sftwolayer (4.0)
index
user/fomels/Mtwolayer.py
Two layer NN training

 
Synopsis
        sftwolayer < dat.rsf label=lbl.rsf valdata=valdat.rsf vallabel=vallbl.rsf weight1=wt1.rsf weight2=wt2.rsf bias1=bs1.rsf bias2=bs2.rsf > loss.rsf weight1out=wt1out.rsf weight2out=wt2out.rsf bias1out=bs1out.rsf bias2out=bs2out.rsf valloss=valloss.rsf lr= niter= act= opt= seed= stop= lossfunc= reg= alpha=

 
Parameters
       
 
float act=
Activation function - 0:sigmoid 1:tanh 2:relu 3:identity
 
float alpha=
Regularization coeff. If not, set alpha=0
 
file bias1=
auxiliary input file name
 
file bias1out=
auxiliary output file name
 
file bias2=
auxiliary input file name
 
file bias2out=
auxiliary output file name
 
file label=
auxiliary input file name
 
float lossfunc=
Loss function - 0:MSE 1:L1
 
float lr=

 
float niter=

 
float opt=
Optimization method - 0:SGD 1:momentum 2:Adam
 
float reg=
Regularization - 0:L2 1:L1
 
float seed=

 
float stop=

 
file valdata=
auxiliary input file name
 
file vallabel=
auxiliary input file name
 
file valloss=
auxiliary output file name
 
file weight1=
auxiliary input file name
 
file weight1out=
auxiliary output file name
 
file weight2=
auxiliary input file name
 
file weight2out=
auxiliary output file name