Brain Inspired Adaptive Memory Dual-Net for Few-Shot Image Classification

2025-03-10

AI Summary

Single-label supervision in few-shot image classification induces “supervisory collapse”—characterized by high intra-class variability and difficulty in localizing semantically discriminative regions. Method: Inspired by the human complementary learning system, we propose a hippocampal–neocortical dual-network architecture. Its core innovation is the first incorporation of neuroscientific systems consolidation into few-shot learning, realized via an adaptive memory module that dynamically consolidates structured category representations and optimizes generalization. Crucially, we eliminate reliance on explicit local feature alignment, instead employing a generalization-oriented long-term memory gating and retrieval mechanism to enhance robustness. Contribution/Results: Our approach achieves state-of-the-art performance on standard few-shot benchmarks, with particularly pronounced gains under high intra-class variation—outperforming existing methods significantly in such challenging scenarios.

Technology Category

• Machine Learning: Neuro-Symbolic Learning
• Computer Vision: Learning & Optimization for CV
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Social networks and social learning

Abstract

Few-shot image classification has become a popular research topic for its wide application in real-world scenarios, however the problem of supervision collapse induced by single image-level annotation remains a major challenge. Existing methods aim to tackle this problem by locating and aligning relevant local features. However, the high intra-class variability in real-world images poses significant challenges in locating semantically relevant local regions under few-shot settings. Drawing inspiration from the human's complementary learning system, which excels at rapidly capturing and integrating semantic features from limited examples, we propose the generalization-optimized Systems Consolidation Adaptive Memory Dual-Network, SCAM-Net. This approach simulates the systems consolidation of complementary learning system with an adaptive memory module, which successfully addresses the difficulty of identifying meaningful features in few-shot scenarios. Specifically, we construct a Hippocampus-Neocortex dual-network that consolidates structured representation of each category, the structured representation is then stored and adaptively regulated following the generalization optimization principle in a long-term memory inside Neocortex. Extensive experiments on benchmark datasets show that the proposed model has achieved state-of-the-art performance.

Problem

• Addresses supervision collapse in few-shot image classification.
• Overcomes high intra-class variability in real-world images.
• Simulates human complementary learning for feature integration.

Innovation

• Simulates human complementary learning system
• Uses adaptive memory module for feature identification
• Implements Hippocampus-Neocortex dual-network for structured representation