Python: 'numpy.float64' object is not callable

Question

The code below throws the error 'numpy.float64' object is not callable at last_mae = mae(val_scaled_price_client, cv_model). Nevertheless the mae function with the same parameters works just fine when used outside the loop.

loss_history = [1000.00]

for i in range(10000):

  # train
  iterations = 10

  train_auto_encoder(train_latent_customers=train_latentvars,
                   train_product_customers=train_scaled_price_client,
                   auto_encoder=auto_model,
                   iters=iterations,                              
                   batch_size=128,
                   display_step=20)                                  

  cv_model = auto_model.predict([val_latentvars, val_scaled_price_client_corrupted])
  last_mae = mae(val_scaled_price_client, cv_model)          
  loss_history.append(last_mae)

  if loss_history[-1] < loss_history[-2]:
    iterations += 10

  else:
    break

I declared a function mae in previous cells as follow

# define function to calculate MAE between true and reconstructed values
def mae(y_true, y_pred):

    # get non-zero positions
    cond = np.not_equal(y_true, 0)
    # get number of non-zero elements
    num_non_zero = np.sum(cond)
    # initialize zer matrix
    zero_matrix = np.zeros(shape=y_true.shape)
    # replace 
    predictions_corrected = np.where(cond, y_pred, zero_matrix)
    # get rmse
    mae = np.sum(np.abs(y_true - predictions_corrected)) / num_non_zero
    # return
    return(mae)

It seems the issue is due to the type conversion error. Is there a specific reason why you are declaring loss_history as a float. — Gokul Alex
– Gokul Alex, Commented Jun 14, 2020 at 11:25
mae is a number, you use it as a function. Most probably you unintentionally overwrote you function definition of mae with a number. — Stef
– Stef, Commented Jun 14, 2020 at 11:27
The reason I declared loss_history as a float is just because the mae function returns a float and I thought that might have been the problem. @Stef, I edited the post since I did define a mae function before. — Nele
– Nele, Commented Jun 14, 2020 at 11:58
As I said: you redefine mae in mae = np.sum(np.abs(y_true - predictions_corrected)) / num_non_zero. Change this to return np.sum(np.abs(y_true - predictions_corrected)) / num_non_zero and everything will be OK :) — Stef
– Stef, Commented Jun 14, 2020 at 12:00

Stef · Accepted Answer · 2020-06-14 12:08:03Z

2

The problem is in your mae function. In the last but one row you overwrite your function definition of mae with a number. If you call this function once everything is OK. As soon as you call it again (as in the loop), you try to call a number instead of a function, which is impossible.

Just change

mae = np.sum(np.abs(y_true - predictions_corrected)) / num_non_zero
# return
return(mae)

to

return np.sum(np.abs(y_true - predictions_corrected)) / num_non_zero

Python is not Basic or Fortran where you assign the result to the function name to return it :).

answered Jun 14, 2020 at 12:08

Stef

30.9k3 gold badges34 silver badges60 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Yustina Ivanova Over a year ago

I had a similar issue. I used the name of a function as a variable. In the end, renaming the variable solved the issue

Collectives™ on Stack Overflow

Python: 'numpy.float64' object is not callable

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related