PonderNet-TensorFlow

This is an Unofficial Implementation of the paper: PonderNet: Learning to Ponder in TensorFlow.

Official PyTorch Implementation: Link

PonderNet adapts the computation based on the input. It changes the number of steps to take on a recurrent network based on the input. PonderNet learns this with end-to-end gradient descent.

Citations

@misc{banino2021pondernet,
      title={PonderNet: Learning to Ponder}, 
      author={Andrea Banino and Jan Balaguer and Charles Blundell},
      year={2021},
      eprint={2107.05407},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}

License

MIT