We generalize the ‘bits back with ANS’ method to time-series models with a latent Markov structure. This family of models includes hidden Markov models (HMMs), linear Gaussian state space models (LGSSMs) and many more. We provide experimental evidence that our method is effective for small scale models, and discuss its applicability to larger scale settings such as video compression.
|Title of host publication||Neural Compression Workshop at ICLR 2021|
|Number of pages||5|
|Publication status||Published - 7 May 2021|
|Event||Neural Compression Workshop @ ICLR 2021 - Online|
Duration: 7 May 2021 → 7 May 2021
|Workshop||Neural Compression Workshop @ ICLR 2021|
|Period||7/05/21 → 7/05/21|