Added experiments and figures

main.tex

@@ -253,34 +253,37 @@ We solve this problem by decoupling the gradient averaging from the weight avera
    \end{algorithmic}
 \end{algorithm}
 
-
-
 \section{Evaluation}
 
 \subsection{MNIST and Linear Model}
 
      \begin{figure}[htbp]
      \centering
-% To regenerate the figure, from directory results/mnist
-% python ../../../learn-topology/tools/plot_convergence.py 1-node-iid/all/2021-03-10-09:20:03-CET fully-connected/all/2021-03-10-09:25:19-CET clique-ring/all/2021-03-10-10:15:24-CET fully-connected-cliques/all/2021-03-10-10:19:44-CET --add-min-max --yaxis training-loss --labels '1-node IID bsz=12800' '100-nodes non-IID fully-connected bsz=128' '100-nodes non-IID D-Cliques (Ring) bsz=128' '100-nodes non-IID D-Cliques (Fully-Connected) bsz=128' --legend 'upper right' --save-figure ../../figures/d-cliques-mnist-vs-1-node-training-loss.png
-%     \begin{subfigure}[b]{0.70\textwidth}
-%         \centering
-%         \includegraphics[width=\textwidth]{figures/d-cliques-mnist-vs-1-node-training-loss}
-%\caption{\label{fig:d-cliques-mnist-training-loss} Training Loss}
-%     \end{subfigure}
+
 % To regenerate the figure, from directory results/mnist
 % python ../../../learn-topology/tools/plot_convergence.py 1-node-iid/all/2021-03-10-09:20:03-CET fully-connected/all/2021-03-10-09:25:19-CET clique-ring/all/2021-03-10-18:14:35-CET fully-connected-cliques/all/2021-03-10-10:19:44-CET --add-min-max --yaxis test-accuracy --labels '1-node IID bsz=12800' '100-nodes non-IID fully-connected bsz=128' '100-nodes non-IID D-Cliques (Ring) bsz=128' '100-nodes non-IID D-Cliques (Fully-Connected) bsz=128' --legend 'lower right' --ymin 85 --ymax 92.5 --save-figure ../../figures/d-cliques-mnist-vs-1-node-test-accuracy.png
-     %\begin{subfigure}[b]{\textwidth}
-         %\centering
+         \centering
          \includegraphics[width=0.7\textwidth]{figures/d-cliques-mnist-vs-1-node-test-accuracy}
-%\caption{\label{fig:d-cliques-mnist-linear} D-Cliques with Linear Model on MNIST.}
-     %\end{subfigure}
-\caption{\label{fig:d-cliques-mnist-linear} D-Cliques with Linear Model on MNIST.}
-\end{figure}
+         \caption{\label{fig:d-cliques-mnist-linear-w-clique-averaging-w-initial-averaging} MNIST: D-Cliques Convergence Speed}
+        \end{figure}
 
-With and without clique averaging.
+     
+     
+    \begin{figure}[htbp]
+     \centering   
+      \begin{subfigure}[b]{0.48\textwidth}
+         \centering
+         %\includegraphics[width=0.7\textwidth]{figures/d-cliques-mnist-vs-1-node-test-accuracy}
+         \caption{\label{fig:mnist-init-clique-avg-effect-ring} Ring}
+     \end{subfigure}
+       \begin{subfigure}[b]{0.48\textwidth}
+         \centering
+         %\includegraphics[width=0.7\textwidth]{figures/d-cliques-mnist-vs-1-node-test-accuracy}
+         \caption{\label{fig:mnist-init-clique-avg-effect-fcc} Fully-Connected-Cliques}
+     \end{subfigure}
+\caption{\label{fig:d-cliques-mnist-initialization-effect} MNIST: Effects of Clique Averaging and Uniform Initialization on Convergence Speed.}
+\end{figure}
 
-TODO: Update figure with actual Clique-Ring results
 
 \subsection{CIFAR10 and Convolutional Model}
 
@@ -331,17 +334,22 @@ Similar number of maximum hops but no or less clustering than D-Cliques (and no
 
 \subsection{Effect of Scaling}
 
-% To regenerate the figure, from directory results/scaling
-% python ../../../learn-topology/tools/plot_convergence.py 10/mnist/fully-connected-cliques/all/2021-03-10-14:40:35-CET ../mnist/fully-connected-cliques/all/2021-03-10-10:19:44-CET 1000/mnist/fully-connected-cliques/all/2021-03-10-16:44:35-CET --labels '10 nodes bsz=128' '100 nodes bsz=128' '1000 nodes bsz=128 (45)' --legend 'lower right' --yaxis test-accuracy --save-figure ../../figures/d-cliques-mnist-scaling-fully-connected.png --ymin 80 --add-min-max
+
      \begin{figure}[htbp]
          \centering
-         \begin{subfigure}[b]{\textwidth}
+         % To regenerate the figure, from directory results/scaling
+% python ../../../learn-topology/tools/plot_convergence.py 10/mnist/fully-connected-cliques/all/2021-03-10-14:40:35-CET ../mnist/fully-connected-cliques/all/2021-03-10-10:19:44-CET 1000/mnist/fully-connected-cliques/all/2021-03-10-16:44:35-CET --labels '10 nodes bsz=128' '100 nodes bsz=128' '1000 nodes bsz=128 (45)' --legend 'lower right' --yaxis test-accuracy --save-figure ../../figures/d-cliques-mnist-scaling-fully-connected-cst-bsz.png --ymin 80 --add-min-max
+         \begin{subfigure}[b]{0.48\textwidth}
          \centering
-         \includegraphics[width=0.7\textwidth]{figures/d-cliques-mnist-scaling-fully-connected}
+         \includegraphics[width=\textwidth]{figures/d-cliques-mnist-scaling-fully-connected-cst-bsz}
          \caption{Constant Batch-Size}
      \end{subfigure}
-      \begin{subfigure}[b]{\textwidth}
+              % To regenerate the figure, from directory results/scaling
+% python ../../../learn-topology/tools/plot_convergence.py 10/mnist/fully-connected-cliques/all/2021-03-12-09:13:27-CET ../mnist/fully-connected-cliques/all/2021-03-10-10:19:44-CET 1000/mnist/fully-connected-cliques/all/2021-03-10-16:44:35-CET --labels '10 nodes bsz=1280' '100 nodes bsz=128' '1000 nodes bsz=13' --legend 'lower right' --yaxis test-accuracy --save-figure ../../figures/d-cliques-mnist-scaling-fully-connected-cst-updates.png --ymin 80 --add-min-max
+\hfill
+      \begin{subfigure}[b]{0.48\textwidth}
          \centering
+         \includegraphics[width=\textwidth]{figures/d-cliques-mnist-scaling-fully-connected-cst-updates}
          \caption{Constant Nb Updates per Epoch}
      \end{subfigure}
      \caption{\label{fig:d-cliques-mnist-scaling-fully-connected} Scaling Behaviour of Fully-Connected D-Clique}

results/mnist/no-clique-avg/clique-ring/experiments.sh

+#!/usr/bin/env bash
+TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS
+BSZS='
+    128
+    '
+LRS='
+    0.1
+    '
+for BSZ in $BSZS; 
+    do for LR in $LRS;
+        do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 --dist-optimization d-psgd --topology clique-ring --metric dissimilarity --learning-momentum 0. --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --parallel-training --nb-workers 10 --dataset mnist --model linear --initial-averaging
+    done;
+done;
+

results/mnist/no-clique-avg/fully-connected-cliques/experiments.sh

+#!/usr/bin/env bash
+TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS
+BSZS='
+    128
+    '
+LRS='
+    0.1
+    '
+for BSZ in $BSZS; 
+    do for LR in $LRS;
+        do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 --dist-optimization d-psgd --topology fully-connected-cliques --metric dissimilarity --learning-momentum 0. --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --parallel-training --nb-workers 10 --dataset mnist --model linear --initial-averaging
+    done;
+done;
+

results/mnist/no-init-no-clique-avg/clique-ring/experiments.sh

+#!/usr/bin/env bash
+TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS
+BSZS='
+    128
+    '
+LRS='
+    0.1
+    '
+for BSZ in $BSZS; 
+    do for LR in $LRS;
+        do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 --dist-optimization d-psgd --topology clique-ring --metric dissimilarity --learning-momentum 0. --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --parallel-training --nb-workers 10 --dataset mnist --model linear
+    done;
+done;
+

results/mnist/no-init-no-clique-avg/fully-connected-cliques/experiments.sh

+#!/usr/bin/env bash
+TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS
+BSZS='
+    128
+    '
+LRS='
+    0.1
+    '
+for BSZ in $BSZS; 
+    do for LR in $LRS;
+        do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 --dist-optimization d-psgd --topology fully-connected-cliques --metric dissimilarity --learning-momentum 0. --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --parallel-training --nb-workers 10 --dataset mnist --model linear
+    done;
+done;
+

results/mnist/no-init/clique-ring/experiments.sh

+#!/usr/bin/env bash
+TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS
+BSZS='
+    128
+    '
+LRS='
+    0.1
+    '
+for BSZ in $BSZS; 
+    do for LR in $LRS;
+        do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 --dist-optimization d-psgd --topology clique-ring --metric dissimilarity --learning-momentum 0. --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --parallel-training --nb-workers 10 --dataset mnist --model linear --clique-gradient
+    done;
+done;
+

results/mnist/no-init/fully-connected-cliques/experiments.sh

+#!/usr/bin/env bash
+TOOLS=../../../../learn-topology/tools; CWD="$(pwd)"; cd $TOOLS
+BSZS='
+    128
+    '
+LRS='
+    0.1
+    '
+for BSZ in $BSZS; 
+    do for LR in $LRS;
+        do python sgp-mnist.py --nb-nodes 100 --nb-epochs 100 --local-classes 1 --seed 1 --nodes-per-class 10 10 10 10 10 10 10 10 10 10 --global-train-ratios 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 0.802568 --dist-optimization d-psgd --topology fully-connected-cliques --metric dissimilarity --learning-momentum 0. --sync-per-mini-batch 1 --results-directory $CWD/all --learning-rate $LR --batch-size $BSZ "$@" --parallel-training --nb-workers 10 --dataset mnist --model linear --clique-gradient
+    done;
+done;
+