Annealed Multiple Choice Learning: Overcoming limitations of Winner-takes-all with annealing
This work addresses a limitation in learning frameworks for ambiguous tasks, offering an incremental improvement for applications like speech separation.
The paper tackles the problem of Winner-takes-all (WTA) in Multiple Choice Learning (MCL) converging to suboptimal local minima by introducing Annealed Multiple Choice Learning (aMCL), which uses simulated annealing to enhance exploration and improve performance, as validated on synthetic datasets, UCI benchmark, and speech separation tasks.
We introduce Annealed Multiple Choice Learning (aMCL) which combines simulated annealing with MCL. MCL is a learning framework handling ambiguous tasks by predicting a small set of plausible hypotheses. These hypotheses are trained using the Winner-takes-all (WTA) scheme, which promotes the diversity of the predictions. However, this scheme may converge toward an arbitrarily suboptimal local minimum, due to the greedy nature of WTA. We overcome this limitation using annealing, which enhances the exploration of the hypothesis space during training. We leverage insights from statistical physics and information theory to provide a detailed description of the model training trajectory. Additionally, we validate our algorithm by extensive experiments on synthetic datasets, on the standard UCI benchmark, and on speech separation.