def trainCnnPointEstimator(modelType):
hidden,measure = gm.loadData('Generated.Data/' + modelType + '.Train')
testHidden, testMeasure = gm.loadData('Generated.Data/' + modelType + '.Test')
cnnPointEstimator = nn.CnnPointEstimator(hidden.shape[1])
cnnPointEstimator.train(2e-4,100,200,measure,hidden,
'Trained.Models/CNN.Point.Estimator.' + modelType + '.ckpt',testMeasure,
testHidden,1213,showKalman=(modelType=='LG'))
def testCnnPointEstimator(modelType,sampleNo):
sampleNo -= 1
testHidden,testMeasure = gm.loadData('Generated.Data/' + modelType + '.Test')
cnnPointEstimator = nn.CnnPointEstimator(testHidden.shape[1])
estimated = cnnPointEstimator.infer(testMeasure[sampleNo],
'Trained.Models/CNN.Point.Estimator.' + modelType +
'/CNN.Point.Estimator.' + modelType + '.ckpt')
loss = cnnPointEstimator.computeLoss(estimated,testHidden[sampleNo],
'Trained.Models/CNN.Point.Estimator.' + modelType +
'/CNN.Point.Estimator.' + modelType + '.ckpt')
if modelType=='LG':
testKalmanZ,dump = ks.loadResults('Results.Data/' + modelType +
'.Kalman.Results')
plt.figure(figsize=(10,5))
plt.scatter(np.arange(testHidden.shape[1]),testHidden[sampleNo],
marker='o',color='blue',s=4)
if modelType=='LG': plt.plot(testKalmanZ[sampleNo],color='green')
plt.plot(estimated.flatten(),color='red')
plt.show()
