Beginning Anomaly Detection Using

Figure A-8.  Code to evaluate the model given x and y data sets

For model.evaluate(), the parameters are

• 

x: The Numpy array representing the test data. Pass in a list of Numpy 

arrays if the model has multiple inputs.

• 

y: The Numpy array of target or label data that is a part of the test 

data. If there are multiple inputs, pass in a list of Numpy arrays.

Appendix A   intro to KerAs

---

327

• 

batch_size: If none is specified, the default is 32. This parameter 

expects an integer value that dictates how many samples there are 

per evaluation step.

• 

verbose: If set to 0, no output is shown. If set to 1, the progress bar is 

shown and looks like Figure 

A-9

.

Figure A-9.  The evaluate function with verbosity 1

Figure A-10.  The prediction function generates predictions given some data set x

• 

sample_weight: (optional) A Numpy array of weights for each of 

the test samples. Again, either a 1:1 map between the sample and 

the weights, unless it’s temporal data. If you have temporal data (an 

extra time dimension), pass in a 2D array with a shape (samples, 

sequence_length) to apply these weights to each timestep of the 

samples. Don’t forget to set “temporal” for sample_weight_mode in 

model.compile().

•