wantedly / machine-learning-round-table

Gather around the table, and have a discussion to catch up the latest trend of machine learning 🤖
https://www.wantedly.com/projects/391912
297 stars 2 forks source link

[2024/05/22] Machine Learning 輪講 #248

Open nogawanogawa opened 1 month ago

nogawanogawa commented 1 month ago

Why

Machine Learning 輪講は最新の技術や論文を追うことで、エンジニアが「技術で解決できること」のレベルをあげていくことを目的にした会です。

prev. https://github.com/wantedly/machine-learning-round-table/issues/246

What

話したいことがある人はここにコメントしましょう! 面白いものを見つけた時点でとりあえず話すという宣言だけでもしましょう!

chimuichimu commented 1 month ago

GNNX-BENCH: Unravelling the Utility of Perturbation-based GNN Explainers through In-depth Benchmarking

GNN の最新動向 (ICLR 2024) の記事で紹介されていた GNN の説明性手法 のベンチマークをした論文

会議

ICLR 2024

背景

論文の概要

GNN の複数の説明性手法をベンチマークし、各種法の強みや制約、性能の評価を行う

説明性を得るアプローチ

Factual explanations

グラフ全体と同じ予測を導くための最小の部分グラフを特定することで、モデルを解釈するアプローチ

Counterfactual explanations

GNN への入力をどう変えたら出力が変わるか?という情報に基づいて、モデルを解釈するアプローチ

ベンチマークの観点

主な結果

nogawanogawa commented 1 month ago

Query Understanding in the Age of Large Language Models

論文URL

https://arxiv.org/abs/2306.16004

著者

Avishek Anand, Venktesh V, Abhijit Anand, Vinay Setty

会議

GENIR(SIGIR'23)

背景

LLMの登場によって、Web検索システムの様々なコンポーネントにLLMを用いようという動きが出ている。

IRの大きな問題の一つとして、ユーザーとシステム間の語彙の不一致を解消する事が上げられる。

従来の検索エンジンでは、query understandingとしてスペルチェック, クエリ分類等が古典的には行われていた。多くの場合、ユーザーはクエリ提案を用いるくらいで、最終的に使用されたクエリを知ることは無い。

目的

アプローチ

memo

https://github.com/nogawanogawa/paper_memo/issues/103