Training plot taking very long to run

2 views (last 30 days)
Nathaniel Porter
Nathaniel Porter on 21 Dec 2021
Answered: yanqi liu on 21 Dec 2021
How can I improve my network to run faster and use less memory.
clc; clear all; close all;
%Import/Upload data
load generated_data.mat
% change to label vector
CS = categories(categorical(Y1));
Z1 = []; Z2 = [];
for i = 1 : length(Y1)
Z1(i,1) = find(Y1(i)==CS);
end
for i = 1 : length(Y2)
Z2(i,1) = find(Y2(i)==CS);
end
Yo1 = Y1;
Yo2 = Y2;
Y1 = Z1;
Y2 = Z2;
%transposing glucose data
X1_T = X1';
%Shuffling data to take randomly
rand('seed', 0)
ind = randperm(size(X1_T, 1));
X1_T = X1_T(ind, :);
Y1 = Y1(ind);
%Separating data in training, validation and testing data
X1_train = X1_T;
%Partioning data for training 70%
train_X1 = X1_train(1:120,:);
%Corresponding X(input) data to Y(output) data
train_Y1 = Y1(1:120);
%reshaping data into 4D array
XTrain=(reshape(train_X1', [2289,1,1,120]));
%Separating and partioning for validation data 15%
val_X1 = X1_train(121:150,:);
%Corresponding X(input) data to Y(output) data
val_Y1 = Y1(121:150);
%reshaping data into 4D array
XVal=(reshape(val_X1', [2289,1,1,30])); %Train data
%Separating and partioning for test data 15%
test_X1 = X1_train(151:180,:);
%Corresponding X(input) data to Y(output) data
test_Y1 = Y1(151:180);
%reshaping data into 4D array
XTest=(reshape(test_X1', [2289,1,1,30])); %Train data
%% NETWORK ARCHITECTURE
layers = [imageInputLayer([2289 1 1]) % Creating the image layer
convolution2dLayer([102 1],3,'Stride',1)
batchNormalizationLayer
reluLayer
maxPooling2dLayer(2,'Stride',2,'Padding',[0 0 0 1])
dropoutLayer
fullyConnectedLayer(1)
regressionLayer];
% Specify training options.
opts = trainingOptions('adam', ...
'MaxEpochs',1000, ...
'Shuffle','every-epoch', ...
'Plots','training-progress', ...
'Verbose',false, ...
'ValidationData',{XVal,val_Y1},...
'LearnRateDropFactor',0.2,...
'LearnRateDropPeriod',5,...
'ExecutionEnvironment', 'cpu', ...
'ValidationPatience',Inf);
%% Train network
%net = trainNetwork(XTrain,Trainoutfinal,layers,opts);
yc = train_Y1(:);
net1 = trainNetwork(XTrain,yc,layers,opts);
%% Compare against testing Data
Ypredicted = predict(net1, XTest)
predictionError = test_Y1 - Ypredicted;
squares = predictionError.^2;
rmse = sqrt(mean(squares))
figure
scatter(Ypredicted, test_Y1,'+')
title ('True value vs Predicted Value')
xlabel ("Predicted Value")
ylabel ("True Value")
hold on
plot([-3 3], [-7 7], 'b--')

Accepted Answer

KSSV
KSSV on 21 Dec 2021
You don't plot the progress of training..it will eat away lot of time:
'Plots','training-progress',
USe
'Plots','none',
You can save the progress into a variable and check at the end:
[net1,net1_info] = trainNetwork(XTrain,yc,layers,opts);

More Answers (1)

yanqi liu
yanqi liu on 21 Dec 2021
yes,sir,if got gpu device,may be use gpu to run train

Products


Release

R2021b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by