#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
@author: Maziar Raissi
"""
import autograd.numpy as np
def stochastic_update_Adam(w,grad_w,mt,vt,lrate,iteration):
beta1 = 0.9;
beta2 = 0.999;
epsilon = 1e-8;
mt = mt*beta1 + (1.0-beta1)*grad_w;
vt = vt*beta2 + (1.0-beta2)*grad_w**2;
mt_hat = mt/(1.0-beta1**iteration);
vt_hat = vt/(1.0-beta2**iteration);
scal = 1.0/(np.sqrt(vt_hat) + epsilon);
w = w - lrate*mt_hat*scal;
return w,mt,vt
def Normalize(X, X_m, X_s):
return (X-X_m)/(X_s)
def Denormalize(X, X_m, X_s):
return X_s*X + X_m
def fetch_minibatch_rnn(X,Y,N_batch):
N = X.shape[1]
idx = np.random.choice(N, N_batch, replace=False)
X_batch = X[:,idx,:]
Y_batch = Y[idx,:]
return X_batch, Y_batch
def activation(x):
return np.tanh(x)
def sigmoid(x):
return 0.5*(np.tanh(x) + 1.0)
def create_dataset(data, lags):
N = len(data)-lags
data_dim = data.shape[1]
X = np.zeros((lags, N, data_dim))
Y = np.zeros((N, data_dim))
for i in range(0,N):
X[:,i,:] = data[i:(i+lags), :]
Y[i,:] = data[i + lags, :]
return X, Y
