๋ณธ๋ฌธ ๋ฐ”๋กœ๊ฐ€๊ธฐ
728x90
๋ฐ˜์‘ํ˜•

๋Œ€์™ธํ™œ๋™9

TOBIG's [์‹ฌ์ธตํ•™์Šต] ์ œ5์žฅ ๊ธฐ๊ณ„ ํ•™์Šต์˜ ๊ธฐ์ดˆ 5.1 ํ•™์Šต ์•Œ๊ณ ๋ฆฌ์ฆ˜ 5.1.1 ๊ณผ์ œ T - ๊ธฐ๊ณ„ํ•™์Šต์˜ ๊ณผ์ œ๋Š” ์ผ๋ฐ˜์ ์œผ๋กœ ๊ธฐ๊ณ„ ํ•™์Šต ์‹œ์Šคํ…œ์ด ๊ฒฌ๋ณธ์„ ์ฒ˜๋ฆฌํ•˜๋Š” ๋ฐฉ์‹์„ ์„œ์ˆ ํ•˜๋Š” ํ˜„ํƒœ๋กœ ์ •์˜๋œ๋‹ค. ์—ฌ๊ธฐ์„œ ๊ฒฌ๋ณธ์ด๋ž€ ๊ธฐ๊ณ„ ํ•™์Šต ์‹œ์Šคํ…œ์˜ ์ฒ˜๋ฆฌ ๋Œ€์ƒ์ธ ์–ด๋–ค ๋ฌผ์ฒด๋‚˜ ์‚ฌ๊ฑด์œผ๋กœ๋ถ€ํ„ฐ ์ •๋Ÿ‰์ ์œผ๋กœ ์ธก์ •ํ•œ ํŠน์ง•(feature๋“ค์˜ ์ง‘ํ•ฉ)์„ ์˜๋ฏธํ•œ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด ๊ฒฌ๋ณธ ๋””์ง€ํ„ธ ์ด๋ฏธ์ง€์˜ ํŠน์ง•๋“ค์€ ์ด๋ฏธ์ง€๋ฅผ ๊ตฌ์„ฑํ•˜๋Š” ํ”ฝ์…€๊ฐ’์ด๋‹ค. ๋‹ค์Œ์€ ๊ฐ€์žฅ ํ”ํ•œ ๊ธฐ๊ณ„ ํ•™์Šต ๊ณผ์ œ ๋ช‡ ๊ฐ€์ง€์ด๋‹ค. 1. ๋ถ„๋ฅ˜ 2. ๊ฒฐ์ธก ์ž…๋ ฅ์ด ์žˆ๋Š” ์ž๋ฃŒ์˜ ๋ถ„๋ฅ˜ : ์ž…๋ ฅ ๋ฒกํ„ฐ์˜ ๋ชจ๋“  ์ธก๋„๊ฐ€ ํ•ญ์ƒ ๋ณด์žฅ์ด ์—†์„ ๋•Œ๋Š” ๋ถ„๋ฅ˜๊ฐ€ ๋” ์–ด๋ ค์›Œ์ง„๋‹ค. ๋”ฐ๋ผ์„œ, ๋ชจ๋“  ๊ด€๋ จ ๋ณ€์ˆ˜์— ๊ด€ํ•œ ํ™•๋ฅ ๋ถ„ํฌ๋ฅผ ํ•™์Šตํ•˜๊ณ , ๊ฒฐ์ธก๊ฐ’๋“ค์„ ์ฃผ๋ณ€ํ™”ํ•ด์„œ ๋ถ„๋ฅ˜ ๋ฌธ์ œ๋ฅผ ํ’€๊ธฐ๋„ ํ•œ๋‹ค. 3. ํšŒ๊ท€ 4. ์ „์‚ฌ(์˜ฎ๊ฒจ์“ฐ๊ธฐ) : ํ•ด๋‹น ์ข…๋ฅ˜์˜ ๊ณผ์ œ์—์„œ ๊ธฐ๊ณ„ ํ•™์Šต ์‹œ์Šคํ…œ์€ ๋น„๊ต์  ๊ตฌ์กฐ์ ์ด.. 2024. 2. 2.
[TOBIG's] ์‹ฌ์ธตํ•™์Šต - ์ œ4์žฅ ์ˆ˜์น˜๊ณ„์‚ฐ ์ผ๋ฐ˜์ ์œผ๋กœ ๊ธฐ๊ณ„ ํ•™์Šต ์•Œ๊ณ ๋ฆฌ์ฆ˜์—๋Š” ๋Œ€๋Ÿ‰์˜ ์ˆ˜์น˜ ๊ณ„์‚ฐ์ด ํ•„์š”ํ•˜๋‹ค. ์–ด๋–ค ๋ฐ˜๋ณต์ ์ธ ๊ณผ์ •์„ ํ†ตํ•ด ์ •๋‹ต์˜ ์ถ”์ •๊ฐ’์„ ๊ณ„์† ๊ฐฑ์‹ ํ•จ์œผ๋กœ์จ ๋ฌธ์ œ๋ฅผ ํ’€์ง€๋งŒ, ์œ ํ•œํ•œ ๋ฉ”๋ชจ๋ฆฌ๋ฅผ ๊ฐ€์ง„ ๋””์ง€ํ„ธ ์ปดํ“จํ„ฐ๋Š” ์‹ค์ˆ˜๋ฅผ ์ •ํ™•ํ•˜๊ฒŒ ํ‘œํ˜„ํ•  ์ˆ˜ ์—†๊ธฐ์— ์•ฝ๊ฐ„์˜ ํ•œ๊ณ„๊ฐ€ ์กด์žฌํ•œ๋‹ค. 4.1 ๋„˜์นจ๊ณผ ์•„๋ž˜๋„˜์นจ - ์•„๋ž˜๋„˜์นจ(underflow) : 0์— ๊ฐ€๊นŒ์šด ์ˆ˜๊ฐ€ ๋ฐ˜์˜ฌ๋ฆผ ๋•Œ๋ฌธ๋ฐ ์ •ํ™•ํžˆ 0์ด ๋˜๋Š” ๊ฒƒ์„ ๋งํ•œ๋‹ค. - ๋„˜์นจ(overflow) : ํฌ๊ธฐ๊ฐ€ ํฐ ์ˆ˜๊ฐ€ ๋ฌดํ•œ๋Œ€ ํ˜น์€ ๋งˆ์ด๋„ˆ์Šค ๋ฌดํ•œ๋Œ€๋กœ ๊ทผ์‚ฌ๋˜๋Š” ๊ฒƒ์„ ๋งํ•œ๋‹ค. -> ์ด๋Ÿฌํ•œ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๋Š”๋ฐ ์“ฐ์ด๋Š” ํ•จ์ˆ˜๊ฐ€ ๋ฐ”๋กœ ์†Œํ”„ํŠธ๋งฅ์Šค ํ•จ์ˆ˜์ด๋‹ค. ๋ฉ€ํ‹ฐ๋ˆ„์ด ๋ถ„ํฌ๋ฅผ ๋”ฐ๋ฅด๋Š” ํ™•๋ฅ ๋“ค์„ ์˜ˆ์ธกํ•˜๋Š”๋ฐ ํ”ํžˆ ์‚ฌ์šฉ๋œ๋‹ค. 4.2 ๋‚˜์œ ์กฐ๊ฑดํ™” - ์กฐ๊ฑดํ™” : ์ž…๋ ฅ์˜ ์ž‘์€ ๋ณ€ํ™”์— ๋Œ€ํ•ด ํ•จ์ˆ˜๊ฐ€ ์–ผ๋งˆ๋‚˜ ๊ธ‰ํ•˜๊ฒŒ ๋ณ€ํ•˜๋Š”์ง€๋ฅผ ๋œปํ•˜๋Š” ์šฉ์–ด์ด๋‹ค. ๊ณผํ•™ ๊ณ„์‚ฐ์—.. 2024. 2. 2.
LG Aimers 3๊ธฐ ์ˆ˜๋ฃŒ 2023. 12. 31.
Module 7. ๋”ฅ๋Ÿฌ๋‹ (Deep Learning) (KAIST ์ฃผ์žฌ๊ฑธ ๊ต์ˆ˜) ๋‚ ์งœ: 2023๋…„ 7์›” 15์ผ Part 1. Introduction to Deep Neural Networks 1. Deep Learning : ์‹ ๊ฒฝ์„ธํฌ๋“ค์ด ๋ง์„ ์ด๋ฃจ์–ด์„œ ์ •๋ณด๋ฅผ ๊ตํ™˜ํ•˜๊ณ  ์ฒ˜๋ฆฌํ•˜๋Š” ๊ณผ์ •์„ ๋ณธ๋”ฐ์„œ ๋งŒ๋“  ๋ฐฉ์‹์„ ์˜๋ฏธํ•จ 2. ์‹ฌ์ธต ์‹ ๊ฒฝ๋ง์˜ ๊ธฐ๋ณธ ๋™์ž‘ ๊ณผ์ • - Big Data์˜ ํ•„์š” - GPU Acceleration - Algorithm Improvements 3. Perceptron - ํผ์…‰ํŠธ๋ก ์€ ์ƒ๋ฌผํ•™์ ์ธ ์‹ ๊ฒฝ๊ณ„(Neual Network)์˜ ๊ธฐ๋ณธ ๋‹จ์œ„์ธ ์‹ ๊ฒฝ์„ธํฌ(=๋‰ด๋Ÿฐ)์˜ ๋™์ž‘ ๊ณผ์ •์„ ํ†ต๊ณ„ํ•™์ ์œผ๋กœ ๋ชจ๋ธ๋งํ•œ ์•Œ๊ณ ๋ฆฌ์ฆ˜ 4. Forward Propagation - ํ–‰๋ ฌ ๊ณฑ์„ ํ†ตํ•ด sigmoid function๊ณผ ๊ฐ™์€ actiavtion function์„ ์ง€๋‚˜๋ฉด ๊ฒฐ๊ณผ ๊ฐ’์ด ๋‚˜์˜ด 5. MSE - ์—๋Ÿฌ.. 2023. 7. 15.
Module 6. ๊ฐ•ํ™”ํ•™์Šต (Reinforcement Learning) (๊ณ ๋ ค๋Œ€ํ•™๊ต ์ด๋ณ‘์ค€ ๊ต์ˆ˜) ๋‚ ์งœ: 2023๋…„ 7์›” 13์ผ Part 1. MDP and Planning : Markov Decision Process์˜ ์•ฝ์ž Sequential Decision Making under Uncertainty๋ฅผ ์œ„ํ•œ ๊ธฐ๋ฒ• ๊ฐ•ํ™”ํ•™์Šต(Reinforcement Learning, RL)์„ ์œ„ํ•œ ๊ธฐ๋ณธ ๊ธฐ๋ฒ• ์•Œ๊ณ ๋ฆฌ์ฆ˜(transition probability, reward function)์„ ์•Œ๊ณ  ์žˆ์„ ๋•Œ๋Š” MDP(stocasitc control ๊ธฐ๋ฒ•)์„ ์ด์šฉ ์•Œ๊ณ ๋ฆฌ์ฆ˜์„ ๋ชจ๋ฅด๊ณ  simulation ๊ฒฐ๊ณผ(reward ๊ฐ’)๋ฅผ ํ™œ์šฉํ•  ๋•Œ๋Š” ๊ฐ•ํ™”ํ•™์Šต์„ ์ด์šฉ https://velog.io/@recoder/MDP%EC%9D%98%EA%B0%9C%EB%85%90 S : set of states(state space) state .. 2023. 7. 15.
Module 5. ์ง€๋„ํ•™์Šต (๋ถ„๋ฅ˜/ํšŒ๊ท€) (์ดํ™”์—ฌ์ž๋Œ€ํ•™๊ต ๊ฐ•์ œ์› ๊ต์ˆ˜) ๋‚ ์งœ: 2023๋…„ 7์›” 8์ผ Part 1. SL Foundation 1.Supervised Learning - label๊ฐ’์ด ์žˆ๋Š” ๊ฒƒ์„ ๋งํ•จ - training๊ณผ test ๋‹จ๊ณ„๊ฐ€ ์กด์žฌํ•จ - feature์˜ ๊ฒฝ์šฐ, domain ์ง€์‹์ด ์–ด๋Š ์ •๋„ ํ•„์š”ํ•จ - ๋”ฅ๋Ÿฌ๋‹์˜ ๊ฒฝ์šฐ, feature๋ฅผ ์Šค์Šค๋กœ ํ•™์Šตํ•˜๊ธฐ๋„ ํ•จ - SL์˜ ๊ฒฝ์šฐ, training error, val error, test error์„ ํ†ตํ•ด generalization error์„ ์ตœ์†Œํ™”ํ•˜๋„๋ก ํ•˜๋Š” ๋…ธ๋ ฅ์„ ํ•˜๊ฒŒ ๋จ - loss function=cost function 2. Bias-variance trade-off - bias์™€ variance์˜ trade off๋ฅผ ์ž˜ ์กฐ์ •ํ•ด์„œ ์ตœ์ ์˜ generalization error๋ฅผ ๋งŒ๋“œ๋Š” ๊ฒƒ์ด ์ค‘์š”ํ•จ - ๋”ฅ.. 2023. 7. 8.
728x90
๋ฐ˜์‘ํ˜•