Add some tests

Author: ishii-norimi

tests/lib/model/c2p.test.js

@@ -44,4 +44,21 @@ describe('clustering', () => {
 		const ri = randIndex(y, t)
 		expect(ri).toBeGreaterThan(0.9)
 	})
+
+	test('near distance clusters', () => {
+		const model = new C2P(10, 50)
+		const x = [
+			[0, 0],
+			[1, 1],
+			[0, 1],
+		]
+
+		model.fit(x)
+		const y = model.predict(3)
+		expect(y).toHaveLength(x.length)
+
+		const t = [0, 1, 2]
+		const ri = randIndex(y, t)
+		expect(ri).toBeGreaterThan(0.9)
+	})
 })

tests/lib/model/generalized_esd.test.js

@@ -1,20 +1,36 @@
 import Matrix from '../../../lib/util/matrix.js'
 import GeneralizedESD from '../../../lib/model/generalized_esd.js'
 
-test.each([1, 2, 3])('anomaly detection %d', k => {
-	const model = new GeneralizedESD(1, k)
-	const x = Matrix.random(100, 2, 0, 0.2).toArray()
-	x.push([10, 10])
-	const y = model.predict(x, 0.1)
-	let c = 0
-	for (let i = 0; i < y.length - 1; i++) {
-		if (c < k - 1) {
-			if (y[i]) {
-				c++
+describe('anomaly detection', () => {
+	test.each([1, 2, 3])('r: %d', k => {
+		const model = new GeneralizedESD(1, k)
+		const x = Matrix.random(100, 2, 0, 0.2).toArray()
+		x.push([10, 10])
+		const y = model.predict(x, 0.1)
+		let c = 0
+		for (let i = 0; i < y.length - 1; i++) {
+			if (c < k - 1) {
+				if (y[i]) {
+					c++
+				}
+				continue
 			}
-			continue
+			expect(y[i]).toBe(false)
 		}
-		expect(y[i]).toBe(false)
-	}
-	expect(y[y.length - 1]).toBe(true)
+		expect(y[y.length - 1]).toBe(true)
+	})
+
+	test('oidx correction', () => {
+		const model = new GeneralizedESD(1, 3)
+		const x = [
+			[0.06, 0.13],
+			[0.01, 0.06],
+			[0.1, 0],
+			[0.07, 0.01],
+			[0.12, 0.15],
+			[10, 10],
+		]
+		const y = model.predict(x, 0.1)
+		expect(y[5]).toBe(true)
+	})
 })

tests/lib/model/vbgmm.test.js

@@ -6,28 +6,55 @@ import VBGMM from '../../../lib/model/vbgmm.js'
 
 import { randIndex } from '../../../lib/evaluate/clustering.js'
 
-test('clustering', () => {
-	const model = new VBGMM(0.001, 0.001, 5)
-	const n = 20
-	const x = Matrix.concat(Matrix.randn(n, 2, 0, 0.2), Matrix.randn(n, 2, 5, 0.2)).toArray()
+describe('clustering', () => {
+	test('many data', () => {
+		const model = new VBGMM(0.001, 0.001, 5)
+		const n = 20
+		const x = Matrix.concat(Matrix.randn(n, 2, 0, 0.2), Matrix.randn(n, 2, 5, 0.2)).toArray()
 
-	model.init(x)
-	for (let i = 0; i < 2; i++) {
+		model.init(x)
+		for (let i = 0; i < 2; i++) {
+			model.fit()
+		}
+		expect(model.means.sizes).toEqual([5, 2])
+		expect(model.covs).toHaveLength(5)
+		for (let i = 0; i < model.covs.length; i++) {
+			expect(model.covs[i].sizes).toEqual([2, 2])
+		}
+		expect(model.effectivity).toHaveLength(5)
+		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)
+	})
+
+	test('few data', () => {
+		const model = new VBGMM(0.001, 0.001, 1)
+		const x = [
+			[0, 0],
+			[1, 1],
+		]
+
+		model.init(x)
+		model.fit()
+
+		const y = model.predict(x)
+		expect(y).toEqual([0, 0])
+	})
+
+	test('single data', () => {
+		const model = new VBGMM(0.001, 0.001, 1)
+		const x = [[0, 0]]
+
+		model.init(x)
 		model.fit()
-	}
-	expect(model.means.sizes).toEqual([5, 2])
-	expect(model.covs).toHaveLength(5)
-	for (let i = 0; i < model.covs.length; i++) {
-		expect(model.covs[i].sizes).toEqual([2, 2])
-	}
-	expect(model.effectivity).toHaveLength(5)
-	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)
+
+		const y = model.predict(x)
+		expect(y).toEqual([0])
+	})
 })