Model Selection & Hyper-Parameter Tuning

August 26, 2020 | Patrick Bangert

In training AI models, we have to choose what kind of #model to train, e.g. how many layers and how many neurons per layer. This process is called ModelSelection. The Machine Learning #algorithm that trains the model has parameters such as LearningRate or #momentum that need to be chosen, as well. While we can set parameters for both manually by selecting a few options and choosing the best one, this process can be automated! In automating it, we need to define precisely what we mean by “best” in order to navigate the bias-variance-trade-off. Also, we need the training run to be fast. Otherwise, we could never run the many trial-and-error experiments needed. Reducing training time makes #AutoML feature generation, model selection, hyper-parameter tuning possible. Brightics AI Accelerator does all of this for you - yielding a much better model.


  • RSS
  • Samsung SDS
  • Samsung SDS LinkedIn
  • Samsung SDS Youtube
  • Samsung SDS Slideshare

Copyright ©2019 SAMSUNG SDS America, Inc. All rights reserved.