Improved Feature Generating Framework for Transductive Zero-shot Learning

2024-12-24

AI Summary

In zero-shot learning (ZSL), prior bias toward unseen classes impairs discriminative accuracy of generated features—especially when generators are imperfect—while unconditional discriminators introduce noise. To address these challenges in transductive ZSL (TZSL), we propose I-VAEGAN, a novel framework integrating variational autoencoders (VAEs) and generative adversarial networks (GANs). Our approach introduces two key innovations: pseudo-conditioned feature adversarial (PFA) learning and variational embedding regression (VER), which jointly eliminate reliance on unseen-class priors and enable semantics-driven, precise conditional generation alongside class-wise statistical modeling. I-VAEGAN unifies pseudo-conditioned discrimination, semantic regression, reconstruction-based pretraining, and conditional adversarial training. Evaluated on multiple TZSL benchmarks, it achieves state-of-the-art performance, significantly improving unseen-class classification accuracy across diverse prior settings, while demonstrating superior generalization and robustness.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Feature Generative Adversarial Networks have emerged as powerful generative models in producing high-quality representations of unseen classes within the scope of Zero-shot Learning (ZSL). This paper delves into the pivotal influence of unseen class priors within the framework of transductive ZSL (TZSL) and illuminates the finding that even a marginal prior bias can result in substantial accuracy declines. Our extensive analysis uncovers that this inefficacy fundamentally stems from the utilization of an unconditional unseen discriminator - a core component in existing TZSL. We further establish that the detrimental effects of this component are inevitable unless the generator perfectly fits class-specific distributions. Building on these insights, we introduce our Improved Feature Generation Framework, termed I-VAEGAN, which incorporates two novel components: Pseudo-conditional Feature Adversarial (PFA) learning and Variational Embedding Regression (VER). PFA circumvents the need for prior estimation by explicitly injecting the predicted semantics as pseudo conditions for unseen classes premised by precise semantic regression. Meanwhile, VER utilizes reconstructive pre-training to learn class statistics, obtaining better semantic regression. Our I-VAEGAN achieves state-of-the-art TZSL accuracy across various benchmarks and priors. Our code would be released upon acceptance.

Problem

• Zero-shot Learning
• Unseen Class Bias
• Conditional Discriminator

Innovation

• I-VAEGAN
• Pseudo Conditional Feature Adversarial Learning
• Variational Embedding Regression