Matlab use many training algos, like LM, CG, SD etc for ANN training. I am searching for the time complexity for LM & CG based ANN implementation in Matlab. specificly, What is the time complexity of LM & CG method implementd in Matlab. Are they faster than O(n3).
Thanks in advance for your help.
It depends on the network efficiency settings that control the speed/storage tradeoff
For example, the current algorithms for regression and classification are FITNET and PATTERNNET, respectively. BOTH call FEEDFORWARDNET.
The default settings are 1 but can be overwritten.
net = feedforwardnet;
% To verify unity defaults
efficiency = net.efficiency
cacheDelayedInputs = net.efficiency.cacheDelayedInputs
flattentime = net.efficiency.flattenTime
memoryReduction = net.efficiency.memoryReduction
% To overide defaults
net.efficiency.cacheDelayedInputs = cacheDelayedInputs0;
net.efficiency.flattenTime = flattentime0;
net.efficiency.memoryReduction = memoryReduction0;
Search the newsgroup and answers using
neural speed memory
Hope this helps.
Thank you for formally accepting my answer.
As a disclaimer, I have no background in neural networks. I do have some background in function optimization, though, and I think I can confidently say,