λ¨Έμ λ¬λ/AIμμ μ¬μ©λλ "Ground Truth" λ»
λ¨Έμ λ¬λμ κ΄λ ¨ν κΈμ μ½λ€λ³΄λ©΄ "ground-truth"λΌλ μ©μ΄λ₯Ό λ§μ΄ μ νκ² λλ€.
"ground-truth"λ κΈ°μνμμ μ λνμμΌλ©°, μ΄λ ν μ₯μμμ μμ§λ μ 보λ₯Ό μλ―Ένλ μ©μ΄λ‘ μ¬μ©λμλ€κ³ νλ€.
κ·Έλ λ€λ©΄ machine learning/AIμ λ¬Έλ§₯μμ μ¬μ©λλ "ground-truth"μ λ»μ 무μμΈκ°.
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.
cited : towardsdatascience.com/in-ai-the-objective-is-subjective-4614795d179b
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.
μ²μμ λλ "ground-truth"μ΄ "label"κ³Ό κ°μ μλ―ΈμΈ κ² κ°μλ° μ "label"μ΄λΌκ³ μνκ³ "ground-truth"λΌκ³ νμ§? μκ°νμλ€.γ
κ²°λ‘ μ μΌλ‘, "ground-truth"κ³Ό "label"μ κ°μ μλ―Έλ‘ μ°μ΄λ κ²μ΄ μλλ€. λͺ
ννκ² λ§νλ©΄ λμ κ°μ§ μλ€.
"label"μ μ λ΅μ§λΌκ³ λ λΆλ¦¬λ―μ΄ λ΅μ΄ λͺ ννκ² μ ν΄μ Έ μλ κ°μ΄λ€.
"ground-truth"μ 'μ°λ¦¬κ° μ ν μ λ΅', 'μ°λ¦¬μ λͺ¨λΈμ΄ μ°λ¦¬κ° μνλ λ΅μΌλ‘ μμΈ‘ν΄μ£ΌκΈΈ λ°λΌλ λ΅'μ΄λ€.
μλ₯Ό λ€μ΄, μλ μ¬μ§μ΄ λͺ¨λΈμ νλ ¨μν¬ λ°μ΄ν°λ‘ μ¬μ©λλ€κ³ ν΄λ³΄μ. μ¬μ§μ κ³ μμ΄ λΆμ₯μ ν μ¬λμ΄λ€. μ΄ μ¬μ§μ λν label, μ¦ μ λ΅μ μλ€. μ¬λμΈ κ²λ, κ³ μμΈ κ²λ μλλ€. νμ§λ§ μ°λ¦¬λ μ°λ¦¬μ λΆλ₯λͺ¨λΈμ΄ μ΄λ₯Ό 'κ³ μμ΄'λ‘ λΆλ₯νκΈΈ μνλ€. 그리νμ¬ μ΄ λ°μ΄ν°μ λν "ground-truth" κ°μ μ°λ¦¬κ° μνλ λ΅μΈ 'κ³ μμ΄'κ° λλ€.