Using biological networks to improve neural architecture search performance
TRON: Transferring Naturally Evolved Network Architectures into Neuroevolution
Neural and Evolutionary ComputingMachine Learning
Summary
Finding the best design for artificial neural networks is hard because there are many ways to connect them. The authors introduced TRON, a method that starts with structures inspired by real biological protein networks instead of random setups. This approach helps the system evolve better neural networks more reliably across different problems. TRON performed better than other methods on several tests, showing that borrowing ideas from biology can help make artificial intelligence designs work well.
What this means in practice
- •For machine learning engineers: Develop neural networks with improved prediction accuracy by initializing architectures based on biologically inspired network patterns.
- •For bioinformatics software teams: Create AI tools that leverage known biological network structures to enhance model robustness and consistency across genomic datasets.
Authors
Konstantinos Zacharopoulos, Alexandra Bekou, Jose M Serradell, Foteini Paraskevopoulou, Apostolos Fournaris, Ioannis Arapakis, Oscar Lao
Abstract
Neural architecture search (NAS) is challenging because of the vast, non-linear space of artificial neural network (ANN) topologies. Evolutionary methods can generate architectures, but their effectiveness is limited by difficulties in preserving functional organization during mutation and recombination. Naturally evolved biological networks may provide organizational principles that support adaptation and evolvability. Here, we introduce TRON (TRansfer Of Natural networks), a neuroevolution framework that uses experimentally characterized protein-protein interaction (PPI) networks as architectural priors. Instead of starting from random topologies, TRON reuses biological interaction patterns as structural building blocks. It combines classical evolutionary operators with neuroevolution-inspired mechanisms, including speciation, and introduces species quarantine to protect newly generated architectures during early maturation. We evaluated TRON against established evolutionary and neuroevolutionary methods on three classification benchmarks (CORA, CITESEER, and PUBMED) and three regression benchmarks (DIABETES, FRIEDMAN, and CALIFORNIA). TRON achieved the highest predictive performance among the evaluated methods on all classification datasets and most regression tasks. It showed substantially lower variability across independent runs, indicating greater robustness and convergence stability, at increased computational cost. These findings support the use of PPI networks as effective architectural priors and suggest that biologically derived interaction networks provide more favorable substrates for neuroevolution than random topologies. More broadly, TRON demonstrates the feasibility of transferring naturally evolved network architectures into artificial intelligence systems, offering a direction for improving evolvability in NAS through the reuse of biological organizational principles.