Out of memory error

Sir , My project is offline handwritten character recognition.I have some problem during train the network.Size of my feature matrix is 830X57.It will be changed based on my input .i cant predict the size of feature matrix.when i train this feature matrix i have got an error message "Out of memory". Is there any possible methods to reduce the size of matrix without any data loss.Or any other methods.plz help me..

i'm not good at Matlab.

Answer 1

[Jan]

Install more RAM. It is cheap and nothing can beat enough memory.

Answer 2

[Andreas Goser]

You should use a 64 bit MATLAB on a 64 bit OS.

If you already do it, you need to reduce the memory consumption by reducing your model.

Answer 3

[cr]

There could be many reasons for that error from poor memory management to demanding task. Consider posting your code. A lot is available on forums for handling memory management, please do a search. In general, some good memory management tips would be -

1. cut down use of double where not required.
2. clear big temporary variables right at the point they are no longer needed.
3. if your code uses heavy vectorization, you will need to probably reduce that and strike a balance between speed and memory.
4. if using simulink, make scopes not to log more than a specified number of points.
5. if input data is huge, restructure code to deal with smaller chunks of data.