Rethinking Task-Incremental Learning Baselines

TL;DR

This paper introduces a simple adjustment network (SAN) for task-incremental learning that maintains high performance with minimal architecture size and no memory of past instances, applicable across 2D and 3D recognition tasks.

Contribution

The study proposes SAN, a novel, memory-efficient approach for task-incremental learning that outperforms existing methods without increasing model size or relying on past data.

Findings

SAN achieves near state-of-the-art performance.

SAN is unaffected by task order variations.

Effective across both 2D and 3D recognition tasks.

Abstract

It is common to have continuous streams of new data that need to be introduced in the system in real-world applications. The model needs to learn newly added capabilities (future tasks) while retaining the old knowledge (past tasks). Incremental learning has recently become increasingly appealing for this problem. Task-incremental learning is a kind of incremental learning where task identity of newly included task (a set of classes) remains known during inference. A common goal of task-incremental methods is to design a network that can operate on minimal size, maintaining decent performance. To manage the stability-plasticity dilemma, different methods utilize replay memory of past tasks, specialized hardware, regularization monitoring etc. However, these methods are still less memory efficient in terms of architecture growth or input data costs. In this study, we present a simple yet effective adjustment network (SAN) for task incremental learning that achieves near state-of-the-art performance while using minimal architectural size without using memory instances compared to previous state-of-the-art approaches. We investigate this approach on both 3D point cloud object (ModelNet40) and 2D image (CIFAR10, CIFAR100, MiniImageNet, MNIST, PermutedMNIST, notMNIST, SVHN, and FashionMNIST) recognition tasks and establish a strong baseline result for a fair comparison with existing methods. On both 2D and 3D domains, we also observe that SAN is primarily unaffected by different task orders in a task-incremental setting.