gprx 0.2.0

Gaussian process regression in Rust: exact GPR, sparse GPR (VFE), and SVGP.
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
//! Constant kernel `k(x, x') = c`.

use super::{
    Triangle, require_same_shape, require_square_pair, validate_log_positive,
    validate_positive_finite, write_square,
};
use crate::error::GprError;
use crate::kernel::KernelScalar;
use crate::param::{BoundedParam, Interval};
use faer::{MatMut, MatRef};

/// Evaluates the constant kernel `k = c` for every pair of points.
///
/// The optimizer parameter is `θ = log(c)`. Compose with a stationary leaf
/// (for example [`super::RbfKernel`]) to set the signal variance. `dist` is
/// used only for shape checks.
///
/// # Examples
///
/// ```rust
/// use faer::Mat;
/// use gprx::Interval;
/// use gprx::kernel::{ConstantKernel, Triangle};
///
/// # fn main() -> Result<(), gprx::GprError> {
/// // Parameter: `θ = log(c)`; the bounds are on `c`.
/// let k = ConstantKernel::new(1.5)?;
/// assert!((k.constant() - 1.5).abs() < 1e-12);
/// let same = ConstantKernel::from_log_constant(k.log_constant())?;
/// assert!((same.constant() - 1.5).abs() < 1e-12);
/// let mut k = k.with_bounds(Interval::new(0.01, 100.0)?)?;
/// assert!((k.bounds().hi() - 100.0).abs() < 1e-12);
/// let mut theta = vec![0.0; k.num_params()];
/// k.get_params(&mut theta)?;
/// k.set_params(&[2.0_f64.ln()])?; // c = 2
///
/// // `dist` (squared distances) and `x` (coordinates) only give the shape.
/// let dist = Mat::<f64>::zeros(3, 3);
/// let x = Mat::from_fn(3, 1, |i, _| i as f64);
/// let mut gram = Mat::<f64>::zeros(3, 3);
/// k.apply(dist.as_ref(), gram.as_mut(), Triangle::Full)?;
/// assert!((gram[(2, 0)] - 2.0).abs() < 1e-12);
/// k.apply_points(x.as_ref(), gram.as_mut(), Triangle::Lower)?;
/// let xs = Mat::from_fn(1, 1, |_, _| 0.5);
/// let mut k_cross = Mat::<f64>::zeros(3, 1);
/// k.apply_cross(Mat::<f64>::zeros(3, 1).as_ref(), k_cross.as_mut())?;
/// k.apply_cross_points(x.as_ref(), xs.as_ref(), k_cross.as_mut())?;
/// let mut diag = [0.0; 3];
/// k.fill_diag(&mut diag);
///
/// // ∂K/∂θ = c and ∂²K/∂θ² = c, from either view.
/// let mut d_k = Mat::<f64>::zeros(3, 3);
/// k.grad(dist.as_ref(), d_k.as_mut(), 0, Triangle::Lower)?;
/// k.grad_points(x.as_ref(), d_k.as_mut(), 0, Triangle::Lower)?;
/// let mut d2_k = Mat::<f64>::zeros(3, 3);
/// k.hess(dist.as_ref(), d2_k.as_mut(), 0, 0, Triangle::Lower)?;
/// k.hess_points(x.as_ref(), d2_k.as_mut(), 0, 0, Triangle::Lower)?;
/// assert!((d2_k[(1, 0)] - 2.0).abs() < 1e-12);
/// # Ok(())
/// # }
/// ```
#[derive(Clone, Copy, Debug, PartialEq)]
pub struct ConstantKernel {
    constant: BoundedParam,
}

impl ConstantKernel {
    /// Builds a constant kernel from a positive finite value `c`.
    ///
    /// # Errors
    ///
    /// Returns [`GprError::InvalidHyperparameter`] if `constant` is not finite
    /// or not strictly positive, or [`GprError::InvalidInterval`] if it is
    /// not inside [`Interval::DEFAULT_POSITIVE`], the bounds a new leaf
    /// starts with.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn new(constant: f64) -> Result<Self, GprError> {
        validate_positive_finite(constant, "constant value")?;
        Ok(Self {
            constant: BoundedParam::default_positive(constant)?,
        })
    }

    /// Builds a constant kernel from `θ = log(c)`.
    ///
    /// # Errors
    ///
    /// Returns [`GprError::InvalidHyperparameter`] if `θ` is not finite, if
    /// `exp(θ)` overflows, or if `exp(θ)` underflows to zero, or
    /// [`GprError::InvalidInterval`] if `exp(θ)` is not inside
    /// [`Interval::DEFAULT_POSITIVE`].
    ///
    /// See the example on [`ConstantKernel`].
    pub fn from_log_constant(log_constant: f64) -> Result<Self, GprError> {
        let log_constant = validate_log_positive(log_constant, "constant value")?;
        Ok(Self {
            constant: BoundedParam::default_positive(log_constant.exp())?,
        })
    }

    /// Returns `c = exp(θ)`.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn constant(&self) -> f64 {
        self.constant.value()
    }

    /// Returns `θ = log(c)`.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn log_constant(&self) -> f64 {
        self.constant.ln()
    }

    /// Returns the open interval on `c`.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn bounds(&self) -> Interval {
        self.constant.interval()
    }

    /// Rebuilds this kernel with a new interval on `c`.
    ///
    /// # Errors
    ///
    /// Returns [`crate::IntervalError`] if the current `c` is not strictly
    /// inside `interval`.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn with_bounds(self, interval: Interval) -> Result<Self, crate::IntervalError> {
        Ok(Self {
            constant: self.constant.with_interval(interval)?,
        })
    }

    /// Returns the number of optimizer parameters (always 1).
    ///
    /// See the example on [`ConstantKernel`].
    pub fn num_params(&self) -> usize {
        1
    }

    /// Writes `θ` into a length-1 slice.
    ///
    /// # Errors
    ///
    /// Returns [`GprError::LengthMismatch`] if `out` is not length 1.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn get_params(&self, out: &mut [f64]) -> Result<(), GprError> {
        crate::data::require_count(out.len(), 1, "constant parameter")?;
        out[0] = self.constant.ln();
        Ok(())
    }

    /// Replaces `θ` from a length-1 slice.
    ///
    /// # Errors
    ///
    /// Returns [`GprError::LengthMismatch`] if `params` is not length 1,
    /// [`GprError::InvalidHyperparameter`] if the new `θ` is invalid, or
    /// [`GprError::InvalidInterval`] if the new `c` is not inside
    /// [`Self::bounds`].
    ///
    /// See the example on [`ConstantKernel`].
    pub fn set_params(&mut self, params: &[f64]) -> Result<(), GprError> {
        crate::data::require_count(params.len(), 1, "constant parameter")?;
        let log_constant = validate_log_positive(params[0], "constant value")?;
        self.constant = BoundedParam::new(log_constant.exp(), self.constant.interval())?;
        Ok(())
    }

    /// Writes `c` into `out` for the requested triangle.
    ///
    /// # Errors
    ///
    /// Returns [`GprError::ShapeMismatch`] if `dist` is not square or `out` is not
    /// its shape, or [`GprError::EmptyInput`] if they are empty.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn apply<T: KernelScalar>(
        &self,
        dist: MatRef<'_, T>,
        out: MatMut<'_, T>,
        uplo: Triangle,
    ) -> Result<(), GprError> {
        require_square_pair(dist, out.as_ref())?;
        self.write_square(out, uplo)
    }

    /// Writes rectangular `k = c` into `out`.
    ///
    /// # Errors
    ///
    /// Returns [`GprError::ShapeMismatch`] if `out` is not the shape of `dist`, or
    /// [`GprError::EmptyInput`] if they are empty.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn apply_cross<T: KernelScalar>(
        &self,
        dist: MatRef<'_, T>,
        mut out: MatMut<'_, T>,
    ) -> Result<(), GprError> {
        require_same_shape(dist, out.as_ref())?;
        let c = T::from_f64(self.constant());
        for col in 0..out.ncols() {
            for row in 0..out.nrows() {
                out[(row, col)] = c;
            }
        }
        Ok(())
    }

    /// Writes `c` from the number of rows of `x` into a square `out`.
    ///
    /// # Errors
    ///
    /// Returns [`GprError::EmptyInput`] if `x` has no row or no column, or
    /// [`GprError::ShapeMismatch`] if `out` is not `n×n`.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn apply_points<T: KernelScalar>(
        &self,
        x: MatRef<'_, T>,
        out: MatMut<'_, T>,
        uplo: Triangle,
    ) -> Result<(), GprError> {
        require_columns(x)?;
        self.apply_rows(x, out, uplo)
    }

    /// [`Self::apply_points`] without the column check: only the shape of the
    /// rows is read, so a tree on supplied distances passes no columns.
    pub(crate) fn apply_rows<T: KernelScalar>(
        &self,
        x: MatRef<'_, T>,
        out: MatMut<'_, T>,
        uplo: Triangle,
    ) -> Result<(), GprError> {
        require_points_square(x, out.as_ref())?;
        self.write_square(out, uplo)
    }

    /// Writes rectangular `k = c` for train × test coordinates.
    ///
    /// # Errors
    ///
    /// Returns [`GprError::EmptyInput`] if `x` or `xs` has no row or no column, or
    /// [`GprError::ShapeMismatch`] if `out` is not `x.nrows() × xs.nrows()`.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn apply_cross_points<T: KernelScalar>(
        &self,
        x: MatRef<'_, T>,
        xs: MatRef<'_, T>,
        out: MatMut<'_, T>,
    ) -> Result<(), GprError> {
        require_columns(x)?;
        require_columns(xs)?;
        self.apply_cross_rows(x, xs, out)
    }

    /// [`Self::apply_cross_points`] without the column check: only the shape of the
    /// rows is read, so a tree on supplied distances passes no columns.
    pub(crate) fn apply_cross_rows<T: KernelScalar>(
        &self,
        x: MatRef<'_, T>,
        xs: MatRef<'_, T>,
        mut out: MatMut<'_, T>,
    ) -> Result<(), GprError> {
        require_cross_points(x, xs, out.as_ref())?;
        let c = T::from_f64(self.constant());
        for col in 0..out.ncols() {
            for row in 0..out.nrows() {
                out[(row, col)] = c;
            }
        }
        Ok(())
    }

    /// Writes the diagonal `k(x, x) = c` into `out`.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn fill_diag<T: KernelScalar>(&self, out: &mut [T]) {
        out.fill(T::from_f64(self.constant()));
    }

    /// Writes `∂K/∂θ` for `θ = log(c)` into `d_k` (`∂k/∂θ = c`).
    ///
    /// # Errors
    ///
    /// Returns [`GprError::IndexOutOfRange`] if `param_idx` is not 0, or
    /// the same shape errors as [`Self::apply`].
    ///
    /// See the example on [`ConstantKernel`].
    pub fn grad<T: KernelScalar>(
        &self,
        dist: MatRef<'_, T>,
        d_k: MatMut<'_, T>,
        param_idx: usize,
        uplo: Triangle,
    ) -> Result<(), GprError> {
        require_param_idx(param_idx)?;
        require_square_pair(dist, d_k.as_ref())?;
        self.write_square(d_k, uplo)
    }

    /// Writes `∂K/∂θ` from coordinates.
    ///
    /// `x` is used only for shape.
    ///
    /// # Errors
    ///
    /// Same as [`Self::grad`], with `x` in place of `dist`.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn grad_points<T: KernelScalar>(
        &self,
        x: MatRef<'_, T>,
        d_k: MatMut<'_, T>,
        param_idx: usize,
        uplo: Triangle,
    ) -> Result<(), GprError> {
        require_columns(x)?;
        self.grad_rows(x, d_k, param_idx, uplo)
    }

    /// [`Self::grad_points`] without the column check: only the shape of the
    /// rows is read, so a tree on supplied distances passes no columns.
    pub(crate) fn grad_rows<T: KernelScalar>(
        &self,
        x: MatRef<'_, T>,
        d_k: MatMut<'_, T>,
        param_idx: usize,
        uplo: Triangle,
    ) -> Result<(), GprError> {
        require_param_idx(param_idx)?;
        require_points_square(x, d_k.as_ref())?;
        self.write_square(d_k, uplo)
    }

    /// Writes the rectangular `∂K(x, xs)/∂θ` (`∂k/∂θ = c`) for train × test
    /// rows. Only the rows are read: a tree's entry points reject
    /// coordinates without a column for a coordinate tree
    /// ([`crate::kernel::CompiledKernel::require_tree_columns`]).
    ///
    /// # Errors
    ///
    /// Returns [`GprError::IndexOutOfRange`] if `param_idx` is not 0, or the
    /// errors of [`Self::apply_cross_rows`].
    pub(crate) fn grad_cross_rows<T: KernelScalar>(
        &self,
        x: MatRef<'_, T>,
        xs: MatRef<'_, T>,
        d_k: MatMut<'_, T>,
        param_idx: usize,
    ) -> Result<(), GprError> {
        require_param_idx(param_idx)?;
        self.apply_cross_rows(x, xs, d_k)
    }

    /// Writes the rectangular `∂²K(x, xs)/∂θ²` (`∂²k/∂θ² = c`) for train ×
    /// test rows, as [`Self::grad_cross_rows`] reads them.
    ///
    /// # Errors
    ///
    /// Returns [`GprError::IndexOutOfRange`] if `i` or `j` is not 0, or the
    /// errors of [`Self::apply_cross_rows`].
    pub(crate) fn hess_cross_rows<T: KernelScalar>(
        &self,
        x: MatRef<'_, T>,
        xs: MatRef<'_, T>,
        d2_k: MatMut<'_, T>,
        i: usize,
        j: usize,
    ) -> Result<(), GprError> {
        require_hess_idx(i, j)?;
        self.apply_cross_rows(x, xs, d2_k)
    }

    /// Writes `∂²K/∂θ²` for `θ = log(c)` into `d2_k` (`∂²k/∂θ² = c`).
    ///
    /// # Errors
    ///
    /// Returns [`GprError::IndexOutOfRange`] if `i` or `j` is not 0, or
    /// the same shape errors as [`Self::apply`].
    ///
    /// See the example on [`ConstantKernel`].
    pub fn hess<T: KernelScalar>(
        &self,
        dist: MatRef<'_, T>,
        d2_k: MatMut<'_, T>,
        i: usize,
        j: usize,
        uplo: Triangle,
    ) -> Result<(), GprError> {
        require_hess_idx(i, j)?;
        require_square_pair(dist, d2_k.as_ref())?;
        self.write_square(d2_k, uplo)
    }

    /// Writes `∂²K/∂θ²` from coordinates.
    ///
    /// `x` is used only for shape.
    ///
    /// # Errors
    ///
    /// Same as [`Self::hess`], with `x` in place of `dist`.
    ///
    /// See the example on [`ConstantKernel`].
    pub fn hess_points<T: KernelScalar>(
        &self,
        x: MatRef<'_, T>,
        d2_k: MatMut<'_, T>,
        i: usize,
        j: usize,
        uplo: Triangle,
    ) -> Result<(), GprError> {
        require_columns(x)?;
        self.hess_rows(x, d2_k, i, j, uplo)
    }

    /// [`Self::hess_points`] without the column check: only the shape of the
    /// rows is read, so a tree on supplied distances passes no columns.
    pub(crate) fn hess_rows<T: KernelScalar>(
        &self,
        x: MatRef<'_, T>,
        d2_k: MatMut<'_, T>,
        i: usize,
        j: usize,
        uplo: Triangle,
    ) -> Result<(), GprError> {
        require_hess_idx(i, j)?;
        require_points_square(x, d2_k.as_ref())?;
        self.write_square(d2_k, uplo)
    }

    fn write_square<T: KernelScalar>(
        &self,
        out: MatMut<'_, T>,
        uplo: Triangle,
    ) -> Result<(), GprError> {
        let c = T::from_f64(self.constant());
        write_square(out, uplo, |_, _| Ok(c))
    }
}

fn require_param_idx(param_idx: usize) -> Result<(), GprError> {
    if param_idx == 0 {
        Ok(())
    } else {
        Err(GprError::IndexOutOfRange {
            reason: "constant kernel has a single parameter at index 0".to_owned(),
        })
    }
}

fn require_hess_idx(i: usize, j: usize) -> Result<(), GprError> {
    if i == 0 && j == 0 {
        Ok(())
    } else {
        Err(GprError::IndexOutOfRange {
            reason: format!("constant kernel has a single parameter; got pair ({i}, {j})"),
        })
    }
}

fn require_points_square<T: KernelScalar>(
    x: MatRef<'_, T>,
    out: MatRef<'_, T>,
) -> Result<(), GprError> {
    if x.nrows() == 0 {
        return Err(GprError::EmptyInput);
    }
    if out.nrows() != x.nrows() || out.ncols() != x.nrows() {
        return Err(GprError::ShapeMismatch {
            reason: format!(
                "output is {}x{}, expected {}x{}",
                out.nrows(),
                out.ncols(),
                x.nrows(),
                x.nrows()
            ),
        });
    }
    Ok(())
}

fn require_cross_points<T: KernelScalar>(
    x: MatRef<'_, T>,
    xs: MatRef<'_, T>,
    out: MatRef<'_, T>,
) -> Result<(), GprError> {
    if x.nrows() == 0 || xs.nrows() == 0 {
        return Err(GprError::EmptyInput);
    }
    if out.nrows() != x.nrows() || out.ncols() != xs.nrows() {
        return Err(GprError::ShapeMismatch {
            reason: format!(
                "output is {}x{}, expected {}x{}",
                out.nrows(),
                out.ncols(),
                x.nrows(),
                xs.nrows()
            ),
        });
    }
    Ok(())
}

/// Rejects coordinates without a column ([`GprError::EmptyInput`]).
fn require_columns<T>(x: MatRef<'_, T>) -> Result<(), GprError> {
    if x.ncols() == 0 {
        return Err(GprError::EmptyInput);
    }
    Ok(())
}

#[cfg(test)]
mod tests {
    use super::ConstantKernel;
    use crate::error::GprError;
    use crate::kernel::Triangle;
    use faer::{Mat, mat};

    const TOL: f64 = 1e-8;

    use crate::test_check::{assert_close, assert_lower_close, fill};

    fn dummy_dist(n: usize) -> Mat<f64> {
        Mat::from_fn(n, n, |i, j| ((i + j) as f64) * 0.1)
    }

    #[test]
    fn apply_is_constant_and_symmetric() {
        let kernel = ConstantKernel::new(2.5).expect("valid");
        let dist = dummy_dist(3);
        let mut k = fill(3, 0.0);
        kernel
            .apply(dist.as_ref(), k.as_mut(), Triangle::Full)
            .expect("shape");
        for col in 0..3 {
            for row in 0..3 {
                assert_close(k[(row, col)], 2.5, TOL);
                assert_close(k[(row, col)], k[(col, row)], TOL);
            }
        }
    }

    #[test]
    fn lower_matches_full_and_leaves_upper() {
        let kernel = ConstantKernel::new(0.75).expect("valid");
        let dist = dummy_dist(3);
        let mut full = fill(3, 0.0);
        kernel
            .apply(dist.as_ref(), full.as_mut(), Triangle::Full)
            .expect("shape");
        let mut lower = fill(3, 42.0);
        kernel
            .apply(dist.as_ref(), lower.as_mut(), Triangle::Lower)
            .expect("shape");
        assert_lower_close(lower.as_ref(), full.as_ref(), TOL);
        assert_close(lower[(0, 1)], 42.0, TOL);
    }

    #[test]
    fn grad_matches_finite_difference() {
        let kernel = ConstantKernel::from_log_constant(-0.2).expect("valid");
        let theta = kernel.log_constant();
        let h = 1e-6;
        let plus = ConstantKernel::from_log_constant(theta + h).expect("valid");
        let minus = ConstantKernel::from_log_constant(theta - h).expect("valid");
        let dist = dummy_dist(2);
        let mut kp = fill(2, 0.0);
        let mut km = fill(2, 0.0);
        let mut dk = fill(2, 0.0);
        plus.apply(dist.as_ref(), kp.as_mut(), Triangle::Full)
            .expect("shape");
        minus
            .apply(dist.as_ref(), km.as_mut(), Triangle::Full)
            .expect("shape");
        kernel
            .grad(dist.as_ref(), dk.as_mut(), 0, Triangle::Full)
            .expect("idx 0");
        let fd = (kp[(0, 1)] - km[(0, 1)]) / (2.0 * h);
        assert_close(dk[(0, 1)], fd, TOL);
        assert_close(dk[(0, 1)], kernel.constant(), TOL);
    }

    #[test]
    fn get_set_params_roundtrip() {
        let mut kernel = ConstantKernel::new(2.0).expect("valid");
        let mut params = [0.0];
        kernel.get_params(&mut params).expect("len 1");
        assert_close(params[0], 2.0_f64.ln(), TOL);
        params[0] = 0.5_f64.ln();
        kernel.set_params(&params).expect("len 1");
        assert_close(kernel.constant(), 0.5, TOL);
    }

    #[test]
    fn apply_cross_and_diag() {
        let kernel = ConstantKernel::new(3.0).expect("valid");
        let dist = mat![[0.0, 1.0], [4.0, 0.0]];
        let mut out = Mat::zeros(2, 2);
        kernel
            .apply_cross(dist.as_ref(), out.as_mut())
            .expect("rect");
        assert_close(out[(0, 1)], 3.0, TOL);
        let mut diag = [0.0, 0.0];
        kernel.fill_diag(&mut diag);
        assert_close(diag[0], 3.0, TOL);
    }

    #[test]
    fn rejects_non_positive() {
        assert!(matches!(
            ConstantKernel::new(0.0),
            Err(GprError::InvalidHyperparameter { .. })
        ));
        let kernel = ConstantKernel::new(1.0).expect("valid");
        let dist = dummy_dist(2);
        let mut dk = fill(2, 0.0);
        assert!(matches!(
            kernel.grad(dist.as_ref(), dk.as_mut(), 1, Triangle::Lower),
            Err(GprError::IndexOutOfRange { .. })
        ));
    }
}