Scholarlink

The primary artifact of the development and training phases of DRL agents are trained NNs. For policy-based DRL algorithms specifically, the efforts in previous chapters focus on features space improvements and more efficient training curricula result in sets of parameters that define approximate policy functions. However, little research exists on advanced methods for the inference phase of trained NNs. This chapter introduces two new methods to make use of trained state-of-the-art NNs for JSS and exeed their performance.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Effective Inference Strategies with Trained Agents for the JSSP

  • 章节

  • 发表日期：

    2026年

查看全文
  • Constantin Waubert de Puiseau

摘要

The primary artifact of the development and training phases of DRL agents are trained NNs. For policy-based DRL algorithms specifically, the efforts in previous chapters focus on features space improvements and more efficient training curricula result in sets of parameters that define approximate policy functions. However, little research exists on advanced methods for the inference phase of trained NNs. This chapter introduces two new methods to make use of trained state-of-the-art NNs for JSS and exeed their performance.