from LR import Logisticlayer
from mlp import MLP
if __name__=="__main__":
numpy.set_printoptions(threshold=numpy.nan)
input_dim = 4
output_dim = 3
sample_size = 100
#X=numpy.random.normal(0,1,(sample_size,input_dim))
#temp,Y=numpy.nonzero(numpy.random.multinomial(1,[1.0/output_dim]*output_dim,size=sample_size))
mlp = MLP(4,3,[10,10])
with open('debug_nnet.pickle') as f:
init_param = pickle.load(f)
init_param = numpy.concatenate([i.flatten() for i in init_param])
mlp.packParam(init_param)
with open('debug_data.pickle') as f:
data = pickle.load(f)
X = data[0]
Y = data[1]
with open('HJv.pickle') as f:
HJv_theano = pickle.load(f)
num_param = numpy.sum(mlp.sizes)
batch_size = 100
grad,train_nll,train_error=mlp.get_gradient(X,Y,batch_size)
d = 1.0*numpy.ones((num_param,))
delta, next_init, after_cost = mlp.cg(-grad, train_cg_X_cur, train_cg_Y_cur, batch_size, next_init, 1)
Gv = mlp.get_Gv(train_cg_X_cur,train_cg_Y_cur,batch_size,delta)
delta_cost = numpy.dot(delta,grad+0.5*Gv)
before_cost = mlp.quick_cost(numpy.zeros((num_param,)), train_cg_X_cur, train_cg_Y_cur, batch_size)
l2norm = numpy.linalg.norm(Gv + mlp._lambda*delta + grad)
print "Residual Norm: ",l2norm
print 'Before cost: %f, After cost: %f'%(before_cost,after_cost)
param = mlp.flatParam() + delta
mlp.packParam(param)
tune_lambda = (after_cost - before_cost)/delta_cost
if tune_lambda < 0.25:
mlp._lambda = mlp._lambda*1.5
elif tune_lambda > 0.75:
mlp._lambda = mlp._lambda/1.5
print "Training NNL: %f, Error: %f"%(train_nll,train_error)
nll=[]
error=[]
for batch_index in xrange(n_valid_batches):
X=valid_X[batch_index*batch_size:(batch_index+1)*batch_size,:]
Y=valid_Y[batch_index*batch_size:(batch_index+1)*batch_size]
mlp.forward(X)