Revision history [back]

Applying a filter on the number of samples in Keras will affect the time per step by reducing it since there are fewer samples to process. The reduction in time per step may increase the number of steps per epoch, but this depends on the other parameters set in the model such as batch size.

However, applying a filter on the number of samples may not have a significant impact on the epoch time because the total number of samples in the dataset remains the same. It only affects the number of samples that are fed into the model for each step. Hence, the overall training time may not change significantly.

In summary, applying a filter on the number of samples may reduce the time per step, but it may not have a significant impact on the epoch time.