Beginning Anomaly Detection Using

n times according to its time entry.

 UpSampling  2D

keras.layers.UpSampling2D()

Similar to UpSampling1D(), but for 2D inputs. The rows and columns are repeated 

n 

times according to size[0] and size[1].

This is the list of parameters:

• 

size: An integer or tuple of two integers. The integer is the 

upsampling factor for both rows and columns, and the tuple lets you 

specify the upsampling factor for rows and for columns individually.

Appendix A   intro to KerAs

---

338

• 

data_format: ‘channels_first’ or ‘channels_last’. This tells the 

flattening layer how to format the flattened output to preserve the 

formatting of channels first or channels last.

• 

interpolation: ‘nearest’ or ‘bilinear’. CNTK does not support 

‘bilinear’ yet, and Theanos only supports size=(2,2). ‘nearest’ and 

‘bilinear’ are interpolation techniques used in image processing.

 ZeroPadding1D

keras.layers.ZeroPadding1D()

Depending on the input, pads the input sequence with zeroes on both sides or either 

a zero on the left side or a zero on the right side of the input sequence.

This is the list of parameters:

• 

padding: An integer, a tuple of two integers, or a dictionary. The 

integer is a number that tells the layer how many zeroes to add on 

both the left and right side. An input of 

1 adds a zero on both the left 

and right side. The tuple is formatted as (left_pad, right_pad), so 

you can pass in (0, 1) to tell it to add no zeroes on the left side and 

add one zero on the right side.

 ZeroPadding2D

keras.layers.ZeroPadding2D()

Depending on the input, it pads the input sequence with a row and columns of 

zeroes at the top, left, right, and bottom of the image tensor.

This is the list of parameters:

•