In [11]:

    

from systemml import MLContext, dml, jvm_stdout
ml = MLContext(sc)
ml.version()


    

    
Out[11]:





u'1.0.0-SNAPSHOT'

In [18]:

    

script="""
source("nn/layers/cross_entropy_loss.dml") as cross_entropy_loss
source("nn/layers/l2_loss.dml") as l2_loss

source("nn/layers/lstm.dml") as lstm
source("nn/layers/sigmoid.dml") as sigmoid

source("nn/optim/sgd_nesterov.dml") as sgd_nesterov
source("nn/optim/rmsprop.dml") as rmsprop

X = seq(0, 999, 1) * 3.14 / 100 #0 to 10 pi radians

X = sin(X)
#print(toString(X))

in_TS = 75 #in_TS
out_TS = 25 #out_TS
N = nrow(X) / (in_TS + out_TS)
M = out_TS

idx_mat = outer(seq(0,N-1,1), t(seq(0,in_TS+out_TS-1,1)), "+") + 1
#print(toString(idx_mat))

idx_col = matrix(idx_mat, rows=nrow(idx_mat)*ncol(idx_mat), cols=1)
rordrd_X = table(seq(1, nrow(idx_col), 1), idx_col, nrow(idx_col), nrow(idx_col)) %*% X

X = matrix(rordrd_X, rows=nrow(idx_mat), cols=ncol(idx_mat))
#print(toString(X))

Y = X[,in_TS+1:in_TS+out_TS]
X = X[,1:in_TS]

max_iterations = 2000
iter = 0
learningRate = 0.01
decayRate = 0.95

[W, b, out0, c0] = lstm::init(N,1,M)
rmspropCache = rmsprop::init(W)

while( iter < max_iterations ){
    [a1, c, c_out, c_c, c_ifog] = lstm::forward(X, W, b, in_TS, 1, FALSE, out0, c0)
    loss = l2_loss::forward(a1, Y)

    if(iter %% 100 == 0) print("iter=" + iter + " loss=" + loss)

    loss_grad = l2_loss::backward(a1, Y)
    [dX, dW, db, dout0, dc0] = lstm::backward(loss_grad, c0, X, W, b, in_TS, 1, FALSE, out0, c0, c_out, c_c, c_ifog)

    [W, rmspropCache] = rmsprop::update(W, dW, learningRate, decayRate, 1e-6, rmspropCache)

    iter = iter + 1
}

[a1, c, c_out, c_c, c_ifog] = lstm::forward(X, W, b, in_TS, 1, FALSE, out0, c0)
print(toString(cbind(a1, Y)))

"""


    

In [19]:

    

prog = dml(script).output("a1").output("X").output("Y")
with jvm_stdout(True):
    result = ml.execute(prog)


    

    




iter=0 loss=1.482576498588268
iter=100 loss=0.1612714328013795
iter=200 loss=0.058938287626221386
iter=300 loss=0.04897157601036786
iter=400 loss=0.031392670989147206
iter=500 loss=0.03271373060275306
iter=600 loss=0.029034160789182445
iter=700 loss=0.0377526211023614
iter=800 loss=0.009473446532038406
iter=900 loss=0.021804000351799663
iter=1000 loss=0.032295088972713305
iter=1100 loss=0.016885715019442922
iter=1200 loss=0.032142600672863356
iter=1300 loss=0.011399926950937665
iter=1400 loss=0.009869723200786508
iter=1500 loss=0.014297125319484428
iter=1600 loss=0.015026303897219703
iter=1700 loss=0.010265573574590854
iter=1800 loss=0.16290252619536108
iter=1900 loss=0.0019342772124357182
0.701 0.670 0.657 0.628 0.607 0.592 0.557 0.541 0.504 0.478 0.449 0.431 0.392 0.361 0.335 0.302 0.289 0.247 0.235 0.204 0.143 0.164 0.134 0.084 -0.010 0.708 0.685 0.662 0.638 0.614 0.589 0.563 0.537 0.510 0.483 0.455 0.427 0.398 0.369 0.340 0.310 0.280 0.250 0.220 0.189 0.158 0.127 0.096 0.064 0.033
0.691 0.658 0.647 0.619 0.595 0.577 0.546 0.524 0.489 0.465 0.435 0.410 0.377 0.347 0.321 0.291 0.273 0.232 0.213 0.185 0.130 0.139 0.105 0.070 -0.014 0.685 0.662 0.638 0.614 0.589 0.563 0.537 0.510 0.483 0.455 0.427 0.398 0.369 0.340 0.310 0.280 0.250 0.220 0.189 0.158 0.127 0.096 0.064 0.033 0.002
0.679 0.643 0.634 0.608 0.582 0.559 0.534 0.505 0.472 0.448 0.420 0.388 0.361 0.331 0.304 0.278 0.254 0.215 0.189 0.163 0.114 0.113 0.075 0.051 -0.019 0.662 0.638 0.614 0.589 0.563 0.537 0.510 0.483 0.455 0.427 0.398 0.369 0.340 0.310 0.280 0.250 0.220 0.189 0.158 0.127 0.096 0.064 0.033 0.002 -0.030
0.664 0.626 0.619 0.595 0.566 0.539 0.519 0.484 0.453 0.429 0.402 0.365 0.341 0.312 0.285 0.261 0.233 0.197 0.162 0.138 0.094 0.085 0.043 0.027 -0.028 0.638 0.614 0.589 0.563 0.537 0.510 0.483 0.455 0.427 0.398 0.369 0.340 0.310 0.280 0.250 0.220 0.189 0.158 0.127 0.096 0.064 0.033 0.002 -0.030 -0.061
0.646 0.606 0.601 0.578 0.547 0.516 0.500 0.460 0.430 0.406 0.381 0.341 0.319 0.291 0.262 0.240 0.208 0.175 0.133 0.111 0.071 0.054 0.009 -0.002 -0.039 0.614 0.589 0.563 0.537 0.510 0.483 0.455 0.427 0.398 0.369 0.340 0.310 0.280 0.250 0.220 0.189 0.158 0.127 0.096 0.064 0.033 0.002 -0.030 -0.061 -0.092
0.625 0.583 0.579 0.557 0.525 0.490 0.476 0.433 0.404 0.380 0.357 0.315 0.292 0.266 0.236 0.214 0.180 0.150 0.102 0.080 0.043 0.021 -0.026 -0.036 -0.057 0.589 0.563 0.537 0.510 0.483 0.455 0.427 0.398 0.369 0.340 0.310 0.280 0.250 0.220 0.189 0.158 0.127 0.096 0.064 0.033 0.002 -0.030 -0.061 -0.092 -0.124
0.600 0.557 0.553 0.531 0.499 0.461 0.445 0.404 0.375 0.349 0.326 0.287 0.262 0.237 0.204 0.181 0.147 0.120 0.069 0.046 0.011 -0.013 -0.061 -0.074 -0.082 0.563 0.537 0.510 0.483 0.455 0.427 0.398 0.369 0.340 0.310 0.280 0.250 0.220 0.189 0.158 0.127 0.096 0.064 0.033 0.002 -0.030 -0.061 -0.092 -0.124 -0.155
0.570 0.529 0.522 0.499 0.466 0.430 0.408 0.373 0.343 0.314 0.289 0.258 0.228 0.203 0.169 0.142 0.113 0.085 0.037 0.012 -0.024 -0.048 -0.092 -0.111 -0.116 0.537 0.510 0.483 0.455 0.427 0.398 0.369 0.340 0.310 0.280 0.250 0.220 0.189 0.158 0.127 0.096 0.064 0.033 0.002 -0.030 -0.061 -0.092 -0.124 -0.155 -0.186
0.537 0.499 0.489 0.462 0.428 0.399 0.370 0.341 0.309 0.279 0.251 0.229 0.192 0.166 0.132 0.103 0.078 0.047 0.008 -0.020 -0.058 -0.079 -0.115 -0.143 -0.157 0.510 0.483 0.455 0.427 0.398 0.369 0.340 0.310 0.280 0.250 0.220 0.189 0.158 0.127 0.096 0.064 0.033 0.002 -0.030 -0.061 -0.092 -0.124 -0.155 -0.186 -0.216
0.505 0.473 0.459 0.425 0.393 0.372 0.339 0.310 0.278 0.246 0.220 0.203 0.157 0.131 0.097 0.069 0.048 0.012 -0.017 -0.046 -0.087 -0.103 -0.127 -0.165 -0.202 0.483 0.455 0.427 0.398 0.369 0.340 0.310 0.280 0.250 0.220 0.189 0.158 0.127 0.096 0.064 0.033 0.002 -0.030 -0.061 -0.092 -0.124 -0.155 -0.186 -0.216 -0.247

SystemML Statistics:
Total execution time:		66.906 sec.
Number of executed Spark inst:	0.

In [20]:

    

X = result.get("X").toNumPy()
a1 = result.get("a1").toNumPy()


    

In [21]:

    

import matplotlib.pyplot as plt
import numpy as np


    

In [22]:

    

complete = np.vstack((np.transpose(X),np.transpose(a1)))


    

In [23]:

    

plt.plot(complete)
plt.ylabel('some numbers')
plt.show()