Multi-agent ai improves tuning of specialized neural network components
A Multi-Agent LLM System for Automated Tuning of Specialized Neural Network Models
Artificial IntelligenceComputation and LanguageMachine LearningMultiagent SystemsNeural and Evolutionary Computing
Summary
Tuning key parts of machine learning models usually requires expert knowledge and manual work. The authors propose a system where multiple AI agents collaborate, each with a specific role, to automatically improve these parts without changing the entire model. They tested this system on tasks like classifying images and solving physics problems, and it performed better than standard tuning methods. This approach works across different types of models by focusing only on important components rather than redesigning whole networks.
What this means in practice
- •For machine learning engineers: Automatically tune critical components of diverse neural models to improve classification and scientific computing tasks without redesigning entire architectures.
- •For scientific computing teams: Improve solution accuracy and preserve invariants in physics-based neural simulations through automated component tuning powered by multi-agent AI.
Authors
Ivan Moloshnikov, Dmitry Kunitsyn, Alexander Sboev, Maksim Skorokhodov, Roman Rybka, Yuanchao Liu
Abstract
Many machine learning approaches are two-level: an expert selects a class of methods, within this class the choice of key components is critical and usually manual. This paper addresses the automatic tuning of these components and proposes a framework transferable across models of different nature. The multi-agent large language model (LLM) framework has three roles---analyst, coder, and recorder~---with an experiment log and a sandboxed execution environment. Unlike systems that optimize the entire ML script or search architectures in a discrete space, the framework operates over an open but semantically narrow space: the agent generates code only for the specified components, leaving the overall network structure unchanged. The approach is validated in two directions: tuning a classification spiking neural network with correlation learning, and solving high-order differential equations with Physics-Informed Neural Networks (PINN). Across five tasks---spiking classification on MNIST and Digits, and boundary-value problems for the generalized nonlinear Schrödinger equation (GNLSE) and Kuramoto--Sivashinsky (KS) equations---the agent outperforms the baselines: classification accuracy is 13 percentage points (MNIST) and 6 percentage points (Digits) higher; the error against the analytical single-soliton solution is reduced approximately 2491-fold; the invariant-preservation error drops approximately 325-fold in the selected GNLSE run with a Gaussian perturbation and by more than 100-fold on average over repeated initializations; and the mass drift drops 14-fold for the KS problem with a Gaussian perturbation. The framework thus tunes key components on tasks of different nature without changing its internal structure.