Refactor and add tests

js/view/gan.js

@@ -123,7 +123,14 @@ export default function (platform) {
 		{ name: 'dis_rate', title: 'D', value: 0.5 },
 	]) {
 		const grd = ganRatesDiv.div()
-		rateElms[v.name] = grd.input.number({ label: v.title, name: v.name, min: 0, max: 100, step: 0.01, value: v.value })
+		rateElms[v.name] = grd.input.number({
+			label: v.title,
+			name: v.name,
+			min: 0,
+			max: 100,
+			step: 0.01,
+			value: v.value,
+		})
 	}
 	const batch = controller.input.number({ label: ' Batch size ', min: 1, max: 100, value: 10 })
 	let threshold = null

tests/lib/model/categorical_naive_bayes.test.js

@@ -32,3 +32,32 @@ test('predict', () => {
 	const acc = accuracy(y, t)
 	expect(acc).toBeGreaterThan(0.95)
 })
+
+test('predict fit twice', () => {
+	const model = new CategoricalNaiveBayes()
+	const x = [['a']]
+	const t = []
+	for (let i = 0; i < x.length; i++) {
+		t[i] = 'x'
+	}
+
+	model.fit(x, t)
+	model.fit(x, t)
+
+	const y = model.predict([['a']])
+	expect(y).toEqual(['x'])
+})
+
+test('predict unknown data label', () => {
+	const model = new CategoricalNaiveBayes()
+	const x = [['a']]
+	const t = []
+	for (let i = 0; i < x.length; i++) {
+		t[i] = 'x'
+	}
+
+	model.fit(x, t)
+
+	const y = model.predict([['b']])
+	expect(y).toEqual([null])
+})

tests/lib/model/catmull_rom.test.js

@@ -27,23 +27,35 @@ test('CatmullRomSplines', () => {
 	expect(err).toBeLessThan(0.1)
 })
 
-test('CentripetalCatmullRomSplines', () => {
-	const model = new CentripetalCatmullRomSplines()
-	const x = Matrix.random(20, 1, -2, 2).value
-	const t = []
-	for (let i = 0; i < x.length; i++) {
-		t[i] = Math.sin(x[i])
-	}
-	model.fit(x, t)
-
-	const y = model.predict(x)
-	expect(y).toHaveLength(x.length)
-	for (let i = 0; i < y.length; i++) {
-		expect(y[i]).toBeCloseTo(t[i])
-	}
-
-	const x0 = Matrix.random(100, 1, -2, 2).value
-	const y0 = model.predict(x0)
-	const err = rmse(y0, x0.map(Math.sin))
-	expect(err).toBeLessThan(0.1)
+describe('CentripetalCatmullRomSplines', () => {
+	test('default', () => {
+		const model = new CentripetalCatmullRomSplines()
+		const x = Matrix.random(20, 1, -2, 2).value
+		const t = []
+		for (let i = 0; i < x.length; i++) {
+			t[i] = Math.sin(x[i])
+		}
+		model.fit(x, t)
+
+		const y = model.predict(x)
+		expect(y).toHaveLength(x.length)
+		for (let i = 0; i < y.length; i++) {
+			expect(y[i]).toBeCloseTo(t[i])
+		}
+
+		const x0 = Matrix.random(100, 1, -2, 2).value
+		const y0 = model.predict(x0)
+		const err = rmse(y0, x0.map(Math.sin))
+		expect(err).toBeLessThan(0.1)
+	})
+
+	test('predict second largest value', () => {
+		const model = new CentripetalCatmullRomSplines()
+		const x = [0, 2]
+		const t = [0, 1]
+		model.fit(x, t)
+
+		const y = model.predict([1])
+		expect(y[0]).toBeCloseTo(0.5)
+	})
 })

tests/lib/model/change_finder.test.js

@@ -1,11 +1,22 @@
 import Matrix from '../../../lib/util/matrix.js'
 import ChangeFinder from '../../../lib/model/change_finder.js'
 
-test('anomaly detection', () => {
-	const model = new ChangeFinder(5)
-	const n = 50
-	const x = Matrix.concat(Matrix.random(n, 1, 0, 1), Matrix.random(n, 1, 10, 11)).value
-	model.fit(x)
-	const p = model.predict()
-	expect(p).toHaveLength(100)
+describe('anomaly detection', () => {
+	test('default', () => {
+		const model = new ChangeFinder()
+		const n = 50
+		const x = Matrix.concat(Matrix.random(n, 1, 0, 1), Matrix.random(n, 1, 10, 11)).value
+		model.fit(x)
+		const p = model.predict()
+		expect(p).toHaveLength(100)
+	})
+
+	test('with params', () => {
+		const model = new ChangeFinder(5)
+		const n = 50
+		const x = Matrix.concat(Matrix.random(n, 1, 0, 1), Matrix.random(n, 1, 10, 11)).value
+		model.fit(x)
+		const p = model.predict()
+		expect(p).toHaveLength(100)
+	})
 })

tests/lib/model/co_training.test.js

@@ -77,3 +77,33 @@ test('semi-classifier', () => {
 	const acc = accuracy(y, t_org)
 	expect(acc).toBeGreaterThan(0.95)
 })
+
+test('semi-classifier not categorized', () => {
+	const model = new CoTraining(
+		{
+			fit(x, y) {},
+			predict(x) {
+				return x.map(v => ({ category: 0, score: 0 }))
+			},
+			threshold: 0.5,
+		},
+		{
+			fit(x, y) {},
+			predict(x) {
+				return x.map(() => ({ category: 0, score: 0 }))
+			},
+			threshold: 0.5,
+		}
+	)
+	const x = [
+		[0, 0],
+		[1, 1],
+	]
+	const t = [0, null]
+	model.init(x, t)
+	for (let i = 0; i < 1; i++) {
+		model.fit()
+	}
+	const y = model.predict(x)
+	expect(y).toEqual([0, null])
+})

tests/lib/model/confidence_weighted.test.js

@@ -6,18 +6,34 @@ import { ConfidenceWeighted, SoftConfidenceWeighted } from '../../../lib/model/c
 
 import { accuracy } from '../../../lib/evaluate/classification.js'
 
-test('ConfidenceWeighted', () => {
-	const model = new ConfidenceWeighted(0.9)
-	const x = Matrix.concat(Matrix.randn(50, 2, 0, 0.2), Matrix.randn(50, 2, 5, 0.2)).toArray()
-	const t = []
-	for (let i = 0; i < x.length; i++) {
-		t[i] = Math.floor(i / 50) * 2 - 1
-	}
-	model.init(x, t)
-	model.fit()
-	const y = model.predict(x)
-	const acc = accuracy(y, t)
-	expect(acc).toBeGreaterThan(0.95)
+describe('ConfidenceWeighted', () => {
+	test('normal eta', () => {
+		const model = new ConfidenceWeighted(0.9)
+		const x = Matrix.concat(Matrix.randn(50, 2, 0, 0.2), Matrix.randn(50, 2, 5, 0.2)).toArray()
+		const t = []
+		for (let i = 0; i < x.length; i++) {
+			t[i] = Math.floor(i / 50) * 2 - 1
+		}
+		model.init(x, t)
+		model.fit()
+		const y = model.predict(x)
+		const acc = accuracy(y, t)
+		expect(acc).toBeGreaterThan(0.95)
+	})
+
+	test.each([1, 0.5, 0])('eta %p', eta => {
+		const model = new ConfidenceWeighted(eta)
+		const x = Matrix.concat(Matrix.randn(50, 2, 0, 0.2), Matrix.randn(50, 2, 5, 0.2)).toArray()
+		const t = []
+		for (let i = 0; i < x.length; i++) {
+			t[i] = Math.floor(i / 50) * 2 - 1
+		}
+		model.init(x, t)
+		model.fit()
+		const y = model.predict(x)
+		const acc = accuracy(y, t)
+		expect(acc).toBeCloseTo(0.5)
+	})
 })
 
 test.each([1, 2])('SoftConfidenceWeighted %d', version => {

tests/lib/model/crf.test.js

@@ -20,3 +20,21 @@ test('fit', () => {
 	expect(prob).toBeGreaterThan(0.4)
 	expect(prob).toBeLessThanOrEqual(1)
 })
+
+test('fit unknown predict label', () => {
+	const model = new CRF()
+	const x = [['a', 'b', 'c']]
+	const y = [[2, 0, 1]]
+
+	for (let i = 0; i < 100; i++) {
+		model.fit(x, y)
+	}
+
+	const tx = [['a', 'd', 'c']]
+	const ty = [[2, 0, 1]]
+	const p = model.predict(tx)
+	expect(p).toEqual(ty)
+	const prob = model.probability(tx[0], ty[0])
+	expect(prob).toBeGreaterThan(0.4)
+	expect(prob).toBeLessThanOrEqual(1)
+})

tests/lib/model/cubic_hermite_spline.test.js

@@ -3,23 +3,35 @@ import CubicHermiteSpline from '../../../lib/model/cubic_hermite_spline.js'
 
 import { rmse } from '../../../lib/evaluate/regression.js'
 
-test('interpolation', () => {
-	const model = new CubicHermiteSpline(0, 0)
-	const x = Matrix.random(50, 1, -2, 2).value
-	const t = []
-	for (let i = 0; i < x.length; i++) {
-		t[i] = Math.sin(x[i])
-	}
-	model.fit(x, t)
+describe('interpolation', () => {
+	test('default', () => {
+		const model = new CubicHermiteSpline(0, 0)
+		const x = Matrix.random(50, 1, -2, 2).value
+		const t = []
+		for (let i = 0; i < x.length; i++) {
+			t[i] = Math.sin(x[i])
+		}
+		model.fit(x, t)
 
-	const y = model.predict(x)
-	expect(y).toHaveLength(x.length)
-	for (let i = 0; i < y.length; i++) {
-		expect(y[i]).toBeCloseTo(t[i])
-	}
+		const y = model.predict(x)
+		expect(y).toHaveLength(x.length)
+		for (let i = 0; i < y.length; i++) {
+			expect(y[i]).toBeCloseTo(t[i])
+		}
 
-	const x0 = Matrix.random(100, 1, -2, 2).value
-	const y0 = model.predict(x0)
-	const err = rmse(y0, x0.map(Math.sin))
-	expect(err).toBeLessThan(0.1)
+		const x0 = Matrix.random(100, 1, -2, 2).value
+		const y0 = model.predict(x0)
+		const err = rmse(y0, x0.map(Math.sin))
+		expect(err).toBeLessThan(0.1)
+	})
+
+	test('predict second largest value', () => {
+		const model = new CubicHermiteSpline(0, 0)
+		const x = [0, 2]
+		const t = [0, 1]
+		model.fit(x, t)
+
+		const y = model.predict([1])
+		expect(y[0]).toBeCloseTo(0.5)
+	})
 })

tests/lib/model/cubic_interpolation.test.js

@@ -3,23 +3,35 @@ import CubicInterpolation from '../../../lib/model/cubic_interpolation.js'
 
 import { rmse } from '../../../lib/evaluate/regression.js'
 
-test('interpolation', () => {
-	const model = new CubicInterpolation()
-	const x = Matrix.random(50, 1, -2, 2).value
-	const t = []
-	for (let i = 0; i < x.length; i++) {
-		t[i] = Math.sin(x[i])
-	}
-	model.fit(x, t)
+describe('interpolation', () => {
+	test('default', () => {
+		const model = new CubicInterpolation()
+		const x = Matrix.random(50, 1, -2, 2).value
+		const t = []
+		for (let i = 0; i < x.length; i++) {
+			t[i] = Math.sin(x[i])
+		}
+		model.fit(x, t)
 
-	const y = model.predict(x)
-	expect(y).toHaveLength(x.length)
-	for (let i = 0; i < y.length; i++) {
-		expect(y[i]).toBeCloseTo(t[i])
-	}
+		const y = model.predict(x)
+		expect(y).toHaveLength(x.length)
+		for (let i = 0; i < y.length; i++) {
+			expect(y[i]).toBeCloseTo(t[i])
+		}
 
-	const x0 = Matrix.random(100, 1, -2, 2).value
-	const y0 = model.predict(x0)
-	const err = rmse(y0, x0.map(Math.sin))
-	expect(err).toBeLessThan(0.1)
+		const x0 = Matrix.random(100, 1, -2, 2).value
+		const y0 = model.predict(x0)
+		const err = rmse(y0, x0.map(Math.sin))
+		expect(err).toBeLessThan(0.1)
+	})
+
+	test('predict second largest value', () => {
+		const model = new CubicInterpolation()
+		const x = [0, 2]
+		const t = [0, 1]
+		model.fit(x, t)
+
+		const y = model.predict([1])
+		expect(y[0]).toBeCloseTo(0.5)
+	})
 })

tests/lib/model/dann.test.js

@@ -33,4 +33,17 @@ describe('classification', () => {
 		const acc = accuracy(y, t)
 		expect(acc).toBeGreaterThan(0.95)
 	})
+
+	test('same number of class choice', () => {
+		const model = new DiscriminantAdaptiveNearestNeighbor(2)
+		const x = [
+			[-1, -1],
+			[1, 1],
+		]
+		const t = ['a', 'b']
+
+		model.fit(x, t)
+		const y = model.predict([[-1, -1]])
+		expect(y).toEqual(['a'])
+	})
 })

tests/lib/model/denclue.test.js

@@ -78,4 +78,20 @@ describe('clustering', () => {
 		const ri = randIndex(y, t)
 		expect(ri).toBeGreaterThan(0.9)
 	})
+
+	test('large h', () => {
+		const model = new DENCLUE(1e6)
+		const n = 50
+		const x = [
+			[0, 0],
+			[1, 1],
+		]
+
+		model.init(x)
+		for (let i = 0; i < 10; i++) {
+			model.fit()
+		}
+		const y = model.predict()
+		expect(y).toEqual([-1, -1])
+	})
 })