Algorithmic Components

PrismBO is equipped with a collection of state-of-the-art TLBO algorithms, all implemented following object-oriented design principles. The following tutorial pages provide an overview of the different algorithmic components and illustrate how they are initialized and executed within the PrismBO framework.

Overview

• Transfer Learning for Search Space Design: How to use data to reshape the search space, improving search efficiency.

• Transfer Learning for Acquisition Function: How to refine the acquisition function using data to better balance exploration and exploitation.

• Transfer Learning for Initialization Design: How to utilize data to warm-start the optimization process with informative initial configurations.

• Transfer Learning for Surrogate Model: How to enhance the surrogate model by incorporating knowledge from previously learned models or task-specific data.

The List of Components

• Transfer Learning for Acquisition Function
  • TAF
  • FSAF
  • MES
  • EI
  • PI
  • UCB
  • RAF
• Transfer Learning for Initialization Design
  • aLI
  • MI-SMBO
• Transfer Learning for Surrogate Model
  • MTGP
  • MHGP
  • DeepKernel
  • HyperGP
  • RGPE
  • SGPT
  • Transformer
• Transfer Learning for Search Space Design
  • Box
  • Eclipse
  • Prune