Loading views...

Personal Research

Creator
Creator
Seonglae ChoSeonglae Cho
Created
Created
2023 Mar 6 12:14
Editor
Edited
Edited
2026 Jan 12 12:44

Minimize hyper parameters. Do not let research exploration become exponential

Heuristic Exploration

  • Early stage: coarse sweep (log scale)
  • Mid stage: local search around best basin
  • Late stage: novelty search/avoid duplicates (exclude already explored areas)
Parameter layer: polynomial (low-order interaction) coverage
  • Assumption: most important effects are 1st order (single variable) + 2nd order (pairwise interaction)
  • Goal: secure as much "pairwise coverage" as possible with minimal trials
Personal Researches
 
 
Personal paper ideas
 

Tips for AI conference

  • ACL
    prefers stories telling inter disciplines different element

Research core

  • motivation
  • idea
  • baseline
  • statistical robustness
  • insight
Except them, all decoration
 
notion image
notion image
 
 
Submission Dates and Process for EACL/NAACL and ACL 2024 | ACL Member Portal
The use of an ARR-only submission process has been agreed by Program Chairs (PCs), and General Chairs of the upcoming conferences in 2024 (EACL, NAACL and ACL, referred to in this document as *CL conferences), with feedback from the ACL Rolling Review (ARR) Initiative, the ARR board, and the ACL Exec. There will be no direct submission option. This decision was announced during the ACL 2023 business meeting. The purpose of this communication is to update those that were not present, and to provide more details about the coordinated plan.
 
 
 

Recommendations