MeasureTransport / MeasureTransport/MParT

Overhaul to `MapObjective`

Open
#317 1 comment 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
C++
Stars
18
Forks
5
PR merge metrics
No merged PRs in 30d

Description

Currently MapObjective is a nice idea, but has some flaws

  • KLObjective is not really measuring the KL, it's off by a constant. Maybe we just disregard this and call it a quirk.
  • In practice, we need regularization. It would be really cool to have another abstract class like RegularizedObjective::MapObjective, where we create an abstract class RegularizerBase with e.g. child class L2Regularizer(lambda). Then, we could do something like RegularizedObjective obj = KLObjective(train, test, dim) + L2Regularizer(0.05).

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reading the existing MapObjective and KLObjective abstractions, then review the proposed RegularizedObjective::MapObjective and RegularizerBase design described in the issue. Determine how the KL constant and regularization should be specified, including the suggested L2Regularizer(lambda) example. Done means the objective design is agreed and the requested regularization behavior is implemented consistently.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.