Follow up question for #32

[yejiwi]

Hi, As suggested in the reply for #32, I modified the function safe_test

y_pred.append(model.predict(dde.nn.tensorflow_compat_v1.nn.NN.apply_feature_transform(self,X_add)))

but it gives an error saying

TypeError: apply_feature_transform() missing 1 required positional argument: 'transform'

First of all, I am not sure if this is how I am supposed to change the code based on the answer for #32.

If I put apply_feature_transform(self,X_add)),

and it gives

NameError: name 'self' is not defined

I see in nn.py, self is defined. What should I do in this case?

Please refer the full code below. Thanks.

def safe_test(model, data, X_test, y_test, fname=None):
       
    def is_nonempty(X):
        return len(X[0]) > 0 if isinstance(X, (list, tuple)) else len(X) > 0
    
    y_pred = []
    X = X_test
    while is_nonempty(X):
        X_add, X = trim_to_65535(X)
        
        #Original Code
        #y_pred.append(model.predict(data.transform_inputs(X_add))) 
        
        #Modified Code
        y_pred.append(model.predict(dde.nn.tensorflow_compat_v1.nn.NN.apply_feature_transform(self,X_add)))
        
                
    y_pred = np.vstack(y_pred)
    error = np.mean((y_test - y_pred) ** 2)
    print("Test MSE: {}".format(error))
    error = mean_squared_error_outlier(y_test, y_pred)
    print("Test MSE w/o outliers: {}\n".format(error))

    if fname is not None:
        np.savetxt(fname, np.hstack((X_test[1], y_test, y_pred)))

[xuliang5115]

# build a net then train it.
def periodic(inputs):  # just a function
  inputs *= 2 * np.pi
  out = tf.concat([tf.cos(inputs), tf.sin(inputs), tf.cos(2 * inputs), tf.sin(2 * inputs)], 1)
  return out
def main(_,_):
net = MIONetIn4CartesianProd(
      [m, w*2, w*2, m*2], [m, w*2, w*2, m*2],
      [m*_r, w*2, w*2, m*2], [m*_r, w*2, w*2, m*2],
      [2, w, w, m],
      activation,
      "Glorot normal")
  net.apply_feature_transform(periodic)  # without self

if model trained, just predict with model.preict(input_data), because input feature layer is in the graph of the net.

# so, code may be like this:
y_pred = model.predict(X_add)