Continue to Site

Welcome to

Welcome to our site! is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

need help backpropagation algorithm

Not open for further replies.


Junior Member level 2
Oct 30, 2006
Reaction score
Trophy points
Activity points
i am doing my final year project on Brain computer interface, and here in want to classify the 4 brain states,.

i want to know how much hidden neurons i take ,how to assign weights, how mucjh input and output vector should i have to take , for this typical case..

if anyone knows about this all , please help me in making this algo... i urgently requried help ...............\

best regards

Please can you explain more what you want to do. What I understand is that you want to make a computer interface for Brain imaging? so what type of imaging you want? ultrasound or microwave? just explain a little more


every brain signal , is specific for every mental task, if a person imagine to move his right hand , for this a respectve effect on the input data and its shape , also then if he imagine for the left hand movement this a corresponding effect on the input signal , which is different then first .. so this is the classification procedure for the Brain ( EEG) input data .

i want to classify my data with back propagation algorithm.

Q: how it is possible? please help me , if i wan to apply it in the back propagation algorithm, how much neuron requireds, layes, hidden layers

hoping help from every body .... feel free share your ideas...........

best regards

For the number of neuron of input/output layer, this depends on your input/output format. How you represent your input signal and output? In binary? Or any other forms? For example, your input is 8 bit binary data, then your input will be 8 neurons.

For number of hidden layers and hidden neurons, you need to test it out to get the best architecture. There is no fix architecture, as different application needs different architecture to get the best result.

You can try with 1 hidden layer and hidden neurons 2/3 times of your input neuron (if you have few input neurons) or 2/3 times less of your input neuron (if your input is a lot, eg 1000). Then slowly change the number of neuron to test the result is improving or otherwise.

Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to