Darwinian Model Upgrades: Model Evolving with Selective Compatibility

TL;DR

This paper introduces Darwinian Model Upgrades (DMU), a novel approach for model evolution that enhances compatibility and discriminativeness without costly backfilling, demonstrated on large-scale retrieval benchmarks.

Contribution

DMU disentangles inheritance and variation in model upgrades, enabling selective backward compatibility and forward adaptation for efficient large-scale retrieval systems.

Findings

DMU improves new-to-old compatibility.

DMU reduces degradation in new models.

DMU outperforms existing methods on benchmarks.

Abstract

The traditional model upgrading paradigm for retrieval requires recomputing all gallery embeddings before deploying the new model (dubbed as "backfilling"), which is quite expensive and time-consuming considering billions of instances in industrial applications. BCT presents the first step towards backward-compatible model upgrades to get rid of backfilling. It is workable but leaves the new model in a dilemma between new feature discriminativeness and new-to-old compatibility due to the undifferentiated compatibility constraints. In this work, we propose Darwinian Model Upgrades (DMU), which disentangle the inheritance and variation in the model evolving with selective backward compatibility and forward adaptation, respectively. The old-to-new heritable knowledge is measured by old feature discriminativeness, and the gallery features, especially those of poor quality, are evolved in a lightweight manner to become more adaptive in the new latent space. We demonstrate the superiority of DMU through comprehensive experiments on large-scale landmark retrieval and face recognition benchmarks. DMU effectively alleviates the new-to-new degradation and improves new-to-old compatibility, rendering a more proper model upgrading paradigm in large-scale retrieval systems.