Efficient Gaussian Process Model on Class-Imbalanced Datasets for Generalized Zero-Shot Learning

TL;DR

This paper introduces an efficient Gaussian Process-based model for generalized zero-shot learning that effectively handles class imbalance, achieving state-of-the-art results on several benchmark datasets with minimal training time.

Contribution

It proposes a novel neural network and Gaussian Process framework for ZSL that addresses class imbalance and is trained efficiently, setting new performance standards.

Findings

Achieves SOTA performance on AWA2, AWA1, and APY datasets.

Trains in approximately 5 minutes on average.

Performs well on SUN and CUB datasets despite imbalance.

Abstract

Zero-Shot Learning (ZSL) models aim to classify object classes that are not seen during the training process. However, the problem of class imbalance is rarely discussed, despite its presence in several ZSL datasets. In this paper, we propose a Neural Network model that learns a latent feature embedding and a Gaussian Process (GP) regression model that predicts latent feature prototypes of unseen classes. A calibrated classifier is then constructed for ZSL and Generalized ZSL tasks. Our Neural Network model is trained efficiently with a simple training strategy that mitigates the impact of class-imbalanced training data. The model has an average training time of 5 minutes and can achieve state-of-the-art (SOTA) performance on imbalanced ZSL benchmark datasets like AWA2, AWA1 and APY, while having relatively good performance on the SUN and CUB datasets.