Post

Work Log: 2026-05

/// Under construction ///

Posted
By EnderJones
1 min read
Work Log: 2026-05

Info:

  • This is an append-only work log.
  • Main purpose: help future me resume work quickly.
  • Public readability is useful, but this is not a polished monthly report.
  • You can say this is a AGENT.md/CLAUDE.md/GEMINI.md for my future self.

Rules:

  • Keep entries date-stamped and in chronological order.
  • Record changes, decisions, risks, and the next actions.
  • Keep private life, job-process details, and sensitive experiment details out.
  • If one section grows too large, split it into a standalone post later.

Resume Here

Last updated: 2026-05-05 18:26 JST.

Current State

  1. Research:
    • P1-8 results are in and should be reviewed for split metadata and protocol differences before trusting any scores.
    • Evaluation protocol is now refreshed as:
    • 3 seeds × 7-fold subject-level CV, stratified by ctrl/test.
    • For each seed/fold, Stage 2 uses 48 subjects for training and 8 unseen subjects for validation.
    • Stage 1 CCT pre-training is also fold-specific: each fold has separate landmark/rPPG CCT weights.
    • Stage 1 chooses its best CCT weights using only a small validation split inside the 48 Stage-2 training subjects.
    • The 8 Stage-2 validation subjects are never used in Stage 1.
  2. Job preparation:
    • Review 2 past exam question sets.
    • LLM / VLM / Agentic projects/research method steering Planning.

Do Not Forget

  • Do not trust any f1 or result row before checking split metadata.
  • Do not compare against the CCT-LSTM paper without noting protocol differences.

Current Next Actions

  • Verify split metadata before trusting results.
  • Prepare for experiment 2 in CCT-LSTM –> stress level classification with 3 classes (low/med/high).

Standalone Candidates

  • BP4D++ label limitations for stress estimation
  • Single-seed instability in small affective-computing datasets

2026-05-02

Project

Paper writing / experiment results review.

Done

  • Paper related work section.
  • Reviewed all clean p128 results for P1-P8.

    Decision

  • K.I.C.S. paper should only include P0/3/4.

    Risk

  • Deadline is close.

    Next

  • Finish method and experiment sections.
  • Prepare for submission.

2026-05-01

Project

Paper writing

Done

  • Paper introduction.

    Decision

  • None

    Risk

  • Intro and abstract are little too long and may need to be shortened.

    Next

  • Finish the related work section, then move on to method and experiment sections.
This post is licensed under CC BY 4.0 by the author.