Shuaizheng Lu | Searcharxiv

arXiv2024

Data-Driven Decentralized Control Design for Discrete-Time Large-Scale Systems

In this paper, a data-driven approach is developed for controller design for a class of discrete-time large-scale systems, where a large-scale system can be expressed in an equivalent data-driven form and the decentralized controllers can be parameterized by the data collected from its subsystems, i.e., system state, control input, and interconnection input. Based on the developed data-driven method and the Lyapunov approach, a data-driven semi-definite programming problem is constructed to obtain decentralized stabilizing controllers. The proposed approach has been validated on a mass-spring chain model, with the significant advantage of avoiding extensive modeling processes.

eess.SY↗

arXiv2024

Compression Repair for Feedforward Neural Networks Based on Model Equivalence Evaluation

In this paper, we propose a method of repairing compressed Feedforward Neural Networks (FNNs) based on equivalence evaluation of two neural networks. In the repairing framework, a novel neural network equivalence evaluation method is developed to compute the output discrepancy between two neural networks. The output discrepancy can quantitatively characterize the output difference produced by compression procedures. Based on the computed output discrepancy, the repairing method first initializes a new training set for the compressed networks to narrow down the discrepancy between the two neural networks and improve the performance of the compressed network. Then, we repair the compressed FNN by re-training based on the training set. We apply our developed method to the MNIST dataset to demonstrate the effectiveness and advantages of our proposed repair method.

cs.LG↗