@thi.ng/k-means 0.5.39 → 0.6.0

package/CHANGELOG.md

@@ -1,6 +1,6 @@
 # Change Log
 
-- **Last updated**: 2023-01-10T15:20:19Z
+- **Last updated**: 2023-02-10T13:55:29Z
 - **Generator**: [thi.ng/monopub](https://thi.ng/monopub)
 
 All notable changes to this project will be documented in this file.
@@ -9,6 +9,15 @@ See [Conventional Commits](https://conventionalcommits.org/) for commit guidelin
 **Note:** Unlisted _patch_ versions only involve non-code or otherwise excluded changes
 and/or version bumps of transitive dependencies.
 
+## [0.6.0](https://github.com/thi-ng/umbrella/tree/@thi.ng/k-means@0.6.0) (2023-02-10)
+
+#### 🚀 Features
+
+- filter result clusters, minor optimizations ([78169e6](https://github.com/thi-ng/umbrella/commit/78169e6))
+  - use u32 array for internal cluster assignments
+  - remove attempt to start new cluster if one became unused
+  - update buildClusters to filter out empty
+
 ## [0.5.0](https://github.com/thi-ng/umbrella/tree/@thi.ng/k-means@0.5.0) (2021-11-17)
 
 #### 🚀 Features

package/README.md

@@ -48,7 +48,7 @@ For Node.js REPL:
 const kMeans = await import("@thi.ng/k-means");
 ```
 
-Package sizes (brotli'd, pre-treeshake): ESM: 906 bytes
+Package sizes (brotli'd, pre-treeshake): ESM: 878 bytes
 
 ## Dependencies
 
@@ -82,7 +82,7 @@ import { kmeans, meansLatLon } from "@thi.ng/k-means";
 import { HAVERSINE_LATLON } from "@thi.ng/distance";
 
 // data from: https://simplemaps.com/data/world-cities
-const items = [
+const cities = [
     { id: "berlin", latlon: [52.5167, 13.3833] },
     { id: "boston", latlon: [42.3188, -71.0846] },
     { id: "detroit", latlon: [42.3834, -83.1024] },
@@ -99,7 +99,7 @@ const items = [
 // cluster based on lat/lon
 const clusters = kmeans(
     3,
-    items.map((x) => x.latlon),
+    cities.map((x) => x.latlon),
     {
         // custom centroid calc for geo locations
         // https://docs.thi.ng/umbrella/k-means/modules.html#meansLatLon
@@ -111,7 +111,7 @@ const clusters = kmeans(
 
 // print each cluster
 for (let c of clusters) {
-    console.log(c.items.map((i) => items[i].id));
+    console.log(c.items.map((i) => cities[i].id));
 }
 
 // [ 'boston', 'detroit', 'new york', 'philadelphia' ]

package/kmeans.d.ts

@@ -3,7 +3,8 @@ import type { ReadonlyVec } from "@thi.ng/vectors";
 import type { CentroidStrategy, Cluster, KMeansOpts } from "./api.js";
 /**
  * Takes an array of n-dimensional `samples` and attempts to assign them to up
- * to `k` clusters, using the behavior defined by (optionally) given `opts`.
+ * to `k` clusters (might produce less), using the behavior defined by
+ * (optionally) given `opts`.
  *
  * @remarks
  * https://en.wikipedia.org/wiki/K-medians_clustering

package/kmeans.js

@@ -2,7 +2,6 @@ import { argmin } from "@thi.ng/distance/argmin";
 import { DIST_SQ } from "@thi.ng/distance/squared";
 import { assert } from "@thi.ng/errors/assert";
 import { SYSTEM } from "@thi.ng/random/system";
-import { uniqueIndices } from "@thi.ng/random/unique-indices";
 import { weightedRandom } from "@thi.ng/random/weighted-random";
 import { add } from "@thi.ng/vectors/add";
 import { median } from "@thi.ng/vectors/median";
@@ -10,7 +9,8 @@ import { mulN } from "@thi.ng/vectors/muln";
 import { zeroes } from "@thi.ng/vectors/setn";
 /**
  * Takes an array of n-dimensional `samples` and attempts to assign them to up
- * to `k` clusters, using the behavior defined by (optionally) given `opts`.
+ * to `k` clusters (might produce less), using the behavior defined by
+ * (optionally) given `opts`.
  *
  * @remarks
  * https://en.wikipedia.org/wiki/K-medians_clustering
@@ -32,27 +32,20 @@ export const kmeans = (k, samples, opts) => {
     assert(centroidIDs.length > 0, `missing initial centroids`);
     k = centroidIDs.length;
     const centroids = centroidIDs.map((i) => samples[i]);
-    const clusters = [];
+    const clusters = new Uint32Array(num).fill(k);
     let update = true;
-    outer: while (update && maxIter-- > 0) {
+    while (update && maxIter-- > 0) {
         update = assign(samples, centroids, clusters, dist);
+        if (!update)
+            break;
         for (let i = 0; i < k; i++) {
             const impl = strategy(dim);
             for (let j = 0; j < num; j++) {
                 i === clusters[j] && impl.update(samples[j]);
             }
             const centroid = impl.finish();
-            if (centroid) {
+            if (centroid)
                 centroids[i] = centroid;
-            }
-            else {
-                const numC = centroidIDs.length;
-                uniqueIndices(1, num, centroidIDs, undefined, rnd);
-                if (centroidIDs.length === numC)
-                    break outer;
-                centroids[i] = samples[centroidIDs[numC]];
-                update = true;
-            }
         }
     }
     return buildClusters(centroids, clusters);
@@ -104,25 +97,29 @@ export const initKmeanspp = (k, samples, dist = DIST_SQ, rnd = SYSTEM) => {
     }
     return centroidIDs;
 };
-const assign = (samples, centroids, clusters, dist) => {
+const assign = (samples, centroids, assignments, dist) => {
     let update = false;
-    for (let i = 0, n = samples.length; i < n; i++) {
+    for (let i = samples.length; i-- > 0;) {
         const id = argmin(samples[i], centroids, dist);
-        if (id !== clusters[i]) {
-            clusters[i] = id;
+        if (id !== assignments[i]) {
+            assignments[i] = id;
             update = true;
         }
     }
     return update;
 };
-const buildClusters = (centroids, clusters) => {
-    const indices = [];
-    for (let i = 0, n = clusters.length; i < n; i++) {
-        const c = clusters[i];
-        (indices[c] ||
-            (indices[c] = { id: c, centroid: centroids[c], items: [] })).items.push(i);
+const buildClusters = (centroids, assignments) => {
+    const clusters = [];
+    for (let i = 0, n = assignments.length; i < n; i++) {
+        const id = assignments[i];
+        (clusters[id] ||
+            (clusters[id] = {
+                id,
+                centroid: centroids[id],
+                items: [],
+            })).items.push(i);
     }
-    return indices;
+    return clusters.filter((x) => !!x);
 };
 /**
  * Default centroid strategy forming new centroids by averaging the position of

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@thi.ng/k-means",
-  "version": "0.5.39",
+  "version": "0.6.0",
   "description": "Configurable k-means & k-medians (with k-means++ initialization) for n-D vectors",
   "type": "module",
   "module": "./index.js",
@@ -34,19 +34,19 @@
     "test": "testament test"
   },
   "dependencies": {
-    "@thi.ng/api": "^8.6.3",
-    "@thi.ng/distance": "^2.1.39",
-    "@thi.ng/errors": "^2.2.8",
-    "@thi.ng/random": "^3.3.21",
-    "@thi.ng/vectors": "^7.5.31"
+    "@thi.ng/api": "^8.7.1",
+    "@thi.ng/distance": "^2.2.1",
+    "@thi.ng/errors": "^2.2.10",
+    "@thi.ng/random": "^3.3.23",
+    "@thi.ng/vectors": "^7.6.1"
   },
   "devDependencies": {
-    "@microsoft/api-extractor": "^7.33.7",
-    "@thi.ng/testament": "^0.3.9",
-    "rimraf": "^3.0.2",
+    "@microsoft/api-extractor": "^7.34.2",
+    "@thi.ng/testament": "^0.3.10",
+    "rimraf": "^4.1.2",
     "tools": "^0.0.1",
-    "typedoc": "^0.23.22",
-    "typescript": "^4.9.4"
+    "typedoc": "^0.23.24",
+    "typescript": "^4.9.5"
   },
   "keywords": [
     "cluster",
@@ -82,5 +82,5 @@
     "status": "beta",
     "year": 2021
   },
-  "gitHead": "3f0b3e2a7c82aefc7e46fb4338369836b5e1b8cf\n"
+  "gitHead": "2b5a99a8af71670780875637299be9118b01084d\n"
 }