nnbench-v4 is a neural network based training program.
This program rquires:

1. Executable nnbench-v4
2. A file that tells where your trained parameters are.
	This file is called nnbench.rc
	If you have params directory in your working directory, you can create nnbench.rc by typing:
	echo $PWD/params > nnbench.rc
3. In params directory you should have
	tr_opt.dat
	netarch.dat
	wts.dat (will be created during training)
	bias.dat (will be created during training)
4. To train the network:
	1. edit params/netarch.dat 
	2. edir params/tr_opt.dat

	tr_opt.dat format:
			uncomment the lines which you want to use. You may optimize mse or auc. 
			You should prefer mse, unless you know the program very well.
	
			For training use backprop algorithm (recommended)
			although other options also work sometimes.

			Set maxepoch and minepoch as the number of epoch for which training is to be carried out.
			Set learning rate by changing number in front of lrate	(0.01 is recommended).

	If you want to contrinue further trainining from saved parameters, change cont_st to "part" or "old".

		DO NOT change/ delete other options unless you are sure.


