Multipath agents for modular multitask ML systems
This work addresses the challenge of building modular multitask ML systems, offering a novel approach for reusing and combining modules across tasks, though it appears incremental in its specific routing method.
The paper tackles the problem of generating and improving ML models for multiple tasks by introducing a methodology where multiple agents collaborate and compete, demonstrated on over 100 image classification tasks. It shows that a per-sample parallel routing method can boost solution quality by training only a fraction of activated parameters.
A standard ML model is commonly generated by a single method that specifies aspects such as architecture, initialization, training data and hyperparameters configuration. The presented work introduces a novel methodology allowing to define multiple methods as distinct agents. Agents can collaborate and compete to generate and improve ML models for a given tasks. The proposed methodology is demonstrated with the generation and extension of a dynamic modular multitask ML system solving more than one hundred image classification tasks. Diverse agents can compete to produce the best performing model for a task by reusing the modules introduced to the system by competing agents. The presented work focuses on the study of agents capable of: 1) reusing the modules generated by concurrent agents, 2) activating in parallel multiple modules in a frozen state by connecting them with trainable modules, 3) condition the activation mixture on each data sample by using a trainable router module. We demonstrate that this simple per-sample parallel routing method can boost the quality of the combined solutions by training a fraction of the activated parameters.