Computer Science/Data Science

λ¨Έμ‹ λŸ¬λ‹/AIμ—μ„œ μ‚¬μš©λ˜λŠ” "Ground Truth" 뜻

_cactus 2021. 3. 5. 11:19
λ°˜μ‘ν˜•

λ¨Έμ‹ λŸ¬λ‹μ— κ΄€λ ¨ν•œ 글을 읽닀보면 "ground-truth"λΌλŠ” μš©μ–΄λ₯Ό 많이 μ ‘ν•˜κ²Œ λœλ‹€.
"ground-truth"λŠ” κΈ°μƒν•™μ—μ„œ μœ λž˜ν•˜μ˜€μœΌλ©°, μ–΄λŠ ν•œ μž₯μ†Œμ—μ„œ μˆ˜μ§‘λœ 정보λ₯Ό μ˜λ―Έν•˜λŠ” μš©μ–΄λ‘œ μ‚¬μš©λ˜μ—ˆλ‹€κ³  ν•œλ‹€.

 

κ·Έλ ‡λ‹€λ©΄ machine learning/AI의 λ¬Έλ§₯μ—μ„œ μ‚¬μš©λ˜λŠ” "ground-truth"의 λœ»μ€ 무엇인가.

λ°˜μ‘ν˜•

What is ground truth?

Ground truth isn't true. It's an ideal expected result. It might involve hand-labeling example datapoints to collect desirable answer data for training your system.
For example, a set of images might be painstakingly hand-labeled as cat or not-cat according to the opinions of whoever was in charge of the project and those cat/not-cat labels will be called "ground truth" for the project.

cited : towardsdatascience.com/in-ai-the-objective-is-subjective-4614795d179b

μ²˜μŒμ— λ‚˜λŠ” "ground-truth"이 "label"κ³Ό 같은 의미인 것 같은데 μ™œ "label"이라고 μ•ˆν•˜κ³  "ground-truth"라고 ν•˜μ§€? μƒκ°ν–ˆμ—ˆλ‹€.γ…Ž
결둠적으둜, "ground-truth"κ³Ό "label"은 같은 의미둜 μ“°μ΄λŠ” 것이 μ•„λ‹ˆλ‹€. λͺ…ν™•ν•˜κ²Œ λ§ν•˜λ©΄ λ‘˜μ€ 같지 μ•Šλ‹€.

 

"label"은 정닡지라고도 λΆˆλ¦¬λ“―μ΄ 닡이 λͺ…ν™•ν•˜κ²Œ μ •ν•΄μ Έ μžˆλŠ” 값이닀.

 

"ground-truth"은 'μš°λ¦¬κ°€ μ •ν•œ μ •λ‹΅', '우리의 λͺ¨λΈμ΄ μš°λ¦¬κ°€ μ›ν•˜λŠ” λ‹΅μœΌλ‘œ μ˜ˆμΈ‘ν•΄μ£ΌκΈΈ λ°”λΌλŠ” λ‹΅'이닀.
예λ₯Ό λ“€μ–΄, μ•„λž˜ 사진이 λͺ¨λΈμ„ ν›ˆλ ¨μ‹œν‚¬ λ°μ΄ν„°λ‘œ μ‚¬μš©λœλ‹€κ³  ν•΄λ³΄μž. 사진은 고양이 λΆ„μž₯을 ν•œ μ‚¬λžŒμ΄λ‹€. 이 사진에 λŒ€ν•œ label, 즉 정닡은 μ—†λ‹€. μ‚¬λžŒμΈ 것도, 고양인 것도 μ•„λ‹ˆλ‹€. ν•˜μ§€λ§Œ μš°λ¦¬λŠ” 우리의 λΆ„λ₯˜λͺ¨λΈμ΄ 이λ₯Ό '고양이'둜 λΆ„λ₯˜ν•˜κΈΈ μ›ν•œλ‹€. κ·Έλ¦¬ν•˜μ—¬ 이 데이터에 λŒ€ν•œ "ground-truth" 값은 μš°λ¦¬κ°€ μ›ν•˜λŠ” 닡인 '고양이'κ°€ λœλ‹€.

좜처 : Movie 'Cats'

728x90
λ°˜μ‘ν˜•