Accept object for kernel argument

js/view/gplvm.js

@@ -1,82 +1,32 @@
 import GPLVM from '../../lib/model/gplvm.js'
 import Controller from '../controller.js'
 
-var dispGPLVM = function (elm, platform) {
+export default function (platform) {
+	platform.setting.ml.usage = 'Click and add data point. Next, click "Fit" button.'
 	const controller = new Controller(platform)
 	let model = null
 	const fitModel = () => {
 		if (!model) {
 			const dim = platform.dimension
-			const alpha = +elm.select('[name=alpha]').property('value')
-			const sigma = +elm.select('[name=sigma]').property('value')
-			const ez = +elm.select('[name=ez]').property('value')
-			const ea = +elm.select('[name=ea]').property('value')
-			const ep = +elm.select('[name=ep]').property('value')
-			model = new GPLVM(dim, alpha, ez, ea, ep, 'gaussian', [1.0, sigma])
+			model = new GPLVM(dim, alpha.value, ez.value, ea.value, ep.value, {
+				name: kernel.value,
+				a: 1.0,
+				b: gauss_sigma.avlue,
+			})
 			model.init(platform.trainInput)
 		}
 		model.fit()
 		const y = model.predict(platform.trainInput)
 		platform.trainResult = y
 	}
 
-	const kernelElm = elm.append('span')
-	kernelElm
-		.append('select')
-		.attr('name', 'kernel')
-		.selectAll('option')
-		.data(['gaussian'])
-		.enter()
-		.append('option')
-		.attr('value', d => d)
-		.text(d => d)
-	const gauss_sigma = kernelElm.append('span')
-	gauss_sigma
-		.append('span')
-		.text(' sigma = ')
-		.append('input')
-		.attr('type', 'number')
-		.attr('name', 'sigma')
-		.attr('value', 1)
-		.attr('min', 0)
-		.attr('max', 10)
-		.attr('step', 0.1)
-	elm.append('span')
-		.text(' alpha = ')
-		.append('input')
-		.attr('type', 'number')
-		.attr('name', 'alpha')
-		.attr('value', 0.05)
-		.attr('min', 0)
-		.attr('max', 10)
-		.attr('step', 0.01)
-	elm.append('span')
-		.text(' ez = ')
-		.append('input')
-		.attr('type', 'number')
-		.attr('name', 'ez')
-		.attr('value', 1)
-		.attr('min', 0)
-		.attr('max', 10)
-		.attr('step', 0.1)
-	elm.append('span')
-		.text(' ea = ')
-		.append('input')
-		.attr('type', 'number')
-		.attr('name', 'ea')
-		.attr('value', 0.005)
-		.attr('min', 0)
-		.attr('max', 10)
-		.attr('step', 0.001)
-	elm.append('span')
-		.text(' ep = ')
-		.append('input')
-		.attr('type', 'number')
-		.attr('name', 'ep')
-		.attr('value', 0.02)
-		.attr('min', 0)
-		.attr('max', 10)
-		.attr('step', 0.001)
+	const kernelElm = controller.span()
+	const kernel = kernelElm.select(['gaussian'])
+	const gauss_sigma = kernelElm.input.number({ label: ' sigma = ', min: 0, max: 10, step: 0.1, value: 1 })
+	const alpha = controller.input.number({ label: ' alpha = ', min: 0, max: 10, step: 0.01, value: 0.05 })
+	const ez = controller.input.number({ label: ' ez = ', min: 0, max: 10, step: 0.1, value: 1 })
+	const ea = controller.input.number({ label: ' ea = ', min: 0, max: 10, step: 0.001, value: 0.005 })
+	const ep = controller.input.number({ label: ' ep = ', min: 0, max: 10, step: 0.001, value: 0.02 })
 	controller
 		.stepLoopButtons()
 		.init(() => {
@@ -86,8 +36,3 @@ var dispGPLVM = function (elm, platform) {
 		.step(fitModel)
 		.epoch()
 }
-
-export default function (platform) {
-	platform.setting.ml.usage = 'Click and add data point. Next, click "Fit" button.'
-	dispGPLVM(platform.setting.ml.configElement, platform)
-}

js/view/ocsvm.js

@@ -50,11 +50,7 @@ export default function (platform) {
 		step: 0.01,
 	})
 	const slbConf = controller.stepLoopButtons().init(() => {
-		const args = []
-		if (kernel.value === 'gaussian') {
-			args.push(gamma.value)
-		}
-		model = new OCSVM(nu.value, kernel.value, args)
+		model = new OCSVM(nu.value, { name: kernel.value, d: gamma.value })
 		model.init(platform.trainInput, platform.trainOutput)
 		learn_epoch = 0
 		platform.init()

js/view/pca.js

@@ -13,13 +13,7 @@ export default function (platform) {
 			} else if (type.value === 'dual') {
 				model = new DualPCA()
 			} else {
-				const args = []
-				if (kernel.value === 'polynomial') {
-					args.push(poly_d.value)
-				} else if (kernel.value === 'gaussian') {
-					args.push(sigma.value)
-				}
-				model = new KernelPCA(kernel.value, args)
+				model = new KernelPCA({ name: kernel.value, sigma: sigma.value, n: poly_d.value })
 			}
 			model.fit(platform.trainInput)
 			const y = model.predict(platform.trainInput, dim)

js/view/s3vm.js

@@ -34,11 +34,7 @@ export default function (platform) {
 		step: 0.01,
 	})
 	const slbConf = controller.stepLoopButtons().init(() => {
-		const kernel_args = []
-		if (kernel.value === 'gaussian') {
-			kernel_args.push(gamma.value)
-		}
-		model = new S3VM(kernel.value, kernel_args)
+		model = new S3VM({ name: kernel.value, d: gamma.value })
 		model.init(
 			platform.trainInput,
 			platform.trainOutput.map(v => (v[0] == null ? null : v[0] === 1 ? -1 : 1))

js/view/svc.js

@@ -35,11 +35,7 @@ export default function (platform) {
 	controller
 		.stepLoopButtons()
 		.init(() => {
-			const kernel_args = []
-			if (kernel.value === 'gaussian') {
-				kernel_args.push(gamma.value)
-			}
-			model = new SVC(kernel.value, kernel_args)
+			model = new SVC({ name: kernel.value, d: gamma.value })
 			model.init(platform.trainInput)
 			platform.init()
 		})

js/view/svm.js

@@ -41,12 +41,8 @@ export default function (platform) {
 		step: 0.01,
 	})
 	const slbConf = controller.stepLoopButtons().init(() => {
-		const kernel_args = []
-		if (kernel.value === 'gaussian') {
-			kernel_args.push(gamma.value)
-		}
 		model = new EnsembleBinaryModel(function () {
-			return new SVM(kernel.value, kernel_args)
+			return new SVM({ name: kernel.value, d: gamma.value })
 		}, method.value)
 		model.init(
 			platform.trainInput,

js/view/svr.js

@@ -31,11 +31,7 @@ export default function (platform) {
 		step: 0.1,
 	})
 	const slbConf = controller.stepLoopButtons().init(() => {
-		const args = []
-		if (kernel.value === 'gaussian') {
-			args.push(gamma.value)
-		}
-		model = new SVR(kernel.value, args)
+		model = new SVR({ name: kernel.value, d: gamma.value })
 		model.init(platform.trainInput, platform.trainOutput)
 		learn_epoch = 0
 		platform.init()

lib/model/bogd.js

@@ -10,7 +10,7 @@ export default class BOGD {
 	 * @param {number} [lambda=0.1] Regularization parameter
 	 * @param {number} [gamma=0.1] Maximum coefficient
 	 * @param {'uniform' | 'nonuniform'} [sampling=nonuniform] Sampling approach
-	 * @param {'gaussian' | 'polynomial' | function (number[], number[]): number} [kernel=gaussian] Kernel name
+	 * @param {'gaussian' | 'polynomial' | { name: 'gaussian', s?: number } | { name: 'polynomial', d?: number } | function (number[], number[]): number} [kernel=gaussian] Kernel name
 	 * @param {'zero_one' | 'hinge'} [loss=hinge] Loss type name
 	 */
 	constructor(
@@ -31,14 +31,17 @@ export default class BOGD {
 		if (typeof kernel === 'function') {
 			this._kernel = kernel
 		} else {
-			switch (kernel) {
+			if (typeof kernel === 'string') {
+				kernel = { name: kernel }
+			}
+			switch (kernel.name) {
 				case 'gaussian':
-					this._s = 1
+					this._s = kernel.s ?? 1
 					this._kernel = (a, b) =>
 						Math.exp(-(a.reduce((s, v, i) => s + (v - b[i]) ** 2, 0) ** 2) / this._s ** 2)
 					break
 				case 'polynomial':
-					this._d = 2
+					this._d = kernel.d ?? 2
 					this._kernel = (a, b) => (1 + a.reduce((s, v, i) => s + v * b[i])) ** this._d
 					break
 			}

lib/model/bpa.js

@@ -10,7 +10,7 @@ export default class BPA {
 	 * @param {number} [c=1] Regularization parameter
 	 * @param {number} [b=10] Budget size
 	 * @param {'simple' | 'projecting' | 'nn'} [version=simple] Version
-	 * @param {'gaussian' | 'polynomial' | function (number[], number[]): number} [kernel=gaussian] Kernel name
+	 * @param {'gaussian' | 'polynomial' | { name: 'gaussian', s?: number } | { name: 'polynomial', d?: number } | function (number[], number[]): number} [kernel=gaussian] Kernel name
 	 */
 	constructor(c = 1, b = 10, version = 'simple', kernel = 'gaussian') {
 		this._c = c
@@ -19,14 +19,17 @@ export default class BPA {
 		if (typeof kernel === 'function') {
 			this._kernel = kernel
 		} else {
-			switch (kernel) {
+			if (typeof kernel === 'string') {
+				kernel = { name: kernel }
+			}
+			switch (kernel.name) {
 				case 'gaussian':
-					this._s = 1
+					this._s = kernel.s ?? 1
 					this._kernel = (a, b) =>
 						Math.exp(-(a.reduce((s, v, i) => s + (v - b[i]) ** 2, 0) ** 2) / this._s ** 2)
 					break
 				case 'polynomial':
-					this._d = 2
+					this._d = kernel.d ?? 2
 					this._kernel = (a, b) => (1 + a.reduce((s, v, i) => s + v * b[i])) ** this._d
 					break
 			}

lib/model/bsgd.js

@@ -11,7 +11,7 @@ export default class BSGD {
 	 * @param {number} [eta=1] Learning rate
 	 * @param {number} [lambda=1] Regularization parameter
 	 * @param {'removal' | 'projection' | 'merging'} [maintenance=removal] Maintenance type
-	 * @param {'gaussian' | 'polynomial' | function (number[], number[]): number} [kernel=gaussian] Kernel name
+	 * @param {'gaussian' | 'polynomial' | { name: 'gaussian', s?: number } | { name: 'polynomial', d?: number } | function (number[], number[]): number} [kernel=gaussian] Kernel name
 	 */
 	constructor(b = 10, eta = 1, lambda = 1, maintenance = 'removal', kernel = 'gaussian') {
 		this._b = b
@@ -21,14 +21,17 @@ export default class BSGD {
 		if (typeof kernel === 'function') {
 			this._kernel = kernel
 		} else {
-			switch (kernel) {
+			if (typeof kernel === 'string') {
+				kernel = { name: kernel }
+			}
+			switch (kernel.name) {
 				case 'gaussian':
-					this._s = 1
+					this._s = kernel.s ?? 1
 					this._kernel = (a, b) =>
 						Math.exp(-(a.reduce((s, v, i) => s + (v - b[i]) ** 2, 0) ** 2) / this._s ** 2)
 					break
 				case 'polynomial':
-					this._d = 2
+					this._d = kernel.d ?? 2
 					this._kernel = (a, b) => (1 + a.reduce((s, v, i) => s + v * b[i])) ** this._d
 					break
 			}
@@ -195,7 +198,7 @@ export class MulticlassBSGD {
 	 * @param {number} [eta=1] Learning rate
 	 * @param {number} [lambda=1] Regularization parameter
 	 * @param {'removal' | 'projection' | 'merging'} [maintenance=removal] Maintenance type
-	 * @param {'gaussian' | 'polynomial' | function (number[], number[]): number} [kernel=gaussian] Kernel name
+	 * @param {'gaussian' | 'polynomial' | { name: 'gaussian', s?: number } | { name: 'polynomial', d?: number } | function (number[], number[]): number} [kernel=gaussian] Kernel name
 	 */
 	constructor(b = 10, eta = 1, lambda = 1, maintenance = 'removal', kernel = 'gaussian') {
 		this._b = b
@@ -205,14 +208,17 @@ export class MulticlassBSGD {
 		if (typeof kernel === 'function') {
 			this._kernel = kernel
 		} else {
-			switch (kernel) {
+			if (typeof kernel === 'string') {
+				kernel = { name: kernel }
+			}
+			switch (kernel.name) {
 				case 'gaussian':
-					this._s = 1
+					this._s = kernel.s ?? 1
 					this._kernel = (a, b) =>
 						Math.exp(-(a.reduce((s, v, i) => s + (v - b[i]) ** 2, 0) ** 2) / this._s ** 2)
 					break
 				case 'polynomial':
-					this._d = 2
+					this._d = kernel.d ?? 2
 					this._kernel = (a, b) => (1 + a.reduce((s, v, i) => s + v * b[i])) ** this._d
 					break
 			}

lib/model/coll.js

@@ -9,22 +9,25 @@ export default class COLL {
 	/**
 	 * @param {number} c Number of clusters
 	 * @param {number} [eta=1] Initial learning rate
-	 * @param {'gaussian' | 'polynomial' | function (number[], number[]): number} [kernel=gaussian] Kernel name
+	 * @param {'gaussian' | 'polynomial' | { name: 'gaussian', s?: number } | { name: 'polynomial', d?: number } | function (number[], number[]): number} [kernel=gaussian] Kernel name
 	 */
 	constructor(c, eta = 1, kernel = 'gaussian') {
 		this._c = c
 		this._eta = eta
 		if (typeof kernel === 'function') {
 			this._kernel = kernel
 		} else {
-			switch (kernel) {
+			if (typeof kernel === 'string') {
+				kernel = { name: kernel }
+			}
+			switch (kernel.name) {
 				case 'gaussian':
-					this._s = 1
+					this._s = kernel.s ?? 1
 					this._kernel = (a, b) =>
 						Math.exp(-(a.reduce((s, v, i) => s + (v - b[i]) ** 2, 0) ** 2) / this._s ** 2)
 					break
 				case 'polynomial':
-					this._d = 2
+					this._d = kernel.d ?? 2
 					this._kernel = (a, b) => (1 + a.reduce((s, v, i) => s + v * b[i])) ** this._d
 					break
 			}

lib/model/denclue.js

@@ -8,7 +8,7 @@ export default class DENCLUE {
 	/**
 	 * @param {number} h Smoothing parameter for the kernel
 	 * @param {1 | 2} [version=1] Version number
-	 * @param {'gaussian' | function (number[]): number} [kernel=gaussian] Kernel name
+	 * @param {'gaussian' | { name: 'gaussian' } | function (number[]): number} [kernel=gaussian] Kernel name
 	 */
 	constructor(h, version = 1, kernel = 'gaussian') {
 		this._version = version
@@ -19,7 +19,7 @@ export default class DENCLUE {
 
 		if (typeof kernel === 'function') {
 			this._kernel = kernel
-		} else if (kernel === 'gaussian') {
+		} else if (kernel === 'gaussian' || kernel.name === 'gaussian') {
 			this._kernel = u => {
 				const d = u.length
 				return Math.exp(-u.reduce((s, v) => s + v ** 2, 0) / 2) / (2 * Math.PI) ** (d / 2)

lib/model/diffusion_map.js

@@ -8,13 +8,13 @@ export default class DiffusionMap {
 	// https://inside.mines.edu/~whereman/papers/delaPorte-Herbst-Hereman-vanderWalt-PRASA-2008.pdf
 	/**
 	 * @param {number} t Power parameter
-	 * @param {'gaussian' | function (number[], number[]): number} [kernel=gaussian] Kernel name
+	 * @param {'gaussian' | { name: 'gaussian' } | function (number[], number[]): number} [kernel=gaussian] Kernel name
 	 */
 	constructor(t, kernel = 'gaussian') {
 		this._t = t
 		if (typeof kernel === 'function') {
 			this._k = kernel
-		} else if (kernel === 'gaussian') {
+		} else if (kernel === 'gaussian' || kernel.name === 'gaussian') {
 			this._k = (x, y) => Math.exp(-x.reduce((s, v, i) => s + (v - y[i]) ** 2, 0) / 2)
 		}
 	}

lib/model/forgetron.js

@@ -9,21 +9,24 @@ export default class Forgetron {
 	// https://github.com/LIBOL/KOL
 	/**
 	 * @param {number} b Budget parameter
-	 * @param {'gaussian' | 'polynomial' | function (number[], number[]): number} [kernel=gaussian] Kernel name
+	 * @param {'gaussian' | 'polynomial' | { name: 'gaussian', s?: number } | { name: 'polynomial', d?: number } | function (number[], number[]): number} [kernel=gaussian] Kernel name
 	 */
 	constructor(b, kernel = 'gaussian') {
 		this._b = b
 		if (typeof kernel === 'function') {
 			this._kernel = kernel
 		} else {
-			switch (kernel) {
+			if (typeof kernel === 'string') {
+				kernel = { name: kernel }
+			}
+			switch (kernel.name) {
 				case 'gaussian':
-					this._s = 1
+					this._s = kernel.s ?? 1
 					this._kernel = (a, b) =>
 						Math.exp(-(a.reduce((s, v, i) => s + (v - b[i]) ** 2, 0) ** 2) / this._s ** 2)
 					break
 				case 'polynomial':
-					this._d = 2
+					this._d = kernel.d ?? 2
 					this._kernel = (a, b) => (1 + a.reduce((s, v, i) => s + v * b[i])) ** this._d
 					break
 			}