GPy/backlog/infrastructure/2025-08-15_parameter-tying-framework.md

2.8 KiB

id title status priority created last_updated owner github_issue dependencies tags
parameter-tying-framework Design parameter tying framework for GPy multioutput kernels Ready High 2025-08-15 2025-08-15 Neil Lawrence
parameter-tying
multioutput
kernel-framework
architecture

Investigate parameter tying limitations and create CIP for discussion

Description

During LFM kernel code review, we identified that GPy lacks systematic parameter tying capabilities compared to GPmat's modelTieParam() functionality. This limitation affects combination kernels such as multiouptut or additive kernels. We need to investigate the scope of this problem and create a CIP to discuss potential solutions with the community.

Problem Statement

  • Current Limitation: GPy's parameter system doesn't support tying parameters across different kernel components
  • Impact on LFM: Forces complex parameter handling in EQ_ODE1 and EQ_ODE2 kernels
  • Broader Impact: May affect other multiple kernel scenarios where parameters should be shared
  • Comparison: GPmat has modelTieParam() functionality that GPy lacks

Investigation Needed

1. Scope Assessment

  • Search existing GitHub issues for parameter tying discussions
  • Identify other kernels/models that could benefit from parameter tying
  • Assess impact on current GPy codebase

2. Community Input

  • Create GitHub issue and associated CIP to discuss parameter tying needs
  • Gather feedback from GPy maintainers and users
  • Identify use cases beyond LFM kernels
  • Assess priority relative to other GPy improvements

3. Technical Analysis

  • Analyze GPmat's parameter tying implementation
  • Review GPy's current parameter system architecture
  • Identify potential integration points
  • Assess complexity and maintenance burden

Acceptance Criteria

  • Complete investigation of existing GitHub issues and discussions
  • Document scope of parameter tying needs across GPy
  • Create CIP for parameter tying framework discussion
  • Gather community feedback on approach and priority
  • Provide recommendations for next steps

Implementation Notes

  • Focus on problem identification and community discussion
  • Avoid prescribing specific solutions until community input is gathered
  • Consider whether this should be a separate CIP or part of broader multioutput improvements
  • Document trade-offs between different approaches
  • CIP: 0001 (LFM kernel implementation) - may depend on parameter tying
  • GPy parameter system design
  • GPmat parameter tying implementation
  • Multioutput kernel architecture discussions

Progress Updates

2025-08-15

Task created after identifying parameter tying as a potential limitation during LFM kernel code review. Need to investigate scope and create CIP for community discussion.