K-means
A classical unsupervised clustering algorithm proposed by Stuart Lloyd in 1957. It iteratively updates cluster centroids to minimize the within-cluster sum of squared errors, partitioning data into k clusters.
Timeline (last 90 days)
Choosing K=3 to summarize any image into three dominant colors reduces computation and keeps cards concise; the 4,096-pixel sampling limit is an empirical trade-off between accuracy and speed.
Palette Walk uses Python together with scikit-learn's K-means clustering to extract dominant colors from an image, and pairs each color swatch with a fixed observation prompt.
Palette Walk's K-means algorithm fits at most three clusters on up to 4,096 sampled RGB pixels to output the dominant colors of an image.
K-means is an unsupervised learning algorithm that iteratively initializes centroids, assigns data points to the nearest centroid, and recomputes the group mean as the new centroid, until the centroids no longer move significantly.
Euclidean distance in RGB space is the default similarity measure for K-means, but it does not fully correspond to the human eye's subjective perception of color differences (perceptual uniformity).
Using bilinear or bicubic interpolation during image downscaling applies a weighted average of surrounding pixels, producing intermediate colors at boundaries that do not exist in the source image and causing K-means to extract incorrect dominant colors; using nearest-neighbor sampling avoids this problem.
Traditional prototype-based learning methods such as K-means and SOM require the number of cluster centers or prototypes to be specified in advance.
Core machine learning algorithms include linear regression, logistic regression, decision trees, random forests, gradient boosting (XGBoost, LightGBM), K-nearest neighbors, and K-means clustering.
Semantic clustering is based on word embedding or sentence embedding technology, mapping keywords to points in a high-dimensional vector space and automatically grouping them using clustering algorithms such as K-means and DBSCAN.
K-means can be viewed as the hard-limit case of the EM algorithm on Gaussian mixture models, where the EM algorithm uses soft assignments while K-means uses hard assignments
1 more timeline events
All Facts (11)
Choosing K=3 to summarize any image into three dominant colors reduces computation and keeps cards concise; the 4,096-pixel sampling limit is an empirical trade-off between accuracy and speed.
50%UnverifiedPalette Walk uses Python together with scikit-learn's K-means clustering to extract dominant colors from an image, and pairs each color swatch with a fixed observation prompt.
50%UnverifiedPalette Walk's K-means algorithm fits at most three clusters on up to 4,096 sampled RGB pixels to output the dominant colors of an image.
50%UnverifiedK-means is an unsupervised learning algorithm that iteratively initializes centroids, assigns data points to the nearest centroid, and recomputes the group mean as the new centroid, until the centroids no longer move significantly.
50%UnverifiedEuclidean distance in RGB space is the default similarity measure for K-means, but it does not fully correspond to the human eye's subjective perception of color differences (perceptual uniformity).
50%UnverifiedUsing bilinear or bicubic interpolation during image downscaling applies a weighted average of surrounding pixels, producing intermediate colors at boundaries that do not exist in the source image and causing K-means to extract incorrect dominant colors; using nearest-neighbor sampling avoids this problem.
50%UnverifiedTraditional prototype-based learning methods such as K-means and SOM require the number of cluster centers or prototypes to be specified in advance.
50%UnverifiedCore machine learning algorithms include linear regression, logistic regression, decision trees, random forests, gradient boosting (XGBoost, LightGBM), K-nearest neighbors, and K-means clustering.
50%UnverifiedSemantic clustering is based on word embedding or sentence embedding technology, mapping keywords to points in a high-dimensional vector space and automatically grouping them using clustering algorithms such as K-means and DBSCAN.
50%UnverifiedK-means can be viewed as the hard-limit case of the EM algorithm on Gaussian mixture models, where the EM algorithm uses soft assignments while K-means uses hard assignments
50%UnverifiedThe K-means clustering algorithm is essentially a specific instance of coordinate descent
50%