# Learning Recurrent Models with Temporally Local Rules

arXiv: https://arxiv.org/abs/2310.13284 · HTML: https://mlsanity.com/paper/2310.13284/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 5 of its 23 references that are in the map.
- First posted on arXiv 20 Oct 2023.

## Builds on (most cited first, 5 of 5)

- [Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks](https://mlsanity.com/paper/1511.06434/) (2015)
- [Learning Latent Dynamics for Planning from Pixels](https://mlsanity.com/paper/1811.04551/) (2018)
- [Monte Carlo Gradient Estimation in Machine Learning](https://mlsanity.com/paper/1906.10652/) (2019)
- [Clockwork Variational Autoencoders](https://mlsanity.com/paper/2102.09532/) (2021)
- [Recurrent Exponential-Family Harmoniums without Backprop-Through-Time](https://mlsanity.com/paper/1605.05799/) (2016)

## Abstract (arXiv)

Fitting generative models to sequential data typically involves two recursive computations through time, one forward and one backward. The latter could be a computation of the loss gradient (as in backpropagation through time), or an inference algorithm (as in the RTS/Kalman smoother). The backward pass in particular is computationally expensive (since it is inherently serial and cannot exploit GPUs), and difficult to map onto biological processes. Work-arounds have been proposed; here we explore a very different one: requiring the generative model to learn the joint distribution over current and previous states, rather than merely the transition probabilities. We show on toy datasets that different architectures employing this principle can learn aspects of the data typically requiring the backward pass.