Added split method for data

examples/ml_multilinreg_plot/main.v

@@ -0,0 +1,34 @@
+module main
+
+import vsl.ml
+
+fn main() {
+	xy := [
+		[0.99, 90.01],
+		[1.02, 89.05],
+		[1.15, 91.43],
+		[1.29, 93.74],
+		[1.46, 96.73],
+		[1.36, 94.45],
+		[0.87, 87.59],
+		[1.23, 91.77],
+		[1.55, 99.42],
+		[1.40, 93.65],
+		[1.19, 93.54],
+		[1.15, 92.52],
+		[0.98, 90.56],
+		[1.01, 89.54],
+		[1.11, 89.85],
+		[1.20, 90.39],
+		[1.26, 93.25],
+		[1.32, 93.41],
+		[1.43, 94.98],
+		[0.95, 87.33],
+	]
+	mut data := ml.data_from_raw_xy(xy)!
+	mut reg := ml.new_multi_lin_reg(mut data, 3, 'linear regression')
+
+	reg.train()
+
+	reg.plot()!
+}

la/matrix.v

@@ -217,9 +217,9 @@ pub fn (o &Matrix[T]) get_col(j int) []T {
 // extract_cols returns columns from j=start to j=endp1-1
 // start -- first column
 // endp1 -- "end-plus-one", the number of the last requested column + 1
-pub fn (o &Matrix[T]) extract_cols(start int, endp1 int) &Matrix[T] {
+pub fn (o &Matrix[T]) extract_cols(start int, endp1 int) !&Matrix[T] {
 	if endp1 <= start {
-		errors.vsl_panic("endp1 'end-plus-one' must be greater than start. start=${start}, endp1=${endp1} invalid",
+		return errors.error("endp1 'end-plus-one' must be greater than start. start=${start}, endp1=${endp1} invalid",
 			.efailed)
 	}
 	ncol := endp1 - start
@@ -235,9 +235,9 @@ pub fn (o &Matrix[T]) extract_cols(start int, endp1 int) &Matrix[T] {
 // extract_rows returns rows from i=start to i=endp1-1
 // start -- first column
 // endp1 -- "end-plus-one", the number of the last requested column + 1
-pub fn (o &Matrix[T]) extract_rows(start int, endp1 int) &Matrix[T] {
+pub fn (o &Matrix[T]) extract_rows(start int, endp1 int) !&Matrix[T] {
 	if endp1 <= start {
-		errors.vsl_panic("endp1 'end-plus-one' must be greater than start. start=${start}, endp1=${endp1} invalid",
+		return errors.error("endp1 'end-plus-one' must be greater than start. start=${start}, endp1=${endp1} invalid",
 			.efailed)
 	}
 	nrow := endp1 - start
@@ -260,6 +260,44 @@ pub fn (mut o Matrix[T]) set_col(j int, value T) {
 	}
 }
 
+// split_by_col splits this matrix into two matrices at column j
+// j -- column index
+pub fn (o &Matrix[T]) split_by_col(j int) !(&Matrix[T], &Matrix[T]) {
+	if j < 0 || j >= o.n {
+		return errors.error('j=${j} must be in range [0, ${o.n})', .efailed)
+	}
+	mut left := new_matrix[T](o.m, j)
+	mut right := new_matrix[T](o.m, o.n - j)
+	for i in 0 .. o.m {
+		for k := 0; k < j; k++ {
+			left.set(i, k, o.get(i, k))
+		}
+		for k := j; k < o.n; k++ {
+			right.set(i, k - j, o.get(i, k))
+		}
+	}
+	return left, right
+}
+
+// split_by_row splits this matrix into two matrices at row i
+// i -- row index
+pub fn (o &Matrix[T]) split_by_row(i int) !(&Matrix[T], &Matrix[T]) {
+	if i < 0 || i >= o.m {
+		return errors.error('i=${i} must be in range [0, ${o.m})', .efailed)
+	}
+	mut top := new_matrix[T](i, o.n)
+	mut bottom := new_matrix[T](o.m - i, o.n)
+	for j in 0 .. o.n {
+		for k := 0; k < i; k++ {
+			top.set(k, j, o.get(k, j))
+		}
+		for k := i; k < o.m; k++ {
+			bottom.set(k - i, j, o.get(k, j))
+		}
+	}
+	return top, bottom
+}
+
 // norm_frob returns the Frobenius norm of this matrix
 // nrm := ‖a‖_F = sqrt(Σ_i Σ_j a[ij]⋅a[ij]) = ‖a‖_2
 pub fn (o &Matrix[T]) norm_frob() T {

la/matrix_test.v

@@ -175,7 +175,7 @@ fn test_extract_cols_and_set_col() {
 		[31.0, 32.0, 33.0],
 	])
 	println(mat.data)
-	mat_e := mat.extract_cols(1, 3)
+	mat_e := mat.extract_cols(1, 3)!
 	assert mat_e.m == 3
 	assert mat_e.n == 2
 	assert mat_e.data == [12.0, 13, 22, 23, 32, 33]

ml/data.v

@@ -1,5 +1,6 @@
 module ml
 
+import math
 import vsl.util
 import vsl.la
 import vsl.errors
@@ -38,7 +39,11 @@ pub mut:
 // Output:
 // new object
 pub fn new_data[T](nb_samples int, nb_features int, use_y bool, allocate bool) !&Data[T] {
-	x := if allocate { la.new_matrix[T](nb_samples, nb_features) } else { la.new_matrix[T](0, 0) }
+	x := if allocate {
+		la.new_matrix[T](nb_samples, nb_features)
+	} else {
+		&la.Matrix[T](unsafe { nil })
+	}
 	mut y := []T{}
 	if allocate && use_y {
 		y = []T{len: nb_samples}
@@ -186,3 +191,29 @@ pub fn (mut o Data[T]) add_observer(obs util.Observer) {
 pub fn (mut o Data[T]) notify_update() {
 	o.observable.notify_update()
 }
+
+// split returns a new object with data split into two parts
+// Input:
+// ratio -- ratio of samples to be put in the first part
+// Output:
+// new object
+pub fn (o &Data[T]) split(ratio f64) !(&Data[T], &Data[T]) {
+	if ratio <= 0.0 || ratio >= 1.0 {
+		return errors.error('ratio must be between 0 and 1', .efailed)
+	}
+	nb_features := o.nb_features
+	nb_samples := o.nb_samples
+
+	nb_samples1 := int(math.floor((ratio * nb_samples)))
+	nb_samples2 := nb_samples - nb_samples1
+
+	m1, m2 := o.x.split_by_row(nb_samples1)!
+
+	mut o1 := new_data[T](nb_samples1, nb_features, false, false)!
+	mut o2 := new_data[T](nb_samples2, nb_features, false, false)!
+
+	o1.set(m1, o.y[..nb_samples1])!
+	o2.set(m2, o.y[nb_samples1..])!
+
+	return o1, o2
+}

ml/data_test.v

@@ -31,3 +31,19 @@ fn test_data_01() {
 	assert data_backup.nb_features == 3
 	assert data_backup.nb_samples == 5
 }
+
+fn test_split() {
+	data := data_from_raw_xy([
+		[-1.0, 0, -3, 0],
+		[-2.0, 3, 3, 1],
+		[3.0, 1, 4, 1],
+		[-4.0, 5, 0, 0],
+		[1.0, -8, 5, 1],
+		[-1.0, 0, -3, 1],
+	])!
+	data1, data2 := data.split(0.2)!
+	assert data1.nb_samples == 1
+	assert data2.nb_samples == 5
+	assert data1.nb_features == 3
+	assert data2.nb_features == 3
+}