Equalization Optimization for SerDes Channels with Constrained Bayesian Optimization

Assigning parameters of a feed-forward equalizer (FFE) can be a challenging and time-consuming task. In this work we introduce a machine learning algorithm to automatically optimize these parameters without the need to a domain expert. Conventional optimizers are not applicable to this problem because of a constraint over the FFE parameters. Therefore, we reformulate the problem and propose a modified Bayesian optimization algorithm to take this constraint into account. The proposed approach is validated with an example.