이 페이지에서3분 읽기
개요
Choosing the right metric matters because different metrics assume different notions of similarity, and the wrong choice can make an otherwise sound algorithm perform poorly.
심층 분석
A distance metric turns raw feature values into a single number representing how far apart two points are, which many algorithms rely on directly. Euclidean distance, the straight-line distance familiar from geometry, is a common choice in many introductory examples and works well when features are continuous, roughly on comparable scales, and the notion of similarity matches physical closeness. Manhattan distance, also called taxicab distance, sums the absolute differences along each dimension rather than taking the square root of squared differences; it suits grid-like movement constraints and is less sensitive to outliers in individual dimensions than Euclidean distance. Cosine distance measures orientation between nonzero vectors rather than their magnitudes, making it useful in some text and recommendation systems where a document or user profile's overall direction, such as topic balance, matters more than its raw length or intensity. Mahalanobis distance generalizes Euclidean distance by accounting for the correlations and differing variances between features, effectively normalizing the space so that features are compared fairly regardless of their original scale, which is useful in anomaly detection and multivariate outlier analysis. Hamming distance counts the number of positions at which two equal-length strings or categorical vectors differ, making it suited for categorical data, error-correcting codes, and genetic sequence comparison rather than continuous numeric data. A common misconception is that Euclidean distance is always the safe default; in high-dimensional spaces, all pairwise Euclidean distances tend to become similar, a phenomenon sometimes called the curse of dimensionality, which can make nearest-neighbor comparisons less informative; checking distance distributions and task performance can reveal when that matters.
전략적 영향
더 명확한 결정들
이는 명확한 기술적 주장과 마케팅 언어를 구분하는 데 도움이 됩니다.
비용 및 예산
돈이나 시간을 들이기 전에 더 나은 구현 질문을 할 수 있습니다.
팀과 워크플로우
이해를 공유한 팀은 더 나은 제품, 정책 및 학습 결정을 내립니다.
The Future of Distance Metrics in Machine Learning
Distance metric choice remains a foundational, largely stable part of machine learning practice, though learned distance metrics, sometimes called metric learning, continue to gain traction for specialized applications like face verification, where a neural network learns an embedding space in which a simple distance, often cosine or Euclidean, becomes meaningful after training. Expect distance metrics to remain relevant even as deep learning grows, since most embedding-based systems still rely on a classical distance function applied to learned representations rather than replacing the concept of distance entirely.
실제 구현
A k-nearest neighbors model predicting house prices uses Euclidean distance across square footage, number of bedrooms, and age, treating all numeric differences as straight-line distance in feature space.
A city-grid delivery routing tool uses Manhattan distance instead of Euclidean distance, since vehicles must travel along street grids rather than in straight lines, matching the metric to the real movement constraint.
A recommendation system comparing user preference vectors uses cosine distance rather than Euclidean distance, since it cares about the direction of preference patterns, such as genre balance, rather than the raw magnitude of ratings.
A DNA sequence comparison tool uses Hamming distance to count the number of positions where two equal-length genetic sequences differ, since the data is categorical rather than continuous.
위험 및 가드레일
팀마다 동일한 용어를 다르게 사용할 수 있으므로 범위를 조기에 정의하세요.
벤치마크는 강력해 보이지만 실제 성능은 고르지 않을 수 있습니다.
데이터 품질 및 평가 계획을 무시하면 취약한 결과가 발생하는 경우가 많습니다.
구현 로드맵
필요한 결과에 대한 일반 언어 정의부터 시작하세요.
테스트하기 전에 하나의 성공 지표와 하나의 실패 조건을 선택하세요.
세련된 데모 세트가 아닌 대표 데이터를 사용하여 소규모 파일럿을 실행하세요.
Document where Distance Metrics in Machine Learning helps and where simpler methods are better.
계속 탐색하세요
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Distance Metrics in Machine Learning quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
자주 묻는 질문
What is Distance Metrics in Machine Learning?
Distance metrics are mathematical functions that quantify how similar or different two data points are, forming the basis for algorithms like k-nearest neighbors and clustering. Choosing the right metric matters because different metrics assume different notions of similarity, and the wrong choice can make an otherwise sound algorithm perform poorly.
Why did the delivery routing example choose Manhattan distance over Euclidean distance?
Manhattan distance sums differences along grid axes, matching how vehicles actually move along city streets rather than straight-line paths.
What does cosine distance measure, as defined in the guide, that makes it suited to recommendation systems?
Cosine distance captures direction, such as genre balance in a preference vector, rather than raw magnitude, which fits recommendation use cases.
What additional information does Mahalanobis distance incorporate that Euclidean distance does not?
Mahalanobis distance uses the inverse of the feature covariance matrix, accounting for correlation and scale differences that plain Euclidean distance ignores.
For what type of data is Hamming distance specifically suited, according to the guide?
The guide describes Hamming distance as counting differing positions in equal-length strings or categorical vectors, fitting genetic sequence comparison.
What preprocessing step does the guide say is typically required before computing Euclidean distance across features?
The technical section notes that features must usually be scaled, typically via standardization, so a larger-range feature doesn't dominate the distance calculation.
계속 학습하세요
관련 가이드
이 주제에 대해 선택된 추가 가이드