Activity regularization not supported

[johannpayer]

System information

• TensorFlow.js version (you are using): 4.12.0
• Are you willing to contribute it (Yes/No): No

Describe the feature and the current behavior/state.
Added support for activity regularization. Currently, layers contain an activityRegularizer property, but on invocation throw NotImplementedError: Layer invocation in the presence of activity regularizer(s) is not supported yet.

Will this change the current api? How?
No, because the activityRegularizer property already exists in the API. Only an implementation needs to be added.

Who will benefit with this feature?
It could help anyone trying to regularize their neural networks as it provides another type of regularization they can use.

Any Other info.
Nothing I can think of.

[gaikwadrahul8]

Hi, @johannpayer

Thank you for bringing this issue to our attention and if possible could you please help us with code-snippet/codepen-example to replicate the same behaviour from our end ? Thank you!

[johannpayer]

Sure. Here's a CodePen example.

import {
  layers,
  regularizers,
  ones
} from "https://cdn.skypack.dev/@tensorflow/tfjs@4.12.0";

const dense = layers.dense({
  units: 1,
  activityRegularizer: regularizers.l2()
});

try {
  dense.apply(ones([1, 1]));
  console.log("Layer invocation succeeded.");
} catch (error) {
  console.error(error.message);
}

"Layer invocation in the presence of activity regularizer(s) is not supported yet."

[johannpayer]

Sorry, I accidentally clicked close.

[gaikwadrahul8]

Hi, @johannpayer

No problem, thank you for providing the codepen-example, I was trying to replicate the same issue from my end with your codepen-example and I'm also getting the same error message so at the moment looking at the error message Layer invocation in the presence of activity regularizer(s) is not supported yet, Activity regularization not supported yet so either we'll have to dig more into this issue or most probably this issue will be considered as feature request.

Thank you for bringing this issue to our attention, I really appreciate your patience and valuable time. Thank you!