What Do You See in Common? Learning Hierarchical Prototypes over Tree-of-Life to Discover Evolutionary Traits

2024-09-03

arXiv.org

AI Summary

This work addresses the challenge of automated evolutionary trait discovery in biological images. We propose HComP-Net, the first framework that aligns prototype learning with the hierarchical structure of the phylogenetic tree. Methodologically, it integrates a hierarchical prototype network, phylogeny-aware regularization, multi-granularity commonality modeling, and cross-species zero-shot transfer—jointly mitigating critical issues including internal-node overfitting, prototype semantic inconsistency, and poor generalization. Evaluated on avian, lepidopteran, and fish image datasets, HComP-Net significantly improves prototype accuracy, semantic coherence, and zero-shot generalization to unseen species. The code and datasets are publicly released. Our approach establishes a novel, interpretable, phylogenetically aligned, and scalable paradigm for image-based evolutionary phenotypic analysis.

Technology Category

• Machine Learning: Evolutionary Learning
• Search and Optimization: Evolutionary Computation
• Computer Vision: Representation Learning for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Models for Web evolution
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

A grand challenge in biology is to discover evolutionary traits - features of organisms common to a group of species with a shared ancestor in the tree of life (also referred to as phylogenetic tree). With the growing availability of image repositories in biology, there is a tremendous opportunity to discover evolutionary traits directly from images in the form of a hierarchy of prototypes. However, current prototype-based methods are mostly designed to operate over a flat structure of classes and face several challenges in discovering hierarchical prototypes, including the issue of learning over-specific features at internal nodes. To overcome these challenges, we introduce the framework of Hierarchy aligned Commonality through Prototypical Networks (HComP-Net). We empirically show that HComP-Net learns prototypes that are accurate, semantically consistent, and generalizable to unseen species in comparison to baselines on birds, butterflies, and fishes datasets. The code and datasets are available at https://github.com/Imageomics/HComPNet.

Problem

• Discover evolutionary traits from biological images using hierarchical prototypes
• Overcome challenges in learning hierarchical prototypes with over-specificity issues
• Introduce HComP-Net framework to improve prototype accuracy and generalizability

Innovation

• Hierarchical prototype learning over tree-of-life
• Novel over-specificity and discriminative losses
• Masking module for prototype exclusion