๊ณ ์ฐจ์›์—์„œ ๊ตญ์†Œ์  ๋ฐฉ๋ฒ•(interpolate->extrapolate)
ยท
๐Ÿ“ Statistics
K-NN (k-์ตœ๊ทผ์ ‘์ด์›ƒ) *์•ˆ์ •์ ์ด์ง€๋งŒ ํŽธํ–ฅ๋œ ์„ ํ˜• ๋ชจ๋ธ vs ๋œ ์•ˆ์ •์ ์ด์ง€๋งŒ ๋œ ํŽธํ–ฅ์ ์ธ ๋ชจ๋ธ *์–ด๋– ํ•œ x๋“ ์ง€ ์ด์— ๊ฐ€๊นŒ์šด ๊ด€์ธก์น˜์˜ ์ด์›ƒ์„ ๊ฝค ๋งŽ์ด ์ฐพ๊ณ  ํ‰๊ท ์„ ๊ตฌํ•  ์ˆ˜ ์žˆ๋‹ค. -> ํ•˜์ง€๋งŒ high dimension์—์„œ ์ด ๋ฐฉ์‹์€ ํ†ตํ•˜์ง€ ์•Š๋Š”๋‹ค. curse of dimensionality(Bellman, 1961) p์ฐจ์› ์œ ๋‹›์˜ ์ดˆ์ž…๋ฐฉ์ฒด๋‚ด ๊ท ์ผํ•˜๊ฒŒ ๋ถ„ํฌ๋œ ์ž…๋ ฅ๊ฐ’์— ๊ด€ํ•œ K-NN ๋ชจ๋ธ ๊ฐ€์ • $e_{r}$= $r^\frac{1}{p}$ ๊ณ ์ฐจ์›์—์„œ ํ‘œ์ง‘์˜ ๋‹ค๋ฅธ ์ค‘์š”ํ•œ ์ ์€ ๋ชจ๋“  ํ‘œ๋ณธ ์ง€์ ๋“ค์ด ํ‘œ๋ณธ์˜ ๋ชจ์„œ๋ฆฌ์™€ ๊ฐ€๊น๋‹ค๋Š” ๊ฒƒ์ด๋‹ค. N๊ฐœ์˜ ๋ฐ์ดํ„ฐ ์ง€์ ์ด ์žˆ์„ ๋•Œ ์›์ ์œผ๋กœ๋ถ€ํ„ฐ ๊ฐ€์žฅ ๊ฐ€๊นŒ์šด ๋ฐ์ดํ„ฐ ์ง€์ ์˜ ์ค‘์•™๊ฐ’์˜ ๊ฑฐ๋ฆฌ๋ฅผ $d(p,N) = (1 - \frac{1}{2}^(1/N))^(1/p)$ N=500, p=10 d..
[M1 Transformers] M1 Mac Transformers Install Error
ยท
๐Ÿ—ฃ๏ธ Natural Language Processing
https://medium.com/@alfarruggia/from-zero-to-transformers-on-apple-m1-chip-e5d9afa35f5 From Zero to Transformers on Apple M1 Chip A copy-and-paste guide to install Tensorflow and Transformers on Apple M1 Chip. medium.com pip์—์„œ rust๋ฅผ ํ†ตํ•ด ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ๋ฅผ ์„ค์น˜ํ•˜๋Š”๋ฐ ๋ฌธ์ œ๊ฐ€ ์žˆ์–ด,, ์•„๋ฌด๋ฆฌ ํ™˜๊ฒฝ ์„ค์ •์„ ์žก์•„๋„ ๋˜์ง€ ์•Š์•„ conda install -c huggingface transformers -y conda๋ฅผ ํ†ตํ•ด ํ•ด๊ฒฐ
[python 3.8-] Print Format
ยท
๐Ÿ Python
from datetime import datetime user = 'eric_idle' member_since = datetime(1975, 7, 31) f'{user=} {member_since=}' result -> "user='eric_idle' member_since=datetime.date(1975, 7, 31)" python์—์„œ f-format์œผ๋กœ print ์‹œ ํŒŒ๋ผ๋ฏธํ„ฐ์™€ ํ•จ๊ป˜ ์ ์–ด ์ค„ ์ˆ˜ ์žˆ๋Š” ๋ฐฉ๋ฒ•์ด Release๋˜์—ˆ๋‹ค. ๋ฐ”๋‹ค ์ฝ”๋ผ๋ฆฌ (walrus operator) :=
[*args, **kwargs] ํ•จ์ˆ˜ ํŒŒ๋ผ๋ฏธํ„ฐ ๊ตฌ๋ถ„์ž
ยท
๐Ÿ Python
https://docs.python.org/3/whatsnew/3.8.html What’s New In Python 3.8 — Python 3.10.5 documentation What’s New In Python 3.8 Editor Raymond Hettinger This article explains the new features in Python 3.8, compared to 3.7. Python 3.8 was released on October 14, 2019. For full details, see the changelog. Summary – Release highlights New Features Assignm docs.python.org ์œ„์น˜ ์ „์šฉ ๋งค๊ฐœ ๋ณ€์ˆ˜ ์ผ๋ถ€ ํ•จ์ˆ˜ ๋งค๊ฐœ ๋ณ€์ˆ˜๋ฅผ ์œ„์น˜์ ์œผ๋กœ..
Monitor ๋น„๊ต ์‚ฌ์ดํŠธ
ยท
๐Ÿƒ Routine
https://www.displayspecifications.com/ DisplaySpecifications - Specifications and features of desktop monitors and TVs Detailed specifications of desktop monitors, smart TVs and other types of displays. The latest display-related news. Comparisons of the specifications of different models, user reviews and ratings. www.displayspecifications.com https://www.displayspecifications.com/en/model/e6d8..
[Model Train Error] ๋ชจ๋ธ ํ•™์Šต ์‹œ ๋ฌดํ•œ ๋ฃจํ”„ /Unknown batch
ยท
๐Ÿ‘พ Deep Learning
3498/Unknown - 10ms/step - loss: 0.1356 - 3585/Unknown - 1029s 282ms/step - loss: 0.13 3586/Unknown - 1029s 282ms/step - loss: 0.1323 - accuracy: 0. 3587/1030s 282ms/step - loss: 0.1322 - accuracy: 0. ImageGenerator๋ฅผ ์‚ฌ์šฉํ•˜๊ณ  model์„ ํ•™์Šต ์‹œํ‚ค๋ฉด generator๊ฐ€ image์˜ ๊ฐœ์ˆ˜๋ฅผ ์•Œ์ง€ ๋ชปํ•ด ๋ฌดํ•œ์ด ํ•™์Šต์„ ๋ฐ˜๋ณตํ•œ๋‹ค. (28์‹œ๊ฐ„ ํ•™์Šต.. ) https://stackoverflow.com/questions/61266275/epoch-1-2-103-unknown-8s-80ms-step-loss-0-0175-model-fit-keeps..
๋‹คํ–ˆ๋‹ค
B's