Emulation Learning for Neuromimetic Systems
This work addresses computational bottlenecks in neuromimetic emulation for researchers in neural systems and control, but it is incremental as it builds on existing neural heuristic quantization systems.
The paper tackles the combinatorial complexity of solving optimal quantization problems in neuromimetic systems via model predictive control by proposing a Deep Q Network algorithm that learns trajectories and demonstrates resilience to channel dropouts, with a mapping-based transfer learning approach for adapting to other emulation problems.
Building on our recent research on neural heuristic quantization systems, results on learning quantized motions and resilience to channel dropouts are reported. We propose a general emulation problem consistent with the neuromimetic paradigm. This optimal quantization problem can be solved by model predictive control (MPC), but because the optimization step involves integer programming, the approach suffers from combinatorial complexity when the number of input channels becomes large. Even if we collect data points to train a neural network simultaneously, collection of training data and the training itself are still time-consuming. Therefore, we propose a general Deep Q Network (DQN) algorithm that can not only learn the trajectory but also exhibit the advantages of resilience to channel dropout. Furthermore, to transfer the model to other emulation problems, a mapping-based transfer learning approach can be used directly on the current model to obtain the optimal direction for the new emulation problems.