# Variational Prototype Replays for Continual Learning

arXiv: https://arxiv.org/abs/1905.09447 · HTML: https://mlsanity.com/paper/1905.09447/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 23 May 2019.

## Built on by (most cited first, 7 of 7)

- [Continual Learning via Bit-Level Information Preserving](https://mlsanity.com/paper/2105.04444/) (2021)
- [Prototype Guided Federated Learning of Visual Feature Representations](https://mlsanity.com/paper/2105.08982/) (2021)
- [ProtoDepth: Unsupervised Continual Depth Completion with Prototypes](https://mlsanity.com/paper/2503.12745/) (2025)
- [Improving generalization by mimicking the human visual diet](https://mlsanity.com/paper/2206.07802/) (2022)
- [Shared and Private VAEs with Generative Replay for Continual Learning](https://mlsanity.com/paper/2105.07627/) (2021)
- [Retrospective Feature Estimation for Continual Learning](https://mlsanity.com/paper/2406.17381/) (2024)
- [BPG: Balancing Plasticity and Generalization for Domain Incremental Learning](https://mlsanity.com/paper/2608.10804/) (2026)

## Abstract (arXiv)

Continual learning refers to the ability to acquire and transfer knowledge without catastrophically forgetting what was previously learned. In this work, we consider few-shot continual learning in classification tasks, and we propose a novel method, Variational Prototype Replays, that efficiently consolidates and recalls previous knowledge to avoid catastrophic forgetting. In each classification task, our method learns a set of variational prototypes with their means and variances, where embedding of the samples from the same class can be represented in a prototypical distribution and class-representative prototypes are separated apart. To alleviate catastrophic forgetting, our method replays one sample per class from previous tasks, and correspondingly matches newly predicted embeddings to their nearest class-representative prototypes stored from previous tasks. Compared with recent continual learning approaches, our method can readily adapt to new tasks with more classes without requiring the addition of new units. Furthermore, our method is more memory efficient since only class-representative prototypes with their means and variances, as well as only one sample per class from previous tasks need to be stored. Without tampering with the performance on initial tasks, our method learns novel concepts given a few training examples of each class in new tasks.