Add DiSH

README.md

@@ -121,7 +121,7 @@ for (let i = 0; i < n; i++) {
 
 | task | model |
 | ---- | ----- |
-| clustering | (Soft / Kernel / Genetic / Weighted / Bisecting) k-means, k-means++, k-medois, k-medians, x-means, G-means, LBG, ISODATA, Fuzzy c-means, Possibilistic c-means, k-harmonic means, MacQueen, Hartigan-Wong, Elkan, Hamelry, Drake, Yinyang, Agglomerative (complete linkage, single linkage, group average, Ward's, centroid, weighted average, median), DIANA, Monothetic, Mutual kNN, Mean shift, DBSCAN, OPTICS, DTSCAN, HDBSCAN, DENCLUE, DBCLASD, BRIDGE, CLUES, PAM, CLARA, CLARANS, BIRCH, CURE, ROCK, C2P, PLSA, Latent dirichlet allocation, GMM, VBGMM, Affinity propagation, Spectral clustering, Mountain, (Growing) SOM, GTM, (Growing) Neural gas, Growing cell structures, LVQ, ART, SVC, CAST, CHAMELEON, COLL, CLIQUE, PROCLUS, ORCLUS, FINDIT, DOC, FastDOC, NMF, Autoencoder |
+| clustering | (Soft / Kernel / Genetic / Weighted / Bisecting) k-means, k-means++, k-medois, k-medians, x-means, G-means, LBG, ISODATA, Fuzzy c-means, Possibilistic c-means, k-harmonic means, MacQueen, Hartigan-Wong, Elkan, Hamelry, Drake, Yinyang, Agglomerative (complete linkage, single linkage, group average, Ward's, centroid, weighted average, median), DIANA, Monothetic, Mutual kNN, Mean shift, DBSCAN, OPTICS, DTSCAN, HDBSCAN, DENCLUE, DBCLASD, BRIDGE, CLUES, PAM, CLARA, CLARANS, BIRCH, CURE, ROCK, C2P, PLSA, Latent dirichlet allocation, GMM, VBGMM, Affinity propagation, Spectral clustering, Mountain, (Growing) SOM, GTM, (Growing) Neural gas, Growing cell structures, LVQ, ART, SVC, CAST, CHAMELEON, COLL, CLIQUE, PROCLUS, ORCLUS, FINDIT, DOC, FastDOC, DiSH, NMF, Autoencoder |
 | classification | (Fisher's) Linear discriminant, Quadratic discriminant, Mixture discriminant, Least squares, (Multiclass / Kernel) Ridge, (Complement / Negation / Universal-set / Selective) Naive Bayes (gaussian), AODE, (Fuzzy / Weighted) k-nearest neighbor, Radius neighbor, Nearest centroid, ENN, ENaN, NNBCA, ADAMENN, DANN, IKNN, Decision tree, Random forest, Extra trees, GBDT, XGBoost, ALMA, (Aggressive) ROMMA, (Bounded) Online gradient descent, (Budgeted online) Passive aggressive, RLS, (Selective-sampling) Second order perceptron, AROW, NAROW, Confidence weighted, CELLIP, IELLIP, Normal herd, Stoptron, (Kernelized) Pegasos, MIRA, Forgetron, Projectron, Projectron++, Banditron, Ballseptron, (Multiclass) BSGD, ILK, SILK, (Multinomial) Logistic regression, (Multinomial) Probit, SVM, Gaussian process, HMM, CRF, Bayesian Network, LVQ, (Average / Multiclass / Voted / Kernelized / Selective-sampling / Margin / Shifting / Budget / Tighter / Tightest) Perceptron, PAUM, RBP, ADALINE, MADALINE, MLP, ELM, LMNN |
 | semi-supervised classification | k-nearest neighbor, Radius neighbor, Label propagation, Label spreading, k-means, GMM, S3VM, Ladder network |
 | regression | Least squares, Ridge, Lasso, Elastic net, RLS, Bayesian linear, Poisson, Least absolute deviations, Huber, Tukey, Least trimmed squares, Least median squares, Lp norm linear, SMA, Deming, Segmented, LOWESS, LOESS, spline, Naive Bayes, Gaussian process, Principal components, Partial least squares, Projection pursuit, Quantile regression, k-nearest neighbor, Radius neighbor, IDW, Nadaraya Watson, Priestley Chao, Gasser Muller, RBF Network, RVM, Decision tree, Random forest, Extra trees, GBDT, XGBoost, SVR, MLP, ELM, GMR, Isotonic, Ramer Douglas Peucker, Theil-Sen, Passing-Bablok, Repeated median |

js/model_selector.js

@@ -150,6 +150,7 @@ const AIMethods = [
 				{ value: 'orclus', title: 'ORCLUS' },
 				{ value: 'findit', title: 'FINDIT' },
 				{ value: 'doc', title: 'DOC / FastDOC' },
+				{ value: 'dish', title: 'DiSH' },
 				{ value: 'plsa', title: 'PLSA' },
 				{ value: 'latent_dirichlet_allocation', title: 'Latent Dirichlet Allocation' },
 				{ value: 'nmf', title: 'NMF' },

js/view/dish.js

@@ -0,0 +1,24 @@
+import DiSH from '../../lib/model/dish.js'
+import Controller from '../controller.js'
+
+export default function (platform) {
+	platform.setting.ml.usage = 'Click and add data point. Then, click "Fit" button.'
+	platform.setting.ml.reference = {
+		author: 'E. Achtert, C. Bohm, H. P. Kriegel, P. Kroger, I. Muller-Gorman, A. Zimek',
+		title: 'Detection and Visualization of Subspace Cluster Hierarchies',
+		year: 2007,
+	}
+	const controller = new Controller(platform)
+
+	const fitModel = () => {
+		let model = new DiSH(mu.value, e.value)
+
+		const pred = model.predict(platform.trainInput).map(v => v + 1)
+		platform.trainResult = pred
+	}
+
+	const mu = controller.input.number({ label: ' mu ', min: 1, max: 1000, value: 20 })
+	const e = controller.input.number({ label: ' e ', min: 0, max: 100, step: 0.1, value: 0.1 })
+
+	controller.input.button('Fit').on('click', fitModel)
+}

lib/model/dish.js

@@ -0,0 +1,238 @@
+class PriorityQueue {
+	constructor(arr) {
+		this._value = arr || []
+	}
+
+	get length() {
+		return this._value.length
+	}
+
+	[Symbol.iterator]() {
+		return this._value[Symbol.iterator]()
+	}
+
+	_sort() {
+		this._value.sort((a, b) => a[1] - b[1])
+	}
+
+	push(value, priority) {
+		this._value.push([value, priority])
+		this._sort()
+	}
+
+	move(value, priority) {
+		for (let i = 0; i < this.length; i++) {
+			if (this._value[i][0] === value) {
+				this._value[i][1] = priority
+				this._sort()
+				return
+			}
+		}
+		this.push(value, priority)
+	}
+
+	shift() {
+		const [value] = this._value.shift()
+		return value
+	}
+}
+
+/**
+ * Detecting Subspace cluster Hierarchies
+ */
+export default class DiSH {
+	// Detection and Visualization of Subspace Cluster Hierarchies
+	// https://imada.sdu.dk/u/zimek/publications/DASFAA2007/detection.pdf
+	/**
+	 * @param {number} mu Number of neighborhood
+	 * @param {number} e Neighborhood range
+	 */
+	constructor(mu, e) {
+		this._mu = mu
+		this._e = e
+	}
+
+	/**
+	 * Returns predicted categories.
+	 * @param {Array<Array<number>>} x Training data
+	 * @returns {number[]} Predicted values
+	 */
+	predict(x) {
+		const n = x.length
+		const a = x[0].length
+		const w = []
+		for (let i = 0; i < n; i++) {
+			const nears = []
+			for (let k = 0; k < a; k++) {
+				nears[k] = new Set()
+				for (let j = 0; j < n; j++) {
+					if (Math.abs(x[i][k] - x[j][k]) <= this._e) {
+						nears[k].add(j)
+					}
+				}
+			}
+			const c = new Set()
+			let max_n = -1
+			let max_k = -1
+			for (let k = 0; k < a; k++) {
+				if (nears[k].size >= this._mu) {
+					if (max_n < 0) {
+						max_n = nears[k].size
+						max_k = k
+					} else if (nears[k].size < max_n) {
+						c.add(max_k)
+						max_n = nears[k].size
+						max_k = k
+					} else {
+						c.add(k)
+					}
+				}
+			}
+			w[i] = Array(a).fill(0)
+			if (max_k < 0) {
+				continue
+			}
+			w[i][max_k] = 1
+
+			let int = nears[max_k]
+			while (c.length > 0) {
+				let max_n = -1
+				let max_k = -1
+				for (const k of c) {
+					const intn = new Set()
+					for (const a of int) {
+						if (nears[k].has(a)) {
+							intn.add(a)
+						}
+					}
+					nears[k] = intn
+					if (nears[k].size >= max_n) {
+						max_n = nears[k].size
+						max_k = k
+					}
+				}
+				if (max_n < this._mu) {
+					break
+				}
+				w[i][max_k] = 1
+				c.delete(max_k)
+				int = nears[max_k]
+			}
+		}
+
+		const queue = new PriorityQueue()
+		for (let i = 0; i < n; i++) {
+			queue.push(i, Infinity)
+		}
+
+		const sdists = []
+		for (let i = 0; i < n; i++) {
+			sdists[i] = []
+			sdists[i][i] = [i, 0]
+			for (let j = 0; j < i; j++) {
+				let lambda = 0
+				let sd = 0
+				for (let k = 0; k < a; k++) {
+					if (w[i][k] === 0 || w[j][k] === 0) {
+						lambda++
+					}
+					if (w[i][k] === 1 && w[j][k] === 1) {
+						sd += (x[i][k] - x[j][k]) ** 2
+					}
+				}
+				const d = lambda + (Math.sqrt(sd) > 2 * this._e ? 1 : 0)
+				sdists[i][j] = [j, d]
+				sdists[j][i] = [i, d]
+			}
+		}
+
+		const co = []
+		while (queue.length > 0) {
+			const o = queue.shift()
+			const ss = sdists[o].concat()
+			ss.sort((a, b) => a[1] - b[1])
+			const [r] = ss[this._mu]
+			for (const [p] of [...queue]) {
+				const sr = Math.max(sdists[o][r][1], sdists[o][p][1])
+				queue.move(p, sr)
+			}
+			co.push(o)
+		}
+
+		const clusters = []
+		for (let i = 0; i < n; i++) {
+			let c = null
+			for (let t = 0; t < clusters.length; t++) {
+				if (clusters[t].w.some((v, k) => v !== w[co[i]][k] * w[co[i - 1]][k])) {
+					continue
+				}
+				let dist = 0
+				for (let k = 0; k < a; k++) {
+					if (clusters[t].w[k] === 1) {
+						dist += (clusters[t].center[k] - x[co[i]][k]) ** 2
+					}
+				}
+				if (Math.sqrt(dist) > 2 * this._e) {
+					continue
+				}
+				c = t
+				break
+			}
+			if (c == null) {
+				c = clusters.length
+				clusters.push({
+					i: [],
+					center: Array(a).fill(0),
+					w: w[co[i]].concat(),
+					l: w[co[i]].reduce((s, v) => s + (v === 0 ? 1 : 0), 0),
+					parents: [],
+				})
+			}
+			clusters[c].center = clusters[c].center.map(
+				(v, k) => (v * clusters[c].i.length + x[co[i]][k]) / (clusters[c].i.length + 1)
+			)
+			clusters[c].i.push(i)
+		}
+
+		for (let i = 0; i < clusters.length; i++) {
+			for (let j = 0; j < clusters.length; j++) {
+				if (clusters[j].l <= clusters[i].l) {
+					continue
+				}
+				if (clusters[j].l === a) {
+					clusters[i].parents.push(j)
+					continue
+				}
+				let dist = 0
+				for (let k = 0; k < a; k++) {
+					if (clusters[i].w[k] === 1 && clusters[j].w[k] === 1) {
+						dist += (clusters[i].center[k] - clusters[j].center[k]) ** 2
+					}
+				}
+				if (Math.sqrt(dist) > 2 * this._e) {
+					continue
+				}
+
+				let target = true
+				for (let t = 0; t < clusters[i].parents.length; t++) {
+					if (clusters[clusters[i].parents[t]].l < clusters[j].l) {
+						target = false
+						break
+					}
+				}
+				if (target) {
+					clusters[i].parents.push(j)
+				}
+			}
+		}
+		this._clusters = clusters
+
+		const p = []
+		for (let c = 0; c < clusters.length; c++) {
+			for (let i = 0; i < clusters[c].i.length; i++) {
+				p[clusters[c].i[i]] = c
+			}
+		}
+		return p
+	}
+}

tests/gui/view/dish.test.js

@@ -0,0 +1,45 @@
+import { getPage } from '../helper/browser'
+
+describe('clustering', () => {
+	/** @type {Awaited<ReturnType<getPage>>} */
+	let page
+	beforeEach(async () => {
+		page = await getPage()
+		const taskSelectBox = await page.waitForSelector('#ml_selector dl:first-child dd:nth-child(5) select')
+		await taskSelectBox.selectOption('CT')
+		const modelSelectBox = await page.waitForSelector('#ml_selector .model_selection #mlDisp')
+		await modelSelectBox.selectOption('dish')
+	})
+
+	afterEach(async () => {
+		await page?.close()
+	})
+
+	test('initialize', async () => {
+		const methodMenu = await page.waitForSelector('#ml_selector #method_menu')
+		const buttons = await methodMenu.waitForSelector('.buttons')
+
+		const mu = await buttons.waitForSelector('input:nth-of-type(1)')
+		await expect(mu.getAttribute('value')).resolves.toBe('20')
+		const e = await buttons.waitForSelector('input:nth-of-type(2)')
+		await expect(e.getAttribute('value')).resolves.toBe('0.1')
+	})
+
+	test('learn', async () => {
+		const methodMenu = await page.waitForSelector('#ml_selector #method_menu')
+		const buttons = await methodMenu.waitForSelector('.buttons')
+
+		const fitButton = await buttons.waitForSelector('input[value=Fit]')
+		await fitButton.evaluate(el => el.click())
+
+		const svg = await page.waitForSelector('#plot-area svg')
+		await svg.waitForSelector('.datas circle')
+		const circles = await svg.$$('.datas circle')
+		const colors = new Set()
+		for (const circle of circles) {
+			const fill = await circle.evaluate(el => el.getAttribute('fill'))
+			colors.add(fill)
+		}
+		expect(colors.size).toBeGreaterThan(1)
+	})
+})

tests/lib/model/dish.test.js

@@ -0,0 +1,20 @@
+import Matrix from '../../../lib/util/matrix.js'
+import DiSH from '../../../lib/model/dish.js'
+
+import { randIndex } from '../../../lib/evaluate/clustering.js'
+
+test('clustering', () => {
+	const model = new DiSH(5, 2.0)
+	const n = 50
+	const x = Matrix.concat(Matrix.randn(n, 6, 0, 0.1), Matrix.randn(n, 6, 5, 0.1)).toArray()
+
+	const y = model.predict(x)
+	expect(y).toHaveLength(x.length)
+
+	const t = []
+	for (let i = 0; i < x.length; i++) {
+		t[i] = Math.floor(i / n)
+	}
+	const ri = randIndex(y, t)
+	expect(ri).toBeGreaterThan(0.9)
+})