Search papers, labs, and topics across Lattice.
This paper introduces a novel meta-learning algorithm that enables agents to optimize their own weights and biases through a dual-loop structure, combining high-dimensional inner optimization with low-dimensional outer optimization. The approach addresses the limitations of traditional gradient descent and gradient-free methods, which struggle with stability and scalability over long time horizons. Key results demonstrate that this greedy dynamical meta-learning framework significantly enhances the learning process in large models, allowing for more effective adaptation over extended periods.
Agents that learn to optimize their own weights can achieve unprecedented stability and efficiency over long time horizons.
Gradient descent scales well to large models, but becomes unstable over long time horizons. Gradient-free optimizers can scale to arbitrary timespans, but are hobbled by high dimensions. Since learning occurs in large models over long timescales, neither of these approaches is likely to produce traits which can accelerate the learning process. Instead, we propose a meta-learning algorithm in which the agent learns to modify its own weights and biases. Our algorithm consists of an inner loop, wherein the agent performs some high-dimensional optimization upon itself, and an outer loop, wherein we perform some low-dimensional optimization upon the inner loop. Since the outer loop handles very few parameters, standard zeroth-order methods may be used.
