fix: 1d example

README.md

@@ -29,14 +29,29 @@ npm install kalman-filter
 ```js
 const {KalmanFilter} = require('kalman-filter');
 
-const observations = [0, 0.1, 0.5, 0.2];
+const observations = [0, 0.1, 0.5, 0.2, 3, 4, 2, 1, 2, 3, 5, 6];
 const kFilter = new KalmanFilter();
 const res = kFilter.filterAll(observations)
-
-console.log(res);
+// res is a list of 1x1 matrices
+// [
+//   [ [ 0 ] ],
+//   [ [ 0.06666665555510715 ] ],
+//   [ [ 0.3374999890620582 ] ],
+//   [ [ 0.25238094852592136 ] ],
+//   [ [ 1.9509090885288296 ] ],
+//   [ [ 3.2173611101031616 ] ],
+//   [ [ 2.4649867370240965 ] ],
+//   [ [ 1.5595744679428254 ] ],
+//   [ [ 1.831772445766021 ] ],
+//   [ [ 2.5537767922925685 ] ],
+//   [ [ 4.065625882212133 ] ],
+//   [ [ 5.26113483436549 ] ]
+// ]
 ```
+Result is :
+
+![Kalman Filter 1d example](./demo/example-1D.png)
 
-**TO DO** add a screenshot of the resulting curve
 
 ## How to instantiate your kalman filter
 

lib/kalman-filter.js

@@ -20,7 +20,7 @@ const distanceMat = require('../lib/linalgebra/distance-mat.js');
 *@param {ObservationConfig} options.observation
 */
 
-const setupModelsParameters = function ({observation, dynamic}) {
+const setupModelsParameters = function ({observation = {name: 'sensors'}, dynamic = {name: 'constant-position', dimension: 1}}) {
 	if (typeof (observation.name) === 'string') {
 		observation = modelCollection.buildObservation(observation);
 	}

lib/state.js

@@ -1,3 +1,9 @@
+const sub = require('../lib/linalgebra/sub.js');
+const transpose = require('../lib/linalgebra/transpose.js');
+const matMul = require('../lib/linalgebra/mat-mul.js');
+const invert = require('../lib/linalgebra/invert.js');
+const arrayToMatrix = require('../lib/utils/array-to-matrix.js');
+
 const checkMatrix = function (matrix, shape) {
 	if (matrix.reduce((a, b) => a.concat(b)).filter(a => Number.isNaN(a)).length > 0) {
 		throw (new Error('Matrix should not have a NaN'));
@@ -61,6 +67,30 @@ class State {
 		// 	throw (new TypeError('t must be a number'));
 		// }
 	}
+
+	mahalanobis({kf, observation}) {
+		const correctlySizedObservation = arrayToMatrix({observation, dimension: observation.length});
+		const stateProjection = kf.observation.stateProjection();
+		const diff = sub(matMul(stateProjection, this.mean), correctlySizedObservation);
+		const diffTransposed = transpose(diff);
+
+		const covarianceInObservationSpace = matMul(
+			matMul(stateProjection, this.covariance),
+			transpose(stateProjection)
+		);
+		// Console.log('covariance in obs space', covarianceInObservationSpace);
+		const covarianceInvert = invert(covarianceInObservationSpace);
+		// Console.log('invert cov', covarianceInvert);
+		return Math.sqrt(
+			matMul(
+				matMul(
+					diffTransposed,
+					covarianceInvert
+				),
+				diff
+			)
+		);
+	}
 }
 
 module.exports = State;

lib/utils/array-to-matrix.js

@@ -8,6 +8,10 @@
 
 module.exports = function ({observation, dimension}) {
 	if (!Array.isArray(observation)) {
+		if (dimension === 1 && typeof (observation) === 'number') {
+			return [[observation]];
+		}
+
 		throw (new TypeError('The observation should be an array'));
 	}
 

package.json

@@ -4,7 +4,7 @@
   "description": "Kalman filter (and Extended Kalman Filter) Multi-dimensional implementation in Javascript",
   "main": "index.js",
   "scripts": {
-    "test": "xo && ava test/unit/kalman-filter/*",
+    "test": "xo && ava",
     "build-demo": "browserify -d -r ./demo/src/main.js:main > demo/dist/demo.js"
   },
   "repository": {

test/api/basic.js

@@ -8,11 +8,11 @@ const State = require('../../lib/state.js');
 const observations = [[0, 2], [0.1, 4], [0.5, 9], [0.2, 12]];
 
 test('Default filter : Constant-position on 1D Data', t => {
-	const observations = [0, 0.1, 0.5, 0.2];
+	const observations = [0, 0.1, 0.5, 0.2, 3, 4, 2, 1, 2, 3, 5, 6];
 	const kFilter = new KalmanFilter();
-	const res = kFilter.filterAll(observations)
-	t.true(Array.isArray(res));
-	t.is(res.length, observations.length);
+	const result = kFilter.filterAll(observations);
+	t.true(Array.isArray(result));
+	t.is(result.length, observations.length);
 });
 
 test('Constant-position on 2D Data', t => {
@@ -134,7 +134,7 @@ test('Sensor observation', t => {
 		observation: {
 			sensorDimension: 2, // Observation.dimension == observation.sensorDimension * observation.nSensors
 			nSensors: 2,
-			sensorCovariance: [3, 3, 4, 4], 
+			sensorCovariance: [3, 3, 4, 4],
 			name: 'sensors'
 		},
 		dynamic: {
@@ -192,7 +192,7 @@ test('Model fits ', t => {
 		},
 		dynamic: {
 			name: 'constant-speed', // Observation.sensorDimension == dynamic.dimension
-			covariance: [3, 4]// Equivalent to diag([3, 4])
+			covariance: [3, 3, 4, 4]
 		}
 	});
 	const observations = [[0, 2], [0.1, 4], [0.5, 9], [0.2, 12]];
@@ -205,7 +205,7 @@ test('Model fits ', t => {
 			previousCorrected
 		});
 
-		const dist = predicted.mahalanobis(observation);
+		const dist = predicted.mahalanobis({observation, kf: kFilter});
 
 		previousCorrected = kFilter.correct({
 			predicted,