e822a02daa1cd1c42f6d2bf309d5996a21979ef1,qucumber/tests/test_grads_positive.py,,algorithmic_gradKL,#Any#Any#Any#,75

Before Change

    Z = partition(nn_state,vis)
    for i in range(len(vis)):
        for rbmType in nn_state.gradient(vis[i]):
            for pars in nn_state.gradient(vis[i])[rbmType]:
                grad_KL[rbmType][pars] += ((target_psi[i,0])**2)*nn_state.gradient(vis[i])[rbmType][pars]            
                grad_KL[rbmType][pars] -= probability(nn_state,vis[i], Z)*nn_state.gradient(vis[i])[rbmType][pars]
    return grad_KL            

def algorithmic_gradNLL(qr,data,k):
    qr.nn_state.set_visible_layer(data)

After Change

    grad_KL = torch.zeros(nn_state.rbm_am.num_pars,dtype=torch.double)
    for i in range(len(vis)):
        grad_KL += ((target_psi[i,0])**2)*nn_state.gradient(vis[i]) 
        grad_KL -=probability(nn_state,vis[i], Z)*nn_state.gradient(vis[i])
        //for rbmType in nn_state.gradient(vis[i]):
        //    for pars in nn_state.gradient(vis[i])[rbmType]:
        //        grad_KL[rbmType][pars] += ((target_psi[i,0])**2)*nn_state.gradient(vis[i])[rbmType][pars]            
        //        grad_KL[rbmType][pars] -= probability(nn_state,vis[i], Z)*nn_state.gradient(vis[i])[rbmType][pars]

Instances