๊ฐ•์˜๋…ธํŠธ

๊ฐ•์˜ยท๊ฐ•์—ฐ์„ ๋“ค์œผ๋ฉฐ ์ •๋ฆฌํ•œ ๋…ธํŠธ.

๊ฐ•์˜ยท๊ฐ•์—ฐ์„ ๋“ค์œผ๋ฉฐ ์ •๋ฆฌํ•œ ๋…ธํŠธ.

32 posts · ์•„์นด์ด๋ธŒ์—์„œ ํ•„ํ„ฐ๋กœ ๋ณด๊ธฐ

[CS50] CS50x 2026 - Lecture 0 - Scratch

https://youtu.be/7ZJ4oo4h8vI ์ฐธ๊ณ  ์ž๋ฃŒ Harvard CS50: Introduction to Computer Science (https://cs50.harvard.edu/) ์›๋ณธ ์˜์ƒ: [ํ•œ๊ธ€๋”๋น™] โ€œAI๊ฐ€ ์ฝ”๋”ฉ ๋‹ค ํ•˜๋Š”๋ฐ ์™œ ๋ฐฐ์›Œ์š”?โ€ ํ•˜๋ฒ„๋“œ ๊ต์ˆ˜์˜ ์‚ฌ์ด๋‹ค ์ฐธ๊ต์œก (https://www.youtube.c...

[ํ† ํฌ] LLM ์™„๋ฒฝ ์ž…๋ฌธ ๊ฐ€์ด๋“œ: Andrej Karpathy ๊ฐ•์˜ ์ •๋ฆฌ

์›๋ณธ ๊ฐ•์˜: Intro to Large Language Models - Andrej Karpathy (https://youtu.be/zjkBMFhNj_g) Slides as PDF: https://drive.google.com/file/d/1pxx_ZI7O-Nwl7ZLNk5hI3WzAsTLwvNU7/view (42MB) All the imag...

[์ •๋ฆฌ] GPT-5 โ€˜์›์‚ฌ์ด์ฆˆ ํ•โ€™์˜ ์ฐฉ๊ฐ๊ณผ ๊ณจ๋””๋ฝ์Šค ํ”„๋กฌํ”„ํŠธ

๋ณธ ํฌ์ŠคํŠธ๋Š” ํ‹ฐํƒ€์ž„์ฆˆTV์™€ ๊ฐ•์ˆ˜์ง„ ๋ฐ•์‚ฌ๋‹˜์˜ ํŒจ๋„ ํ† ํฌ ์˜์ƒ์„ ๋ฐ”ํƒ•์œผ๋กœ ์ •๋ฆฌํ•œ ๋‚ด์šฉ์ž…๋‹ˆ๋‹ค. ์ถœ์ฒ˜: ํ‹ฐํƒ€์ž„์ฆˆTV - YouTube GPT-5์—์„œ ๋“œ๋Ÿฌ๋‚œ ์ƒ˜ ์•ŒํŠธ๋งŒ์˜ ์ฐฉ๊ฐ - ์˜์ƒ ๊ฐ•์ˆ˜์ง„ ๋ฐ•์‚ฌ๋‹˜ Linkedin Post ๐Ÿ“š TL;DR ํ•ต์‹ฌ ํ‰๊ฐ€: GPT-5๋Š” ์ฝ”๋”ฉ/์ˆ˜ํ•™ยท๊ธด ์ปจํ…์ŠคํŠธ๋Š” ๊ฐœ์„ ๋˜์—ˆ์ง€๋งŒ, ๊ธฐ๋Œ€ํ•œ ...

[NLP] 6. Topic Modeling์ด๋ž€?

1. Topic Modeling์ด๋ž€? ๋ณธ ๊ฐ•์˜๋Š” DSBA ๊ฐ•ํ•„์„ฑ ๊ต์ˆ˜๋‹˜์˜ ๊ฐ•์˜๋ฅผ ์ฐธ์กฐํ•˜์—ฌ ์ž‘์„ฑ๋˜์—ˆ์Šต๋‹ˆ๋‹ค. Topic Modeling์€ ๊ธฐ๊ณ„ ํ•™์Šต ๋ฐ ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ ๋ถ„์•ผ์—์„œ ๋ฌธ์„œ ์ง‘ํ•ฉ ๋‚ด์—์„œ ์ž ์žฌ์ ์ธ ์ฃผ์ œ(Latent Topic)๋ฅผ ๋ฐœ๊ฒฌํ•˜๊ธฐ ์œ„ํ•ด ์‚ฌ์šฉํ•˜๋Š” ํ†ต๊ณ„์  ๋ชจ๋ธ๋ง ๊ธฐ๋ฒ•์ž…๋‹ˆ๋‹ค. ์ฃผ์–ด์ง„ ๋ฌธ์„œ์—์„œ ๋ฐ˜๋ณต์ ์œผ๋กœ ๋“ฑ์žฅํ•˜๋Š” ๋‹จ์–ด ํŒจํ„ด์„ ๋ถ„์„ํ•˜์—ฌ...

[NLP] 5. ์ž์—ฐ์–ด ์ฐจ์› ์ถ•์†Œ(Dimension Reduction) ๊ธฐ๋ฒ•

์ฐจ์› ์ถ•์†Œ (Dimensionality Reduction) ๋ณธ ๊ฐ•์˜๋Š” DSBA ๊ฐ•ํ•„์„ฑ ๊ต์ˆ˜๋‹˜์˜ ๊ฐ•์˜๋ฅผ ์ฐธ์กฐํ•˜์—ฌ ์ž‘์„ฑ๋˜์—ˆ์Šต๋‹ˆ๋‹ค. 1. ์ฐจ์› ์ถ•์†Œ๋ž€ ๋ฌด์—‡์ธ๊ฐ€? ์ฐจ์› ์ถ•์†Œ๋Š” ๊ณ ์ฐจ์›์˜ ๋ฐ์ดํ„ฐ๋ฅผ ์ €์ฐจ์›์˜ ๋ฐ์ดํ„ฐ๋กœ ๋ณ€ํ™˜ํ•˜๋Š” ๊ธฐ๋ฒ•์ž…๋‹ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ๊ณ„์‚ฐ ํšจ์œจ์„ฑ์„ ๋†’์ด๊ณ , ๋ฐ์ดํ„ฐ ๋ถ„์„ ๋ฐ ์‹œ๊ฐํ™”๋ฅผ ์šฉ์ดํ•˜๊ฒŒ ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์ฐจ์› ์ถ•์†Œ๋Š” ๋‹ค์Œ ๋‘ ...

[์ž๋ฃŒ] KT ์‚ฌ๋‚ด ์ง์› ๋Œ€์ƒ RAG ๊ต์•ˆ ๋Œ€๋ฐฉ์ถœ?!

์˜ค๋Š˜๋„ RAG์— ๋Œ€ํ•œ ์ž๋ฃŒ๋ฅผ ์ฐพ์•„๋ณด๋˜ ์ค‘, KT DS์˜ ๊น€์„ฑ์šฐ ๊ธฐ์ˆ ํ˜์‹ ๋‹จ์žฅ๋‹˜๊ป˜์„œ ๊ณต์œ ํ•˜์‹  RAG ๊ด€๋ จ ์ž๋ฃŒ๋ฅผ ๋ฐœ๊ฒฌํ–ˆ์Šต๋‹ˆ๋‹ค. ์ถœ์ฒ˜ : LinkedIn - KTDS ๊น€์„ฑ์šฐ ๊ธฐ์ˆ ํ˜์‹ ๋‹จ์žฅ KT ์‚ฌ๋‚ด ์ง์› RAG ๊ต์•ˆ์ด๋ผ๋Š” ๋ง์— ์–ด๋–ค ๋‚ด์šฉ์ผ์ง€ ๊ถ๊ธˆํ•˜์—ฌ, ์ด๋ฒˆ ํฌ์ŠคํŒ…์—์„œ๋Š” ํ•ด๋‹น ์ž๋ฃŒ๋ฅผ ์ž์„ธํ•˜๊ฒŒ ๊ณต๋ถ€ํ•˜๋ฉฐ ๋‚ด์šฉ์„ ์ •๋ฆฌํ•ด๋ณด์•˜์œผ๋‹ˆ ํ•จ๊ป˜ ์‚ดํŽด๋ณด์‹œ์ฃ . ...

[๊ฐ•์˜๋…ธํŠธ] LangChain Academy : Introduction to LangGraph (Module 4)

๋žญ์ฒด์ธ(LangChain)๊ณผ ๋žญ๊ทธ๋ž˜ํ”„(LangGraph)๋Š” ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ(LLM)์„ ํ™œ์šฉํ•œ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ๊ฐœ๋ฐœ์„ ์œ„ํ•œ ๋„๊ตฌ๋“ค์ž…๋‹ˆ๋‹ค. ์œ„ ๊ฐ•์˜๋Š” LangChain์—์„œ ์šด์˜ํ•˜๋Š” LangChain Academy์—์„œ ์ œ์ž‘ํ•œ โ€œIntroduction to LangGraphโ€ ๊ฐ•์˜์˜ ๋‚ด์šฉ์„ ์ •๋ฆฌ ๋ฐ ์ถ”๊ฐ€ ์„ค๋ช…ํ•œ ๋‚ด์šฉ์ž…๋‹ˆ๋‹ค. ๊ฐ•์˜ ๋งํฌ : https://...

[๊ฐ•์˜๋…ธํŠธ] LangChain Academy : Introduction to LangGraph (Module 3)

๋žญ์ฒด์ธ(LangChain)๊ณผ ๋žญ๊ทธ๋ž˜ํ”„(LangGraph)๋Š” ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ(LLM)์„ ํ™œ์šฉํ•œ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ๊ฐœ๋ฐœ์„ ์œ„ํ•œ ๋„๊ตฌ๋“ค์ž…๋‹ˆ๋‹ค. ์œ„ ๊ฐ•์˜๋Š” LangChain์—์„œ ์šด์˜ํ•˜๋Š” LangChain Academy์—์„œ ์ œ์ž‘ํ•œ โ€œIntroduction to LangGraphโ€ ๊ฐ•์˜์˜ ๋‚ด์šฉ์„ ์ •๋ฆฌ ๋ฐ ์ถ”๊ฐ€ ์„ค๋ช…ํ•œ ๋‚ด์šฉ์ž…๋‹ˆ๋‹ค. ๊ฐ•์˜ ๋งํฌ : https://...

[๊ฐ•์˜๋…ธํŠธ] LangChain Academy : Introduction to LangGraph (Module 2)

๋žญ์ฒด์ธ(LangChain)๊ณผ ๋žญ๊ทธ๋ž˜ํ”„(LangGraph)๋Š” ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ(LLM)์„ ํ™œ์šฉํ•œ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ๊ฐœ๋ฐœ์„ ์œ„ํ•œ ๋„๊ตฌ๋“ค์ž…๋‹ˆ๋‹ค. ์œ„ ๊ฐ•์˜๋Š” LangChain์—์„œ ์šด์˜ํ•˜๋Š” LangChain Academy์—์„œ ์ œ์ž‘ํ•œ โ€œIntroduction to LangGraphโ€ ๊ฐ•์˜์˜ ๋‚ด์šฉ์„ ์ •๋ฆฌ ๋ฐ ์ถ”๊ฐ€ ์„ค๋ช…ํ•œ ๋‚ด์šฉ์ž…๋‹ˆ๋‹ค. ๊ฐ•์˜ ๋งํฌ : https://...

[๊ฐ•์˜๋…ธํŠธ] LangChain Academy : Introduction to LangGraph (Module 1)

๋žญ์ฒด์ธ(LangChain)๊ณผ ๋žญ๊ทธ๋ž˜ํ”„(LangGraph)๋Š” ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ(LLM)์„ ํ™œ์šฉํ•œ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ๊ฐœ๋ฐœ์„ ์œ„ํ•œ ๋„๊ตฌ๋“ค์ž…๋‹ˆ๋‹ค. ์œ„ ๊ฐ•์˜๋Š” LangChain์—์„œ ์šด์˜ํ•˜๋Š” LangChain Academy์—์„œ ์ œ์ž‘ํ•œ โ€œIntroduction to LangGraphโ€ ๊ฐ•์˜์˜ ๋‚ด์šฉ์„ ์ •๋ฆฌ ๋ฐ ์ถ”๊ฐ€ ์„ค๋ช…ํ•œ ๋‚ด์šฉ์ž…๋‹ˆ๋‹ค. ๊ฐ•์˜ ๋งํฌ : https:...

[๊ฐ•์˜๋…ธํŠธ] Text Splitting For Retrieval

Introduction ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ(LLM, Large Language Model)์„ ์ด์šฉํ•œ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์˜ ์„ฑ๋Šฅ์„ ํ–ฅ์ƒ์‹œํ‚ค๋Š” ๊ฐ€์žฅ ํšจ๊ณผ์ ์ธ ์ „๋žต ์ค‘ ํ•˜๋‚˜๋Š” ํฐ ํ…์ŠคํŠธ ๋ฐ์ดํ„ฐ๋ฅผ ๋” ์ž‘์€ ์กฐ๊ฐ์œผ๋กœ ๋ถ„ํ• ํ•˜๋Š” ๊ฒƒ์ž…๋‹ˆ๋‹ค. LLM์—๊ฒŒ ํ•„์š”ํ•œ ์ •๋ณด๋งŒ์„ ์ œ๊ณตํ•จ์œผ๋กœ์จ ๋ชจ๋ธ์˜ ์ž‘์—… ํšจ์œจ์„ฑ์„ ๊ทน๋Œ€ํ™”ํ•  ์ˆ˜ ์žˆ์Šต๋‹ˆ๋‹ค. ์ด๋Ÿฌํ•œ ํ…์ŠคํŠธ ๋ถ„ํ•  ๊ธฐ์ˆ ์€ ๋‹จ์ˆœํ•œ ๋ฐฉ๋ฒ• ๊ฐ™์ง€๋งŒ...

[๊ฐ•์˜๋…ธํŠธ] RAG From Scratch : RAG for long context LLMs

RAG for long context LLMs Introduction ์ตœ๊ทผ LLM(Long Large Models)์˜ ๊ธ‰๊ฒฉํ•œ ๋ฐœ์ „์œผ๋กœ ์ธํ•ด ๊ฑฐ๋Œ€ํ•œ ์–‘์˜ ๋ฐ์ดํ„ฐ๋ฅผ ํ•œ ๋ฒˆ์— ์ฒ˜๋ฆฌํ•  ์ˆ˜ ์žˆ๋Š” ๊ฐ€๋Šฅ์„ฑ์ด ์—ด๋ฆฌ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ํŠนํžˆ, 100๋งŒ ํ† ํฐ ์ด์ƒ์˜ ์ •๋ณด๋ฅผ ํ•œ ๋ฒˆ์— ์ฒ˜๋ฆฌํ•  ์ˆ˜ ์žˆ๋Š” ๋ชจ๋ธ๋“ค์ด ๋“ฑ์žฅํ•˜๋ฉด์„œ, ๊ณผ์—ฐ RAG(Retrieval-Augmented G...

[๊ฐ•์˜๋…ธํŠธ] RAG From Scratch : Query Translation

ํ•ด๋‹น ๋ธ”๋กœ๊ทธ ํฌ์ŠคํŠธ๋Š” RAG From Scratch : Coursework ๊ฐ•์˜ ํŒŒํŠธ 5 - 9 ๋‚ด์šฉ์„ ๋‹ค๋ฃจ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋น„๋””์˜ค ์š”์•ฝ ๊ฐ•์˜ ๋งํฌ ์Šฌ๋ผ์ด๋“œ Part 5 (๋‹ค์ค‘ ์ฟผ๋ฆฌ) ๋‹ค์–‘ํ•œ ๋ฌธ์„œ ๊ฒ€์ƒ‰์„ ์œ„ํ•ด ์ฟผ๋ฆฌ ์žฌ์ž‘์„ฑ ๊ธฐ๋ฒ•์„ ์„ค๋ช…ํ•ฉ...

[๊ฐ•์˜๋…ธํŠธ] RAG From Scratch : Query Routing & Structuring

ํ•ด๋‹น ๋ธ”๋กœ๊ทธ ํฌ์ŠคํŠธ๋Š” RAG From Scratch : Coursework ๊ฐ•์˜ ํŒŒํŠธ 10 - 11 ๋‚ด์šฉ์„ ๋‹ค๋ฃจ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋น„๋””์˜ค ์š”์•ฝ ๊ฐ•์˜ ๋งํฌ ์Šฌ๋ผ์ด๋“œ Part 10 (๋ผ์šฐํŒ…) ์ฟผ๋ฆฌ๋ฅผ ๊ด€๋ จ ๋ฐ์ดํ„ฐ ์†Œ์Šค๋กœ ์œ ๋„ํ•˜๊ธฐ ์œ„ํ•œ ๋…ผ๋ฆฌ์  ...

[๊ฐ•์˜๋…ธํŠธ] RAG From Scratch : Query Retrieval ๊ธฐ๋ฒ•

ํ•ด๋‹น ๋ธ”๋กœ๊ทธ ํฌ์ŠคํŠธ๋Š” RAG From Scratch : Coursework ๊ฐ•์˜ ํŒŒํŠธ 15 - 18 ๋‚ด์šฉ์„ ๋‹ค๋ฃจ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋ณ„๋„์˜ ๊ฐ•์ขŒ ๋‚ด์šฉ์ด ๋ณด์ด์ง€ ์•Š์•„์„œ ์‹ค์Šต ์ฝ”๋“œ๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ Reverse Enginneeringํ•ด์„œ ์ž๋ฃŒ๋ฅผ ์ •๋ฆฌํ•œ ๋‚ด์šฉ์ž…๋‹ˆ๋‹ค. ์ฐธ๊ณ  ๋ถ€ํƒ๋“œ๋ฆฝ๋‹ˆ๋‹ค! 1. Re-ranking (์žฌ์ •๋ ฌ) Re-ranking์€ ...

[๊ฐ•์˜๋…ธํŠธ] RAG From Scratch : Query Indexing ๊ธฐ๋ฒ•

ํ•ด๋‹น ๋ธ”๋กœ๊ทธ ํฌ์ŠคํŠธ๋Š” RAG From Scratch : Coursework ๊ฐ•์˜ ํŒŒํŠธ 12 - 14 ๋‚ด์šฉ์„ ๋‹ค๋ฃจ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋น„๋””์˜ค ์š”์•ฝ ๊ฐ•์˜ ๋งํฌ ์Šฌ๋ผ์ด๋“œ Part 12 (๋‹ค์ค‘ ํ‘œํ˜„ ์ธ๋ฑ์‹ฑ) ํšจ์œจ์ ์ธ ๊ฒ€์ƒ‰์„ ์œ„ํ•ด ๋ฌธ์„œ ์š”์•ฝ์„ ์ธ๋ฑ...

[๊ฐ•์˜๋…ธํŠธ] RAG From Scratch : Overview

ํ•ด๋‹น ๋ธ”๋กœ๊ทธ ํฌ์ŠคํŠธ๋Š” RAG From Scratch : Coursework ๊ฐ•์˜ ํŒŒํŠธ 1 - 4 ๋‚ด์šฉ์„ ๋‹ค๋ฃจ๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๋น„๋””์˜ค ์š”์•ฝ ๊ฐ•์˜ ๋งํฌ ์Šฌ๋ผ์ด๋“œ Part 1 (๊ฐœ์š”) RAG๋ฅผ ์†Œ๊ฐœํ•˜๋ฉฐ, ์‹œ๋ฆฌ์ฆˆ๊ฐ€ ๊ธฐ๋ณธ ๊ฐœ๋…๋ถ€ํ„ฐ ๊ณ ๊ธ‰ ๊ธฐ์ˆ ๊นŒ์ง€ ๋‹ค...

[๊ฐ•์˜๋…ธํŠธ] RAG From Scratch : Coursework

์†Œ๊ฐœ ์˜ค๋Š˜๋‚ ์˜ Generative AI๋Š” ๊ธฐ๋ณธ์ ์ธ ๋Œ€ํ˜• ์–ธ์–ด ๋ชจ๋ธ(LLM)์„ ๋„˜์–ด, ์ •๋ณด ํ™œ์šฉ์˜ ๋ฐฉ์‹์„ ํš๊ธฐ์ ์œผ๋กœ ๋ณ€ํ™”์‹œํ‚ค๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค. ๊ทธ ์ค‘์—์„œ๋„ RAG, Retrieval-Augmented Generation์€ ์™ธ๋ถ€ ๋ฐ์ดํ„ฐ ์†Œ์Šค๋ฅผ ํ™œ์šฉํ•˜์—ฌ ๋”์šฑ ์ •๊ตํ•˜๊ณ  ๋„๋ฉ”์ธ์— ํŠนํ™”๋œ ์ •๋ณด๋ฅผ ์ œ๊ณตํ•จ์œผ๋กœ์จ, AI์™€์˜ ์ƒํ˜ธ์ž‘์šฉ ๋ฐฉ์‹์„ ์ƒˆ๋กญ๊ฒŒ ์ •์˜ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค....

[๊ฟ€ํŒ] ํ”„๋กฌํ”„ํŠธ ์—”์ง€๋‹ˆ์–ด๋ง (๊ฐ•์˜ ์š”์•ฝ)

ํ”„๋กฌํ”„ํŠธ ์—”์ง€๋‹ˆ์–ด๋ง: ํ™˜๊ฐ ์ค„์ด๊ณ , ํšจ์œจ์ ์ธ SW ์„œ๋น„์Šค๊นŒ์ง€ ํ”„๋กฌํ”„ํŠธ ์—”์ง€๋‹ˆ์–ด๋ง ๋ถ„์•ผ์—์„œ ์ตœ๊ทผ ๋งŽ์€ ๊ด€์‹ฌ์„ ๋ฐ›๊ณ  ์žˆ๋Š” ์ฃผ์ œ๋Š” ๋ฐ”๋กœ โ€œAI ๋ชจ๋ธ์˜ ํ™˜๊ฐ(Hallucination) ํ˜„์ƒ์„ ์ค„์ด๊ณ , ๋ณด๋‹ค ์‹ ๋ขฐ์„ฑ ์žˆ๋Š” ๊ฒฐ๊ณผ๋ฅผ ์–ป๋Š” ๋ฐฉ๋ฒ•โ€์ž…๋‹ˆ๋‹ค. AI ๋ชจ๋ธ์˜ ํ™˜๊ฐ(Hallucination) ํ˜„์ƒ์ด๋ž€? AI ๋ชจ๋ธ์˜ ํ™˜๊ฐ(Hallucination) ํ˜„์ƒ์€...

[CV Notes] Lecture 18 - Videos

๋‹ค์Œ์€ ์•„๋ž˜ โ€œLecture 18. Videosโ€์— ๋Œ€ํ•œ ์š”์•ฝ ๋ฐ ํ•„๊ธฐ ๋‚ด์šฉ์„ ์ •๋ฆฌํ•œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ํ‹€๋ฆฐ ๋‚ด์šฉ์ด ์žˆ๋‹ค๋ฉด ๋Œ“๊ธ€ ๋ถ€ํƒ๋“œ๋ฆฝ๋‹ˆ๋‹ค ๐Ÿ™Œ Course Website: https://web.eecs.umich.edu/~justincj/teaching/eecs498/ Instructor: Justin Johnson Lecture 18: Vi...

[CV Notes] Lecture 17 - 3D Vision

๋‹ค์Œ์€ ์•„๋ž˜ Lecture์— ๋Œ€ํ•œ ์š”์•ฝ ๋ฐ ํ•„๊ธฐ ๋‚ด์šฉ์„ ์ •๋ฆฌํ•œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ํ‹€๋ฆฐ ๋‚ด์šฉ์ด ์žˆ๋‹ค๋ฉด ๋Œ“๊ธ€ ๋ถ€ํƒ๋“œ๋ฆฝ๋‹ˆ๋‹ค ๐Ÿ™Œ Course Website: https://web.eecs.umich.edu/~justincj/teaching/eecs498/ Instructor: Justin Johnson Lecture 17: 3D Vision 1...

[CS294] Deep Unsupervised Learning: Introduction

๋‹ค์Œ์€ ์•„๋ž˜ โ€œBerkeley CS294 ๊ฐ•์˜โ€์— ๋Œ€ํ•œ ์š”์•ฝ ๋ฐ ํ•„๊ธฐ ๋‚ด์šฉ์„ ์ •๋ฆฌํ•œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ํ‹€๋ฆฐ ๋‚ด์šฉ์ด ์žˆ๋‹ค๋ฉด ๋Œ“๊ธ€ ๋ถ€ํƒ๋“œ๋ฆฝ๋‹ˆ๋‹ค ๐Ÿ™Œ Course : CS294-158 SP24 Deep Unsupervised Learning Instructor: Pieter Abbeel Lecture # : L1. Introduction ๋ณธ ๊ฐ•์˜๋Š”...

[NLP] 4. Natural Language Embeddings

Natural Language Embeddings ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ(NLP)์—์„œ ํ…์ŠคํŠธ ๋ฐ์ดํ„ฐ๋ฅผ ํšจ์œจ์ ์œผ๋กœ ๋‹ค๋ฃจ๊ธฐ ์œ„ํ•ด ๋‹ค์–‘ํ•œ ์ž์—ฐ์–ด ์ž„๋ฒ ๋”ฉ ๊ธฐ๋ฒ•(Natural Language Embedding)์ด ์‚ฌ์šฉ๋ฉ๋‹ˆ๋‹ค. ์ด๋ฒˆ ํฌ์ŠคํŠธ์—์„œ๋Š” ์ด๋Ÿฌํ•œ ๊ธฐ๋ฒ•๋“ค์„ ์ž์„ธํžˆ ์„ค๋ช…ํ•˜๊ณ , ๊ฐ ๊ธฐ๋ฒ•์˜ ์˜ˆ์‹œ๋ฅผ ํ†ตํ•ด ์ดํ•ด๋ฅผ ๋•๊ณ ์ž ํ•ฉ๋‹ˆ๋‹ค. ๐Ÿ”Ž Text Representatio...

[NLP] 3. Natural Language Preprocessing

1. ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ(NLP) ๊ฐœ์š” ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ์˜ ์ผ๋ฐ˜์ ์ธ ์ˆœ์„œ ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ๋Š” ์Œ์„ฑ์„ ํ…์ŠคํŠธ๋กœ ๋ณ€ํ™˜ํ•˜๊ณ , ํ•ด๋‹น ํ…์ŠคํŠธ๋ฅผ ๋ถ„์„ ๋ฐ ์˜๋ฏธ๋ฅผ ์ถ”์ถœํ•œ ๋’ค, ์ด๋ฅผ ๋‹ค์‹œ ์Œ์„ฑ์œผ๋กœ ๋ณ€ํ™˜ํ•˜๋Š” ๊ณผ์ •์„ ํฌํ•จํ•ฉ๋‹ˆ๋‹ค. (์•„๋ž˜ ๊ทธ๋ฆผ ์ฐธ๊ณ ) ์ด ๊ณผ์ •์€ ํฌ๊ฒŒ STT(Speech to Text)์™€ TTS(Text to Speech)๋กœ ๋‚˜๋‰ฉ๋‹ˆ๋‹ค. 1.1 ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ...

[NLP] 2. Steps of Text Analytics

Steps of Text Analytics ํ…์ŠคํŠธ ๋ถ„์„์€ ๋น„์ •ํ˜• ํ…์ŠคํŠธ ๋ฐ์ดํ„ฐ๋ฅผ ๊ตฌ์กฐํ™”ํ•˜๊ณ  ์œ ์˜๋ฏธํ•œ ์ •๋ณด๋ฅผ ์ถ”์ถœํ•˜๊ธฐ ์œ„ํ•œ ์ผ๋ จ์˜ ๊ณผ์ •์ž…๋‹ˆ๋‹ค. ํ•ด๋‹น ํฌ์ŠคํŠธ์—์„œ๋Š” ํ…์ŠคํŠธ ๋ถ„์„์˜ ์ฃผ์š” ๋‹จ๊ณ„๋ฅผ ์ˆœ์ฐจ์ ์œผ๋กœ ์†Œ๊ฐœํ•˜์—ฌ, ๊ฐ ๋‹จ๊ณ„์—์„œ ์ˆ˜ํ–‰๋˜๋Š” ์ž‘์—…๊ณผ ๊ทธ ์ค‘์š”์„ฑ์„ ์„ค๋ช…ํ•˜๊ฒ ์Šต๋‹ˆ๋‹ค. โ‘  STEP 1. ์ •์˜ ๋ฐ ๋ฐ์ดํ„ฐ ์ˆ˜์ง‘(Define & Collect) ...

[NLP] 1. Introduction to Text Analytics

Introduction to Text Analytics Text Analytics๋Š” ๋ฌธ์„œ(text)์™€ ๊ฐ™์€ ๋น„์ •ํ˜• ๋ฐ์ดํ„ฐ๋ฅผ ๋ถ„์„ํ•˜์—ฌ ๊ทธ ์†์— ํฌํ•จ๋œ ์˜๋ฏธ ์žˆ๋Š” ์ •๋ณด๋ฅผ ์ถ”์ถœํ•˜๋Š” ๊ณผ์ •์„ ์˜๋ฏธํ•ฉ๋‹ˆ๋‹ค. ์ด๋Š” ์ž์—ฐ์–ด ์ฒ˜๋ฆฌ(Natural Language Processing, NLP) ๊ธฐ์ˆ ์„ ํ™œ์šฉํ•˜์—ฌ ๋‹ค์–‘ํ•œ ํ˜•ํƒœ์˜ ํ…์ŠคํŠธ ๋ฐ์ดํ„ฐ๋ฅผ ์ฒ˜๋ฆฌํ•˜๊ณ  ๋ถ„์„ํ•จ์œผ๋กœ์จ ...

[๋จธ์‹ ๋Ÿฌ๋‹][์ฐจ์›์ถ•์†Œ] ๋ณ€์ˆ˜ ์ถ”์ถœ๋ฒ• - Multi-Dimensional Scaling (MDS)

๋ณธ ํฌ์ŠคํŠธ๋Š” ๊ณ ๋ ค๋Œ€ํ•™๊ต ๊ฐ•ํ•„์„ฑ ๊ต์ˆ˜๋‹˜์˜ ๊ฐ•์˜๋ฅผ ์ˆ˜๊ฐ• ํ›„ ์ •๋ฆฌ๋ฅผ ํ•œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ž‘์„ฑ ๋ฐ ์„ค๋ช…์˜ ํŽธ์˜๋ฅผ ์œ„ํ•ด ์•„๋ž˜๋Š” ํŽธํ•˜๊ฒŒ ์ž‘์„ฑํ•œ ์  ์–‘ํ•ด๋ถ€ํƒ๋“œ๋ฆฝ๋‹ˆ๋‹ค. Dimensionality Reduction Supervised Variable Extraction ์ฐจ์›์ถ•์†Œ๋Š”, ๋ชจ๋ธ๋ง์„ ํ•˜๊ธฐ ์œ„ํ•ด ๋‚ด๊ฐ€ ๊ฐ€์ง„ ๋ฐ์ดํ„ฐ์˜ ์ •๋ณด๋ฅผ ์ตœ๋Œ€ํ•œ ๋ณด์กดํ•˜๋ฉด์„œ, ํ›จ์”ฌ ๋” compac...

[๋จธ์‹ ๋Ÿฌ๋‹] ์ด์ƒ ํƒ์ง€ ๊ฐœ์š” ๋ฐ ๋ฐ€๋„ ๊ธฐ๋ฐ˜ ์ด์ƒ์น˜ํƒ์ง€

๋ณธ ํฌ์ŠคํŠธ๋Š” ๊ณ ๋ ค๋Œ€ํ•™๊ต ๊ฐ•ํ•„์„ฑ ๊ต์ˆ˜๋‹˜์˜ ๊ฐ•์˜๋ฅผ ์ˆ˜๊ฐ• ํ›„ ์ •๋ฆฌ๋ฅผ ํ•œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ž‘์„ฑ ๋ฐ ์„ค๋ช…์˜ ํŽธ์˜๋ฅผ ์œ„ํ•ด ์•„๋ž˜ ํฌ์ŠคํŠธ๋Š” ๋ฐ˜๋ง๋กœ ์ž‘์„ฑํ•œ ์  ์–‘ํ•ด๋ถ€ํƒ๋“œ๋ฆฝ๋‹ˆ๋‹ค. Abnormal Data๋ž€ Anomaly Data๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์ด Hawkins์™€ Harmeling์— ์˜ํ•ด ์ •์˜๋œ๋‹ค. Observations that deviate so much f...

[๋จธ์‹ ๋Ÿฌ๋‹] ๊ฑฐ๋ฆฌโ€ข๊ตฐ์ง‘โ€ข์„œํฌํŠธ๋ฒกํ„ฐ ๊ธฐ๋ฐ˜ ์ด์ƒํƒ์ง€ ๊ธฐ๋ฒ•

๋ณธ ํฌ์ŠคํŠธ๋Š” ๊ณ ๋ ค๋Œ€ํ•™๊ต ๊ฐ•ํ•„์„ฑ ๊ต์ˆ˜๋‹˜์˜ ๊ฐ•์˜๋ฅผ ์ˆ˜๊ฐ• ํ›„ ์ •๋ฆฌ๋ฅผ ํ•œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ž‘์„ฑ ๋ฐ ์„ค๋ช…์˜ ํŽธ์˜๋ฅผ ์œ„ํ•ด ์•„๋ž˜ ํฌ์ŠคํŠธ๋Š” ๋ฐ˜๋ง๋กœ ์ž‘์„ฑํ•œ ์  ์–‘ํ•ด๋ถ€ํƒ๋“œ๋ฆฝ๋‹ˆ๋‹ค. ๊ฑฐ๋ฆฌ ๊ธฐ๋ฐ˜ ์ด์ƒ์น˜ํƒ์ง€ K-Nearest Neighbor-based Anomaly Detection ๊ฐ ๋ฐ์ดํ„ฐ์— ๋Œ€ํ•œ Anomaly Score๋ฅผ K๊ฐœ์˜ ๊ทผ์ ‘ ์ด์›ƒ๊นŒ์ง€์˜ ๊ฑฐ๋ฆฌ๋ฅผ ์ด์šฉํ•˜์—ฌ ๊ณ„์‚ฐ...

[๋จธ์‹ ๋Ÿฌ๋‹][์‹œ๊ณ„์—ด] AR, MA, ARMA, ARIMA์˜ ๋ชจ๋“  ๊ฒƒ - ๊ฐœ๋…ํŽธ

์˜ค๋Š˜์€ ๋จธ์‹ ๋Ÿฌ๋‹ ์‹œ๊ณ„์—ด์—์„œ ๊ฐ€์žฅ ๋งŽ์ด ์“ฐ์ด๋Š” AR, MA, ARMA, ARIMA์— ๋Œ€ํ•ด ์ •๋ฆฌํ•ด๋ณด๋Š” ์‹œ๊ฐ„์„ ๊ฐ€์ง€๋ ค๊ณ  ํ•ฉ๋‹ˆ๋‹ค. ํ•ด๋‹น ํฌ์ŠคํŠธ๋Š” ๊ณ ๋ ค๋Œ€ํ•™๊ต ๊น€์„ฑ๋ฒ” ๊ต์ˆ˜๋‹˜์˜ ๊ฐ•์˜๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ ์ œ์ž‘๋˜์—ˆ์Šต๋‹ˆ๋‹ค. ๋ชฉ์ฐจ ์ •์ƒ ํ”„๋กœ์„ธ์Šค์™€ ๋น„์ •์ƒ ํ”„๋กœ์„ธ์Šค Autoregressive (AR) Models Moving Average (MA) Models A...

[๋จธ์‹ ๋Ÿฌ๋‹][์ฐจ์›์ถ•์†Œ] ๋ณ€์ˆ˜ ์ถ”์ถœ๋ฒ• - Principal Component Analysis (PCA)

๋ณธ ํฌ์ŠคํŠธ๋Š” ๊ณ ๋ ค๋Œ€ํ•™๊ต ๊ฐ•ํ•„์„ฑ ๊ต์ˆ˜๋‹˜์˜ ๊ฐ•์˜๋ฅผ ์ˆ˜๊ฐ• ํ›„ ์ •๋ฆฌ๋ฅผ ํ•œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ž‘์„ฑ ๋ฐ ์„ค๋ช…์˜ ํŽธ์˜๋ฅผ ์œ„ํ•ด ์•„๋ž˜๋Š” ํŽธํ•˜๊ฒŒ ์ž‘์„ฑํ•œ ์  ์–‘ํ•ด๋ถ€ํƒ๋“œ๋ฆฝ๋‹ˆ๋‹ค. Dimensionality Reduction Supervised Variable Extraction ์ฐจ์›์ถ•์†Œ๋Š”, ๋ชจ๋ธ๋ง์„ ํ•˜๊ธฐ ์œ„ํ•ด ๋‚ด๊ฐ€ ๊ฐ€์ง„ ๋ฐ์ดํ„ฐ์˜ ์ •๋ณด๋ฅผ ์ตœ๋Œ€ํ•œ ๋ณด์กดํ•˜๋ฉด์„œ, ํ›จ์”ฌ ๋” compa...

[๋จธ์‹ ๋Ÿฌ๋‹][์ฐจ์›์ถ•์†Œ] ๋ณ€์ˆ˜ ์„ ํƒ๋ฒ•

๋ณธ ํฌ์ŠคํŠธ๋Š” ๊ณ ๋ ค๋Œ€ํ•™๊ต ๊ฐ•ํ•„์„ฑ ๊ต์ˆ˜๋‹˜์˜ ๊ฐ•์˜๋ฅผ ์ˆ˜๊ฐ• ํ›„ ์ •๋ฆฌ๋ฅผ ํ•œ ๊ฒƒ์ž…๋‹ˆ๋‹ค. ์ž‘์„ฑ ๋ฐ ์„ค๋ช…์˜ ํŽธ์˜๋ฅผ ์œ„ํ•ด ์•„๋ž˜ ํฌ์ŠคํŠธ๋Š” ๋ฐ˜๋ง๋กœ ์ž‘์„ฑํ•œ ์  ์–‘ํ•ด๋ถ€ํƒ๋“œ๋ฆฝ๋‹ˆ๋‹ค. Dimensionality Reduction Curse of dimensionality ์ •์˜ ์ด๋ก ์ (theory)์œผ๋กœ๋Š” ๋ณ€์ˆ˜์˜ ๊ฐœ์ˆ˜๊ฐ€ ์ฆ๊ฐ€ํ•  ๋•Œ ๋ชจ๋ธ์˜ ์„ฑ๋Šฅ๋„ ์ฆ๊ฐ€ํ•œ๋‹ค. ํ•˜์ง€๋งŒ, ํ˜„์‹ค...