BayesPCN: A Continually Learnable Predictive Coding Associative Memory
This addresses the challenge of continual learning in associative memory for AI systems, offering a novel approach to memory writes and forgetting, though it appears incremental in its specific domain.
The paper tackles the problem of continual one-shot memory writes in associative memory without meta-learning, and shows that BayesPCN can recall corrupted high-dimensional data from hundreds to a thousand timesteps ago with minimal drop in ability compared to state-of-the-art offline models.
Associative memory plays an important role in human intelligence and its mechanisms have been linked to attention in machine learning. While the machine learning community's interest in associative memories has recently been rekindled, most work has focused on memory recall ($read$) over memory learning ($write$). In this paper, we present BayesPCN, a hierarchical associative memory capable of performing continual one-shot memory writes without meta-learning. Moreover, BayesPCN is able to gradually forget past observations ($forget$) to free its memory. Experiments show that BayesPCN can recall corrupted i.i.d. high-dimensional data observed hundreds to a thousand ``timesteps'' ago without a large drop in recall ability compared to the state-of-the-art offline-learned parametric memory models.