \end{table}

2302.09438v1

\begin{table}[] \centering \caption{Confusion Matrix and classification evaluation metrics (TPR, FPR, Accuracy, F1 Score) for the peer prediction model.} \begin{tabular}{|cc|cc|} \hline \multicolumn{2}{|c|}{\multirow{2}{*}{Confusion Matrix}} & \multicolumn{2}{c|}{Actual} \\ \cline{3-4} \multicolumn{2}{|c|}{} & \multicolumn{1}{c|}{Peer} & Not a Peer \\ \hline \multicolumn{1}{|c|}{\multirow{2}{*}{Predicted}} & Peer & \multicolumn{1}{c|}{4} & 16 \\ \cline{2-4} \multicolumn{1}{|c|}{} & Not a Peer & \multicolumn{1}{c|}{1} & 1379 \\ \hline \end{tabular} \label{Table: confusionmatrix} \end{table}

2302.09438v1

\begin{table}[] \centering \caption{Performance of logistic regression components $y_{i,j,t}$ within the overall predictive model} \begin{tabular}{|l|l|} \hline True Positive Rate & 80\% \\ \hline False Positive Rate & 1.14\% \\ \hline Precision & 20\% \\ \hline Accuracy & 98.7\% \\ \hline F1 Score & 0.32 \\ \hline \end{tabular} \label{Table: logitPerformance} \end{table}