Neural Networks in Matlab. Adding large learning patterns -


i new matlab , can't find solution problem...

what problem?

i have create neural network using matlab have 25k inputs , 10 outputs. there 300 patterns learn.

when reading info neural networks in matlab saw input/learing data in 1 matrix. it's ok xor or small. realized have create matrix contains 25 000 * 300 elements (7,5 mln of integers).

1) there possibility can expand matrix adding new rows (learning patterns)?

2) or maybe got like:

learnpatternmatrix1 = [1, 2, 3 , ..., 25 000]; perfectoutputmatrix1 = [1, 2, 3, ... , 10];  network.addpattern(learnpatternmatrix1, perfectoutputmatrix1); network.addpattern(learnpatternmatrix2, perfectoutputmatrix2); % ... network.addpattern(learnpatternmatrix300, perfectoutputmatrix300); network.learn()? 

thanks ;)

i'm sorry don't have answer making matlab deal size of matrix. have comments may relevant problem, however.

neural networks are, machine learning algorithms, unlikely perform when there large number of features (inputs) compared number of data points. unless have order or 2 magnitude more data points 250,000 features describe, approach may not work. seem have 300 cases. support vector machines, supposedly robust problem, unlikely perform under these conditions.

in case of not enough data number of features, can think of guaranteed overfitting, each data point uniquely situated , separated in feature space.

have considered feature reduction? solve matlab problem, , improve performance of ann.


Comments

Popular posts from this blog

java - Jmockit String final length method mocking Issue -

asp.net - Razor Page Hosted on IIS 6 Fails Every Morning -

c++ - wxwidget compiling on windows command prompt -