lab-12-5-rnn_stock_prediction.py about deeplearningzerotoall HOT 10 CLOSED

It can be done in two lines though.

def MinMaxScaler(data):
    numerator = data - np.min(data, 0)
    denominator = np.max(data, 0) -  np.min(data, 0)
    return numerator/ (denominator + 1e-8) # noise should be added to prevent zero division

from deeplearningzerotoall.

Thanks. Do you know which numpy calls in the file cause this issue? Perhaps, we can replace it.

from deeplearningzerotoall.

This problem occurred on the line from sklearn.preprocessing import MinMaxScaler.
If we can edit scaler = MinMaxScaler(feature_range=(0, 1)) and xy = scaler.fit_transform(xy), we will solve the problem another way.

from deeplearningzerotoall.

def MinMaxScaler(data):
    num_row = np.shape(data)[0]
    num_col = np.shape(data)[1]
    array = np.zeros((num_row, num_col))
    for i in range(num_col):
        input = data[:,i]
        array[:,i] = (input - np.min(input)) / (np.max(input) - np.min(input))
    return array

If we use above function, we can do it!! Also, we don't need to use sklearn.preprocessing package.
(lab-12-5 and lab-12-6)

from deeplearningzerotoall.

from deeplearningzerotoall.

#56 #57 Done.

from deeplearningzerotoall.

from deeplearningzerotoall.

from deeplearningzerotoall.

I can do that. The vectorized version is 8 times faster than the loop

In [9]: %timeit -n 1000 MinMaxScaler1(A)
1000 loops, best of 3: 81.8 µs per loop

In [10]: %timeit -n 1000 MinMaxScaler2(A)
1000 loops, best of 3: 11.3 µs per loop

from deeplearningzerotoall.

from deeplearningzerotoall.