NN Architecture

[Joel Ranzer]

Hi,

I am very confused about the "context" argument of the network(Table 2). It is named differently in the tensorflow documentation. https://www.tensorflow.org/api_docs/python/tf/keras/layers/Conv1D

In the TensorFlow dokumentation they call:

Num Filters -> filters
Filter Size -> kernal_size

Dilation -> dilation_rate

Context?

Any ideas what context is?

Greetings

Joel Ranzer

[Oran Lang]

Hi Joel,

The context is not a part of the filter specification. It indicates the size of the receptive field of each neuron (but in time / frequency dimension). It is derived from the sequence of filters and dilation size.