PC freeze at the start of neural network training

4 views (last 30 days)
Hello All,
I am training a 2 hidden layers neural network to predict a single output as a function of 8 inputs of 5425 sample each. I have tested different neurons number successfully until I reach 200 neurons in each hidden layer when the MATLAB gives me an "Out of memory" error.
Then, I have increased my DDR3 RAM from 16 GB to 24 GB. I haven't received the out of memory error but the PC freeze at the start of the training with no response for minutes so I have to unplug it !!!!!!!!!!!!
Is it the RAM or a processor or what please ???
MY PC has an intel core i7-3770-3.4 GHz and an intel DQ77MK motherboard and a GTx 650 Ti - 1 GB - GDDR5 and 16 GB DDR3 10600U RAM 1333 MHz (Upgraded to 24 GB as explained earlier).
My code is very simple:
net = feedforwardnet([200,200],'trainlm');
net.divideParam.trainRatio = 70/100;
net.divideParam.valRatio = 15/100;
net.divideParam.testRatio = 15/100;
[net,tr] = train(net,st_inputs5,st_output5);
y = net(st_inputs5); %Estimate the targets using the trained network.
perf = perform(net,y,st_output5);
time = mean(tr.time);
mse = mean(((st_output5(:)-y(:)).^2));
rmse=sqrt(mse);
R222 = 1 - sum((st_output5 - y).^2)/sum((st_output5 - mean(st_output5)).^2);
mae = mean(abs(((st_output5(:)-y(:)))));
tr.num_epochs;

Accepted Answer

Ameen Bassam
Ameen Bassam on 24 Mar 2022
Problem solved. I have upgraded my RAM to 32 GB RAM with a higher transfer rate of 1600 MHz instead of the old 1333 MHz and installed an SSD instead of the HDD. Training takes longtime but it works.

More Answers (0)

Products


Release

R2019b

Community Treasure Hunt

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

Start Hunting!