Scatterplot branch

.gitignore

@@ -1,2 +1 @@
-
-**/.DS_Store
+**/.DS_Store

classes/Datasets/Transforms.sc

@@ -0,0 +1,122 @@
+/*****************************************
+Data Transforms
+(C) 2018 Jonathan Reus
+
+Tools for analyzing and transforming datasets: scaling, normalization, standardization, PCA, etc..
+
+
+******************************************/
+
+// TO NORMALIZE:
+// find min / max of features
+// normalized_value = (val - min) / (max - min)
+
+
+/*
+T for Transform
+Normalization:
+1. find min/max of each feature
+2. normalized_value = (val - min) / (max - min)
+*/
+TNormalizer {
+	var <min, <max, <originalData, <normalizedData, dim;
+
+	/*
+	@param dataset a Matrix of rows as feature vectors
+	*/
+	*new {|dataset|
+		^super.new.init(dataset);
+	}
+
+	init {|dataset|
+		if(dataset.isKindOf(SequenceableCollection).not ) { "Dataset must be a Matrix".error;this.halt; };
+		if(dataset.at(0).isKindOf(SequenceableCollection)) {
+			dim = dataset.cols;
+			min = Array.newClear(dim);
+			max = Array.newClear(dim);
+			normalizedData = Matrix.newClear(dataset.rows, dataset.cols);
+			dim.do {|i|
+				var col = dataset.getCol(i);
+				min[i] = col.minItem;
+				max[i] = col.maxItem;
+				normalizedData.putCol(i, this.normalizeSample(col) );
+			};
+		} {
+			min = dataset.minItem; max = dataset.maxItem;
+			normalizedData = dataset.collect {|it,i| this.normalizeSample(it) };
+		};
+		originalData = dataset;
+	}
+
+	// normalize / denormalize a single sample
+	normalizeSample {|samp| (samp - min) / (max - min) }
+	denormalizeSample {|samp| (samp * (max-min)) + min }
+
+
+	// normalize an entire dataset
+	normalizeData {}
+
+	// denormalize a point-slope form 2-dimensional line
+	// of the form [slope, intercept]
+	denormalizeLine {|line|
+			var p1,p2, new_m, new_b, m=line[0], b=line[1];
+			// calculate two normalized samples & denormalize them
+			p1 = [-1,(-1 * m)+b]; p2 = [1,(1 * m)+b];
+			p1 = this.denormalizeSample(p1);
+			p2 = this.denormalizeSample(p2);
+			// calculate denormalized decision boundary
+			new_m = (p2[1]-p1[1]) / (p2[0]-p1[0]); // slope
+			new_b = p1[1] - (new_m*p1[0]); // y-intercept
+			^[new_m,new_b];
+	}
+
+}
+
+/*
+TO STANDARDIZE:
+1. calculate mean and standard deviation of each feature
+2. subtract mean from each feature
+3. divide features by standard deviation
+
+T is for Transform
+*/
+TStandardizer {
+	var <mean, <stddev;
+
+}
+
+/*
+T for Transform
+*/
+TPCA {
+
+}
+
+
+
+/*
+( // feature scaling of reduced dataset
+~dReduced = Matrix.newFrom(~dReduced);
+dim = ~dReduced.cols;
+mean = Array.newClear(dim);
+stdev = Array.newClear(dim);
+
+~standardizeSample = {|v,mean,stddev| (v-mean) / stddev };
+~destandardizeSample = {|v,mean,stddev| (v*stddev) + mean };
+
+~dStandard = Matrix.newClear(~dReduced.rows, dim);
+dim.do {|i|
+	var col = ~dReduced.getCol(i);
+	mean[i] = col.mean;
+	stdev[i] = col.stdDev(mean[i]);
+	~dStandard.putCol(i, ~standardizeSample.(col, mean[i], stdev[i]));
+};
+"MEAN: %   STDDEV: %".format(mean, stdev).postln;
+s1 = [-0.09, 1.45];
+s2 = ~standardizeSample.(s1, mean, stdev);
+s3 = ~destandardizeSample.(s2, mean, stdev);
+"NEW SAMPLE: %   STANDARDIZED: %  DESTANDARDIZED: %".format(s1,s2,s3).postln;
+
+); // END SCALING & STANDARDIZATION OF FEATURES
+
+*/