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

[2023/09/13]Machine Learning 輪講 #212

Open nogawanogawa opened 10 months ago

nogawanogawa commented 10 months 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/210

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!

zerebom commented 9 months ago

Retentive Network: A Successor to Transformer for Large Language Models

nogawanogawa commented 9 months ago

Vector Search with OpenAI Embeddings: Lucene Is All You Need

論文URL

https://arxiv.org/abs/2308.14963

著者

Jimmy Lin, Ronak Pradeep, Tommaso Teofili, Jasper Xian

会議

(なし)

背景

昨今のNNの検索への応用は、最近ではembeddingによって表現される表現学習に焦点が当てられている。

現在では、embeddingを用いた検索には多数のembeddingを管理する必要があり、ベクトルデータベースが必要といった主張も少なくない。

現在広く普及しているLuceneベースの検索エコシステムは、表現学習の最近のトレンドへの適応が遅れているかもしれないが、一方で大きな投資を行っており、今後ベクトルデータベースなしで検索にembeddingに組み込む際の機能的な差分は見当たらない。

このように、embeddingを用いた検索は有望であるが、そのためにベクトルデータベースが必要であるということにはならない。

目的

追加のベクトルデータベースなしでembedding検索が十分実用的であることを示す。

アプローチ

memo

https://github.com/nogawanogawa/paper_memo/issues/97