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

631

632

633

634

635

636

637

638

639

640

641

642

643

644

645

646

647

648

649

650

651

652

653

654

655

656

657

658

659

660

661

662

663

664

665

666

667

668

669

670

671

672

673

674

675

676

677

678

679

680

681

682

683

684

685

686

687

688

689

690

691

692

693

694

695

696

697

698

699

700

701

702

703

704

705

706

707

708

709

710

711

712

713

714

715

716

717

718

719

720

721

722

723

724

725

726

727

728

729

730

731

732

733

734

735

736

737

738

739

740

741

742

743

744

745

746

747

748

749

750

751

752

753

754

755

756

757

758

759

760

761

762

763

764

765

766

767

768

769

770

771

772

773

774

775

776

777

778

779

780

781

782

783

784

785

786

787

788

789

790

791

792

793

794

795

796

797

798

799

800

801

802

803

804

805

806

807

808

809

810

811

812

813

814

815

816

817

818

819

820

821

822

823

824

825

826

827

828

829

830

831

832

833

834

835

836

837

838

839

840

841

842

843

844

845

846

847

848

849

850

851

852

853

854

855

856

857

858

859

860

861

862

863

864

865

866

867

868

869

870

871

872

873

874

875

876

877

878

879

880

881

882

883

884

885

886

887

888

889

890

891

892

893

894

895

896

897

898

899

900

901

902

903

904

905

906

907

908

909

910

911

912

913

914

915

916

917

918

919

920

921

922

923

924

925

926

927

928

929

930

931

932

933

934

935

936

937

938

939

940

941

942

943

944

945

946

947

948

949

950

951

952

953

954

955

956

957

958

959

960

961

962

963

964

965

966

967

968

969

970

971

972

973

974

975

976

977

978

979

980

981

982

983

984

985

986

987

988

989

990

991

992

993

994

995

996

997

998

999

1000

1001

1002

1003

1004

from __future__ import division, print_function, absolute_import 

 

import functools 

import operator 

 

import numpy as np 

from scipy._lib.six import string_types 

from scipy.linalg import (get_lapack_funcs, LinAlgError, 

cholesky_banded, cho_solve_banded) 

from . import _bspl 

from . import _fitpack_impl 

from . import _fitpack as _dierckx 

 

__all__ = ["BSpline", "make_interp_spline", "make_lsq_spline"] 

 

 

# copy-paste from interpolate.py 

def prod(x): 

"""Product of a list of numbers; ~40x faster vs np.prod for Python tuples""" 

if len(x) == 0: 

return 1 

return functools.reduce(operator.mul, x) 

 

 

def _get_dtype(dtype): 

"""Return np.complex128 for complex dtypes, np.float64 otherwise.""" 

if np.issubdtype(dtype, np.complexfloating): 

return np.complex_ 

else: 

return np.float_ 

 

 

def _as_float_array(x, check_finite=False): 

"""Convert the input into a C contiguous float array. 

 

NB: Upcasts half- and single-precision floats to double precision. 

""" 

x = np.ascontiguousarray(x) 

dtyp = _get_dtype(x.dtype) 

x = x.astype(dtyp, copy=False) 

if check_finite and not np.isfinite(x).all(): 

raise ValueError("Array must not contain infs or nans.") 

return x 

 

 

class BSpline(object): 

r"""Univariate spline in the B-spline basis. 

 

.. math:: 

 

S(x) = \sum_{j=0}^{n-1} c_j B_{j, k; t}(x) 

 

where :math:`B_{j, k; t}` are B-spline basis functions of degree `k` 

and knots `t`. 

 

Parameters 

---------- 

t : ndarray, shape (n+k+1,) 

knots 

c : ndarray, shape (>=n, ...) 

spline coefficients 

k : int 

B-spline order 

extrapolate : bool or 'periodic', optional 

whether to extrapolate beyond the base interval, ``t[k] .. t[n]``, 

or to return nans. 

If True, extrapolates the first and last polynomial pieces of b-spline 

functions active on the base interval. 

If 'periodic', periodic extrapolation is used. 

Default is True. 

axis : int, optional 

Interpolation axis. Default is zero. 

 

Attributes 

---------- 

t : ndarray 

knot vector 

c : ndarray 

spline coefficients 

k : int 

spline degree 

extrapolate : bool 

If True, extrapolates the first and last polynomial pieces of b-spline 

functions active on the base interval. 

axis : int 

Interpolation axis. 

tck : tuple 

A read-only equivalent of ``(self.t, self.c, self.k)`` 

 

Methods 

------- 

__call__ 

basis_element 

derivative 

antiderivative 

integrate 

construct_fast 

 

Notes 

----- 

B-spline basis elements are defined via 

 

.. math:: 

 

B_{i, 0}(x) = 1, \textrm{if $t_i \le x < t_{i+1}$, otherwise $0$,} 

 

B_{i, k}(x) = \frac{x - t_i}{t_{i+k} - t_i} B_{i, k-1}(x) 

+ \frac{t_{i+k+1} - x}{t_{i+k+1} - t_{i+1}} B_{i+1, k-1}(x) 

 

**Implementation details** 

 

- At least ``k+1`` coefficients are required for a spline of degree `k`, 

so that ``n >= k+1``. Additional coefficients, ``c[j]`` with 

``j > n``, are ignored. 

 

- B-spline basis elements of degree `k` form a partition of unity on the 

*base interval*, ``t[k] <= x <= t[n]``. 

 

 

Examples 

-------- 

 

Translating the recursive definition of B-splines into Python code, we have: 

 

>>> def B(x, k, i, t): 

... if k == 0: 

... return 1.0 if t[i] <= x < t[i+1] else 0.0 

... if t[i+k] == t[i]: 

... c1 = 0.0 

... else: 

... c1 = (x - t[i])/(t[i+k] - t[i]) * B(x, k-1, i, t) 

... if t[i+k+1] == t[i+1]: 

... c2 = 0.0 

... else: 

... c2 = (t[i+k+1] - x)/(t[i+k+1] - t[i+1]) * B(x, k-1, i+1, t) 

... return c1 + c2 

 

>>> def bspline(x, t, c, k): 

... n = len(t) - k - 1 

... assert (n >= k+1) and (len(c) >= n) 

... return sum(c[i] * B(x, k, i, t) for i in range(n)) 

 

Note that this is an inefficient (if straightforward) way to 

evaluate B-splines --- this spline class does it in an equivalent, 

but much more efficient way. 

 

Here we construct a quadratic spline function on the base interval 

``2 <= x <= 4`` and compare with the naive way of evaluating the spline: 

 

>>> from scipy.interpolate import BSpline 

>>> k = 2 

>>> t = [0, 1, 2, 3, 4, 5, 6] 

>>> c = [-1, 2, 0, -1] 

>>> spl = BSpline(t, c, k) 

>>> spl(2.5) 

array(1.375) 

>>> bspline(2.5, t, c, k) 

1.375 

 

Note that outside of the base interval results differ. This is because 

`BSpline` extrapolates the first and last polynomial pieces of b-spline 

functions active on the base interval. 

 

>>> import matplotlib.pyplot as plt 

>>> fig, ax = plt.subplots() 

>>> xx = np.linspace(1.5, 4.5, 50) 

>>> ax.plot(xx, [bspline(x, t, c ,k) for x in xx], 'r-', lw=3, label='naive') 

>>> ax.plot(xx, spl(xx), 'b-', lw=4, alpha=0.7, label='BSpline') 

>>> ax.grid(True) 

>>> ax.legend(loc='best') 

>>> plt.show() 

 

 

References 

---------- 

.. [1] Tom Lyche and Knut Morken, Spline methods, 

http://www.uio.no/studier/emner/matnat/ifi/INF-MAT5340/v05/undervisningsmateriale/ 

.. [2] Carl de Boor, A practical guide to splines, Springer, 2001. 

 

""" 

def __init__(self, t, c, k, extrapolate=True, axis=0): 

super(BSpline, self).__init__() 

 

self.k = int(k) 

self.c = np.asarray(c) 

self.t = np.ascontiguousarray(t, dtype=np.float64) 

 

if extrapolate == 'periodic': 

self.extrapolate = extrapolate 

else: 

self.extrapolate = bool(extrapolate) 

 

n = self.t.shape[0] - self.k - 1 

 

if not (0 <= axis < self.c.ndim): 

raise ValueError("%s must be between 0 and %s" % (axis, c.ndim)) 

 

self.axis = axis 

if axis != 0: 

# roll the interpolation axis to be the first one in self.c 

# More specifically, the target shape for self.c is (n, ...), 

# and axis !=0 means that we have c.shape (..., n, ...) 

# ^ 

# axis 

self.c = np.rollaxis(self.c, axis) 

 

if k < 0: 

raise ValueError("Spline order cannot be negative.") 

if int(k) != k: 

raise ValueError("Spline order must be integer.") 

if self.t.ndim != 1: 

raise ValueError("Knot vector must be one-dimensional.") 

if n < self.k + 1: 

raise ValueError("Need at least %d knots for degree %d" % 

(2*k + 2, k)) 

if (np.diff(self.t) < 0).any(): 

raise ValueError("Knots must be in a non-decreasing order.") 

if len(np.unique(self.t[k:n+1])) < 2: 

raise ValueError("Need at least two internal knots.") 

if not np.isfinite(self.t).all(): 

raise ValueError("Knots should not have nans or infs.") 

if self.c.ndim < 1: 

raise ValueError("Coefficients must be at least 1-dimensional.") 

if self.c.shape[0] < n: 

raise ValueError("Knots, coefficients and degree are inconsistent.") 

 

dt = _get_dtype(self.c.dtype) 

self.c = np.ascontiguousarray(self.c, dtype=dt) 

 

@classmethod 

def construct_fast(cls, t, c, k, extrapolate=True, axis=0): 

"""Construct a spline without making checks. 

 

Accepts same parameters as the regular constructor. Input arrays 

`t` and `c` must of correct shape and dtype. 

""" 

self = object.__new__(cls) 

self.t, self.c, self.k = t, c, k 

self.extrapolate = extrapolate 

self.axis = axis 

return self 

 

@property 

def tck(self): 

"""Equivalent to ``(self.t, self.c, self.k)`` (read-only). 

""" 

return self.t, self.c, self.k 

 

@classmethod 

def basis_element(cls, t, extrapolate=True): 

"""Return a B-spline basis element ``B(x | t[0], ..., t[k+1])``. 

 

Parameters 

---------- 

t : ndarray, shape (k+1,) 

internal knots 

extrapolate : bool or 'periodic', optional 

whether to extrapolate beyond the base interval, ``t[0] .. t[k+1]``, 

or to return nans. 

If 'periodic', periodic extrapolation is used. 

Default is True. 

 

Returns 

------- 

basis_element : callable 

A callable representing a B-spline basis element for the knot 

vector `t`. 

 

Notes 

----- 

The order of the b-spline, `k`, is inferred from the length of `t` as 

``len(t)-2``. The knot vector is constructed by appending and prepending 

``k+1`` elements to internal knots `t`. 

 

Examples 

-------- 

 

Construct a cubic b-spline: 

 

>>> from scipy.interpolate import BSpline 

>>> b = BSpline.basis_element([0, 1, 2, 3, 4]) 

>>> k = b.k 

>>> b.t[k:-k] 

array([ 0., 1., 2., 3., 4.]) 

>>> k 

3 

 

Construct a second order b-spline on ``[0, 1, 1, 2]``, and compare 

to its explicit form: 

 

>>> t = [-1, 0, 1, 1, 2] 

>>> b = BSpline.basis_element(t[1:]) 

>>> def f(x): 

... return np.where(x < 1, x*x, (2. - x)**2) 

 

>>> import matplotlib.pyplot as plt 

>>> fig, ax = plt.subplots() 

>>> x = np.linspace(0, 2, 51) 

>>> ax.plot(x, b(x), 'g', lw=3) 

>>> ax.plot(x, f(x), 'r', lw=8, alpha=0.4) 

>>> ax.grid(True) 

>>> plt.show() 

 

""" 

k = len(t) - 2 

t = _as_float_array(t) 

t = np.r_[(t[0]-1,) * k, t, (t[-1]+1,) * k] 

c = np.zeros_like(t) 

c[k] = 1. 

return cls.construct_fast(t, c, k, extrapolate) 

 

def __call__(self, x, nu=0, extrapolate=None): 

""" 

Evaluate a spline function. 

 

Parameters 

---------- 

x : array_like 

points to evaluate the spline at. 

nu: int, optional 

derivative to evaluate (default is 0). 

extrapolate : bool or 'periodic', optional 

whether to extrapolate based on the first and last intervals 

or return nans. If 'periodic', periodic extrapolation is used. 

Default is `self.extrapolate`. 

 

Returns 

------- 

y : array_like 

Shape is determined by replacing the interpolation axis 

in the coefficient array with the shape of `x`. 

 

""" 

if extrapolate is None: 

extrapolate = self.extrapolate 

x = np.asarray(x) 

x_shape, x_ndim = x.shape, x.ndim 

x = np.ascontiguousarray(x.ravel(), dtype=np.float_) 

 

# With periodic extrapolation we map x to the segment 

# [self.t[k], self.t[n]]. 

if extrapolate == 'periodic': 

n = self.t.size - self.k - 1 

x = self.t[self.k] + (x - self.t[self.k]) % (self.t[n] - 

self.t[self.k]) 

extrapolate = False 

 

out = np.empty((len(x), prod(self.c.shape[1:])), dtype=self.c.dtype) 

self._ensure_c_contiguous() 

self._evaluate(x, nu, extrapolate, out) 

out = out.reshape(x_shape + self.c.shape[1:]) 

if self.axis != 0: 

# transpose to move the calculated values to the interpolation axis 

l = list(range(out.ndim)) 

l = l[x_ndim:x_ndim+self.axis] + l[:x_ndim] + l[x_ndim+self.axis:] 

out = out.transpose(l) 

return out 

 

def _evaluate(self, xp, nu, extrapolate, out): 

_bspl.evaluate_spline(self.t, self.c.reshape(self.c.shape[0], -1), 

self.k, xp, nu, extrapolate, out) 

 

def _ensure_c_contiguous(self): 

""" 

c and t may be modified by the user. The Cython code expects 

that they are C contiguous. 

 

""" 

if not self.t.flags.c_contiguous: 

self.t = self.t.copy() 

if not self.c.flags.c_contiguous: 

self.c = self.c.copy() 

 

def derivative(self, nu=1): 

"""Return a b-spline representing the derivative. 

 

Parameters 

---------- 

nu : int, optional 

Derivative order. 

Default is 1. 

 

Returns 

------- 

b : BSpline object 

A new instance representing the derivative. 

 

See Also 

-------- 

splder, splantider 

 

""" 

c = self.c 

# pad the c array if needed 

ct = len(self.t) - len(c) 

if ct > 0: 

c = np.r_[c, np.zeros((ct,) + c.shape[1:])] 

tck = _fitpack_impl.splder((self.t, c, self.k), nu) 

return self.construct_fast(*tck, extrapolate=self.extrapolate, 

axis=self.axis) 

 

def antiderivative(self, nu=1): 

"""Return a b-spline representing the antiderivative. 

 

Parameters 

---------- 

nu : int, optional 

Antiderivative order. Default is 1. 

 

Returns 

------- 

b : BSpline object 

A new instance representing the antiderivative. 

 

Notes 

----- 

If antiderivative is computed and ``self.extrapolate='periodic'``, 

it will be set to False for the returned instance. This is done because 

the antiderivative is no longer periodic and its correct evaluation 

outside of the initially given x interval is difficult. 

 

See Also 

-------- 

splder, splantider 

 

""" 

c = self.c 

# pad the c array if needed 

ct = len(self.t) - len(c) 

if ct > 0: 

c = np.r_[c, np.zeros((ct,) + c.shape[1:])] 

tck = _fitpack_impl.splantider((self.t, c, self.k), nu) 

 

if self.extrapolate == 'periodic': 

extrapolate = False 

else: 

extrapolate = self.extrapolate 

 

return self.construct_fast(*tck, extrapolate=extrapolate, 

axis=self.axis) 

 

def integrate(self, a, b, extrapolate=None): 

"""Compute a definite integral of the spline. 

 

Parameters 

---------- 

a : float 

Lower limit of integration. 

b : float 

Upper limit of integration. 

extrapolate : bool or 'periodic', optional 

whether to extrapolate beyond the base interval, 

``t[k] .. t[-k-1]``, or take the spline to be zero outside of the 

base interval. If 'periodic', periodic extrapolation is used. 

If None (default), use `self.extrapolate`. 

 

Returns 

------- 

I : array_like 

Definite integral of the spline over the interval ``[a, b]``. 

 

Examples 

-------- 

Construct the linear spline ``x if x < 1 else 2 - x`` on the base 

interval :math:`[0, 2]`, and integrate it 

 

>>> from scipy.interpolate import BSpline 

>>> b = BSpline.basis_element([0, 1, 2]) 

>>> b.integrate(0, 1) 

array(0.5) 

 

If the integration limits are outside of the base interval, the result 

is controlled by the `extrapolate` parameter 

 

>>> b.integrate(-1, 1) 

array(0.0) 

>>> b.integrate(-1, 1, extrapolate=False) 

array(0.5) 

 

>>> import matplotlib.pyplot as plt 

>>> fig, ax = plt.subplots() 

>>> ax.grid(True) 

>>> ax.axvline(0, c='r', lw=5, alpha=0.5) # base interval 

>>> ax.axvline(2, c='r', lw=5, alpha=0.5) 

>>> xx = [-1, 1, 2] 

>>> ax.plot(xx, b(xx)) 

>>> plt.show() 

 

""" 

if extrapolate is None: 

extrapolate = self.extrapolate 

 

# Prepare self.t and self.c. 

self._ensure_c_contiguous() 

 

# Swap integration bounds if needed. 

sign = 1 

if b < a: 

a, b = b, a 

sign = -1 

n = self.t.size - self.k - 1 

 

if extrapolate != "periodic" and not extrapolate: 

# Shrink the integration interval, if needed. 

a = max(a, self.t[self.k]) 

b = min(b, self.t[n]) 

 

if self.c.ndim == 1: 

# Fast path: use FITPACK's routine 

# (cf _fitpack_impl.splint). 

t, c, k = self.tck 

integral, wrk = _dierckx._splint(t, c, k, a, b) 

return integral * sign 

 

out = np.empty((2, prod(self.c.shape[1:])), dtype=self.c.dtype) 

 

# Compute the antiderivative. 

c = self.c 

ct = len(self.t) - len(c) 

if ct > 0: 

c = np.r_[c, np.zeros((ct,) + c.shape[1:])] 

ta, ca, ka = _fitpack_impl.splantider((self.t, c, self.k), 1) 

 

if extrapolate == 'periodic': 

# Split the integral into the part over period (can be several 

# of them) and the remaining part. 

 

ts, te = self.t[self.k], self.t[n] 

period = te - ts 

interval = b - a 

n_periods, left = divmod(interval, period) 

 

if n_periods > 0: 

# Evaluate the difference of antiderivatives. 

x = np.asarray([ts, te], dtype=np.float_) 

_bspl.evaluate_spline(ta, ca.reshape(ca.shape[0], -1), 

ka, x, 0, False, out) 

integral = out[1] - out[0] 

integral *= n_periods 

else: 

integral = np.zeros((1, prod(self.c.shape[1:])), 

dtype=self.c.dtype) 

 

# Map a to [ts, te], b is always a + left. 

a = ts + (a - ts) % period 

b = a + left 

 

# If b <= te then we need to integrate over [a, b], otherwise 

# over [a, te] and from xs to what is remained. 

if b <= te: 

x = np.asarray([a, b], dtype=np.float_) 

_bspl.evaluate_spline(ta, ca.reshape(ca.shape[0], -1), 

ka, x, 0, False, out) 

integral += out[1] - out[0] 

else: 

x = np.asarray([a, te], dtype=np.float_) 

_bspl.evaluate_spline(ta, ca.reshape(ca.shape[0], -1), 

ka, x, 0, False, out) 

integral += out[1] - out[0] 

 

x = np.asarray([ts, ts + b - te], dtype=np.float_) 

_bspl.evaluate_spline(ta, ca.reshape(ca.shape[0], -1), 

ka, x, 0, False, out) 

integral += out[1] - out[0] 

else: 

# Evaluate the difference of antiderivatives. 

x = np.asarray([a, b], dtype=np.float_) 

_bspl.evaluate_spline(ta, ca.reshape(ca.shape[0], -1), 

ka, x, 0, extrapolate, out) 

integral = out[1] - out[0] 

 

integral *= sign 

return integral.reshape(ca.shape[1:]) 

 

 

################################# 

# Interpolating spline helpers # 

################################# 

 

def _not_a_knot(x, k): 

"""Given data x, construct the knot vector w/ not-a-knot BC. 

cf de Boor, XIII(12).""" 

x = np.asarray(x) 

if k % 2 != 1: 

raise ValueError("Odd degree for now only. Got %s." % k) 

 

m = (k - 1) // 2 

t = x[m+1:-m-1] 

t = np.r_[(x[0],)*(k+1), t, (x[-1],)*(k+1)] 

return t 

 

 

def _augknt(x, k): 

"""Construct a knot vector appropriate for the order-k interpolation.""" 

return np.r_[(x[0],)*k, x, (x[-1],)*k] 

 

 

def _convert_string_aliases(deriv, target_shape): 

if isinstance(deriv, string_types): 

if deriv == "clamped": 

deriv = [(1, np.zeros(target_shape))] 

elif deriv == "natural": 

deriv = [(2, np.zeros(target_shape))] 

else: 

raise ValueError("Unknown boundary condition : %s" % deriv) 

return deriv 

 

 

def make_interp_spline(x, y, k=3, t=None, bc_type=None, axis=0, 

check_finite=True): 

"""Compute the (coefficients of) interpolating B-spline. 

 

Parameters 

---------- 

x : array_like, shape (n,) 

Abscissas. 

y : array_like, shape (n, ...) 

Ordinates. 

k : int, optional 

B-spline degree. Default is cubic, k=3. 

t : array_like, shape (nt + k + 1,), optional. 

Knots. 

The number of knots needs to agree with the number of datapoints and 

the number of derivatives at the edges. Specifically, ``nt - n`` must 

equal ``len(deriv_l) + len(deriv_r)``. 

bc_type : 2-tuple or None 

Boundary conditions. 

Default is None, which means choosing the boundary conditions 

automatically. Otherwise, it must be a length-two tuple where the first 

element sets the boundary conditions at ``x[0]`` and the second 

element sets the boundary conditions at ``x[-1]``. Each of these must 

be an iterable of pairs ``(order, value)`` which gives the values of 

derivatives of specified orders at the given edge of the interpolation 

interval. 

Alternatively, the following string aliases are recognized: 

 

* ``"clamped"``: The first derivatives at the ends are zero. This is 

equivalent to ``bc_type=((1, 0.0), (1, 0.0))``. 

* ``"natural"``: The second derivatives at ends are zero. This is 

equivalent to ``bc_type=((2, 0.0), (2, 0.0))``. 

* ``"not-a-knot"`` (default): The first and second segments are the same 

polynomial. This is equivalent to having ``bc_type=None``. 

 

axis : int, optional 

Interpolation axis. Default is 0. 

check_finite : bool, optional 

Whether to check that the input arrays contain only finite numbers. 

Disabling may give a performance gain, but may result in problems 

(crashes, non-termination) if the inputs do contain infinities or NaNs. 

Default is True. 

 

Returns 

------- 

b : a BSpline object of the degree ``k`` and with knots ``t``. 

 

Examples 

-------- 

 

Use cubic interpolation on Chebyshev nodes: 

 

>>> def cheb_nodes(N): 

... jj = 2.*np.arange(N) + 1 

... x = np.cos(np.pi * jj / 2 / N)[::-1] 

... return x 

 

>>> x = cheb_nodes(20) 

>>> y = np.sqrt(1 - x**2) 

 

>>> from scipy.interpolate import BSpline, make_interp_spline 

>>> b = make_interp_spline(x, y) 

>>> np.allclose(b(x), y) 

True 

 

Note that the default is a cubic spline with a not-a-knot boundary condition 

 

>>> b.k 

3 

 

Here we use a 'natural' spline, with zero 2nd derivatives at edges: 

 

>>> l, r = [(2, 0.0)], [(2, 0.0)] 

>>> b_n = make_interp_spline(x, y, bc_type=(l, r)) # or, bc_type="natural" 

>>> np.allclose(b_n(x), y) 

True 

>>> x0, x1 = x[0], x[-1] 

>>> np.allclose([b_n(x0, 2), b_n(x1, 2)], [0, 0]) 

True 

 

Interpolation of parametric curves is also supported. As an example, we 

compute a discretization of a snail curve in polar coordinates 

 

>>> phi = np.linspace(0, 2.*np.pi, 40) 

>>> r = 0.3 + np.cos(phi) 

>>> x, y = r*np.cos(phi), r*np.sin(phi) # convert to Cartesian coordinates 

 

Build an interpolating curve, parameterizing it by the angle 

 

>>> from scipy.interpolate import make_interp_spline 

>>> spl = make_interp_spline(phi, np.c_[x, y]) 

 

Evaluate the interpolant on a finer grid (note that we transpose the result 

to unpack it into a pair of x- and y-arrays) 

 

>>> phi_new = np.linspace(0, 2.*np.pi, 100) 

>>> x_new, y_new = spl(phi_new).T 

 

Plot the result 

 

>>> import matplotlib.pyplot as plt 

>>> plt.plot(x, y, 'o') 

>>> plt.plot(x_new, y_new, '-') 

>>> plt.show() 

 

See Also 

-------- 

BSpline : base class representing the B-spline objects 

CubicSpline : a cubic spline in the polynomial basis 

make_lsq_spline : a similar factory function for spline fitting 

UnivariateSpline : a wrapper over FITPACK spline fitting routines 

splrep : a wrapper over FITPACK spline fitting routines 

 

""" 

# convert string aliases for the boundary conditions 

if bc_type is None or bc_type == 'not-a-knot': 

deriv_l, deriv_r = None, None 

elif isinstance(bc_type, string_types): 

deriv_l, deriv_r = bc_type, bc_type 

else: 

deriv_l, deriv_r = bc_type 

 

# special-case k=0 right away 

if k == 0: 

if any(_ is not None for _ in (t, deriv_l, deriv_r)): 

raise ValueError("Too much info for k=0: t and bc_type can only " 

"be None.") 

x = _as_float_array(x, check_finite) 

t = np.r_[x, x[-1]] 

c = np.asarray(y) 

c = np.ascontiguousarray(c, dtype=_get_dtype(c.dtype)) 

return BSpline.construct_fast(t, c, k, axis=axis) 

 

# special-case k=1 (e.g., Lyche and Morken, Eq.(2.16)) 

if k == 1 and t is None: 

if not (deriv_l is None and deriv_r is None): 

raise ValueError("Too much info for k=1: bc_type can only be None.") 

x = _as_float_array(x, check_finite) 

t = np.r_[x[0], x, x[-1]] 

c = np.asarray(y) 

c = np.ascontiguousarray(c, dtype=_get_dtype(c.dtype)) 

return BSpline.construct_fast(t, c, k, axis=axis) 

 

x = _as_float_array(x, check_finite) 

y = _as_float_array(y, check_finite) 

k = int(k) 

 

# come up with a sensible knot vector, if needed 

if t is None: 

if deriv_l is None and deriv_r is None: 

if k == 2: 

# OK, it's a bit ad hoc: Greville sites + omit 

# 2nd and 2nd-to-last points, a la not-a-knot 

t = (x[1:] + x[:-1]) / 2. 

t = np.r_[(x[0],)*(k+1), 

t[1:-1], 

(x[-1],)*(k+1)] 

else: 

t = _not_a_knot(x, k) 

else: 

t = _augknt(x, k) 

 

t = _as_float_array(t, check_finite) 

 

axis = axis % y.ndim 

y = np.rollaxis(y, axis) # now internally interp axis is zero 

 

if x.ndim != 1 or np.any(x[1:] <= x[:-1]): 

raise ValueError("Expect x to be a 1-D sorted array_like.") 

if k < 0: 

raise ValueError("Expect non-negative k.") 

if t.ndim != 1 or np.any(t[1:] < t[:-1]): 

raise ValueError("Expect t to be a 1-D sorted array_like.") 

if x.size != y.shape[0]: 

raise ValueError('x and y are incompatible.') 

if t.size < x.size + k + 1: 

raise ValueError('Got %d knots, need at least %d.' % 

(t.size, x.size + k + 1)) 

if (x[0] < t[k]) or (x[-1] > t[-k]): 

raise ValueError('Out of bounds w/ x = %s.' % x) 

 

# Here : deriv_l, r = [(nu, value), ...] 

deriv_l = _convert_string_aliases(deriv_l, y.shape[1:]) 

if deriv_l is not None: 

deriv_l_ords, deriv_l_vals = zip(*deriv_l) 

else: 

deriv_l_ords, deriv_l_vals = [], [] 

deriv_l_ords, deriv_l_vals = np.atleast_1d(deriv_l_ords, deriv_l_vals) 

nleft = deriv_l_ords.shape[0] 

 

deriv_r = _convert_string_aliases(deriv_r, y.shape[1:]) 

if deriv_r is not None: 

deriv_r_ords, deriv_r_vals = zip(*deriv_r) 

else: 

deriv_r_ords, deriv_r_vals = [], [] 

deriv_r_ords, deriv_r_vals = np.atleast_1d(deriv_r_ords, deriv_r_vals) 

nright = deriv_r_ords.shape[0] 

 

# have `n` conditions for `nt` coefficients; need nt-n derivatives 

n = x.size 

nt = t.size - k - 1 

 

if nt - n != nleft + nright: 

raise ValueError("number of derivatives at boundaries.") 

 

# set up the LHS: the collocation matrix + derivatives at boundaries 

kl = ku = k 

ab = np.zeros((2*kl + ku + 1, nt), dtype=np.float_, order='F') 

_bspl._colloc(x, t, k, ab, offset=nleft) 

if nleft > 0: 

_bspl._handle_lhs_derivatives(t, k, x[0], ab, kl, ku, deriv_l_ords) 

if nright > 0: 

_bspl._handle_lhs_derivatives(t, k, x[-1], ab, kl, ku, deriv_r_ords, 

offset=nt-nright) 

 

# set up the RHS: values to interpolate (+ derivative values, if any) 

extradim = prod(y.shape[1:]) 

rhs = np.empty((nt, extradim), dtype=y.dtype) 

if nleft > 0: 

rhs[:nleft] = deriv_l_vals.reshape(-1, extradim) 

rhs[nleft:nt - nright] = y.reshape(-1, extradim) 

if nright > 0: 

rhs[nt - nright:] = deriv_r_vals.reshape(-1, extradim) 

 

# solve Ab @ x = rhs; this is the relevant part of linalg.solve_banded 

if check_finite: 

ab, rhs = map(np.asarray_chkfinite, (ab, rhs)) 

gbsv, = get_lapack_funcs(('gbsv',), (ab, rhs)) 

lu, piv, c, info = gbsv(kl, ku, ab, rhs, 

overwrite_ab=True, overwrite_b=True) 

 

if info > 0: 

raise LinAlgError("Collocation matix is singular.") 

elif info < 0: 

raise ValueError('illegal value in %d-th argument of internal gbsv' % -info) 

 

c = np.ascontiguousarray(c.reshape((nt,) + y.shape[1:])) 

return BSpline.construct_fast(t, c, k, axis=axis) 

 

 

def make_lsq_spline(x, y, t, k=3, w=None, axis=0, check_finite=True): 

r"""Compute the (coefficients of) an LSQ B-spline. 

 

The result is a linear combination 

 

.. math:: 

 

S(x) = \sum_j c_j B_j(x; t) 

 

of the B-spline basis elements, :math:`B_j(x; t)`, which minimizes 

 

.. math:: 

 

\sum_{j} \left( w_j \times (S(x_j) - y_j) \right)^2 

 

Parameters 

---------- 

x : array_like, shape (m,) 

Abscissas. 

y : array_like, shape (m, ...) 

Ordinates. 

t : array_like, shape (n + k + 1,). 

Knots. 

Knots and data points must satisfy Schoenberg-Whitney conditions. 

k : int, optional 

B-spline degree. Default is cubic, k=3. 

w : array_like, shape (n,), optional 

Weights for spline fitting. Must be positive. If ``None``, 

then weights are all equal. 

Default is ``None``. 

axis : int, optional 

Interpolation axis. Default is zero. 

check_finite : bool, optional 

Whether to check that the input arrays contain only finite numbers. 

Disabling may give a performance gain, but may result in problems 

(crashes, non-termination) if the inputs do contain infinities or NaNs. 

Default is True. 

 

Returns 

------- 

b : a BSpline object of the degree `k` with knots `t`. 

 

Notes 

----- 

 

The number of data points must be larger than the spline degree `k`. 

 

Knots `t` must satisfy the Schoenberg-Whitney conditions, 

i.e., there must be a subset of data points ``x[j]`` such that 

``t[j] < x[j] < t[j+k+1]``, for ``j=0, 1,...,n-k-2``. 

 

Examples 

-------- 

Generate some noisy data: 

 

>>> x = np.linspace(-3, 3, 50) 

>>> y = np.exp(-x**2) + 0.1 * np.random.randn(50) 

 

Now fit a smoothing cubic spline with a pre-defined internal knots. 

Here we make the knot vector (k+1)-regular by adding boundary knots: 

 

>>> from scipy.interpolate import make_lsq_spline, BSpline 

>>> t = [-1, 0, 1] 

>>> k = 3 

>>> t = np.r_[(x[0],)*(k+1), 

... t, 

... (x[-1],)*(k+1)] 

>>> spl = make_lsq_spline(x, y, t, k) 

 

For comparison, we also construct an interpolating spline for the same 

set of data: 

 

>>> from scipy.interpolate import make_interp_spline 

>>> spl_i = make_interp_spline(x, y) 

 

Plot both: 

 

>>> import matplotlib.pyplot as plt 

>>> xs = np.linspace(-3, 3, 100) 

>>> plt.plot(x, y, 'ro', ms=5) 

>>> plt.plot(xs, spl(xs), 'g-', lw=3, label='LSQ spline') 

>>> plt.plot(xs, spl_i(xs), 'b-', lw=3, alpha=0.7, label='interp spline') 

>>> plt.legend(loc='best') 

>>> plt.show() 

 

**NaN handling**: If the input arrays contain ``nan`` values, the result is 

not useful since the underlying spline fitting routines cannot deal with 

``nan``. A workaround is to use zero weights for not-a-number data points: 

 

>>> y[8] = np.nan 

>>> w = np.isnan(y) 

>>> y[w] = 0. 

>>> tck = make_lsq_spline(x, y, t, w=~w) 

 

Notice the need to replace a ``nan`` by a numerical value (precise value 

does not matter as long as the corresponding weight is zero.) 

 

See Also 

-------- 

BSpline : base class representing the B-spline objects 

make_interp_spline : a similar factory function for interpolating splines 

LSQUnivariateSpline : a FITPACK-based spline fitting routine 

splrep : a FITPACK-based fitting routine 

 

""" 

x = _as_float_array(x, check_finite) 

y = _as_float_array(y, check_finite) 

t = _as_float_array(t, check_finite) 

if w is not None: 

w = _as_float_array(w, check_finite) 

else: 

w = np.ones_like(x) 

k = int(k) 

 

axis = axis % y.ndim 

y = np.rollaxis(y, axis) # now internally interp axis is zero 

 

if x.ndim != 1 or np.any(x[1:] - x[:-1] <= 0): 

raise ValueError("Expect x to be a 1-D sorted array_like.") 

if x.shape[0] < k+1: 

raise ValueError("Need more x points.") 

if k < 0: 

raise ValueError("Expect non-negative k.") 

if t.ndim != 1 or np.any(t[1:] - t[:-1] < 0): 

raise ValueError("Expect t to be a 1-D sorted array_like.") 

if x.size != y.shape[0]: 

raise ValueError('x & y are incompatible.') 

if k > 0 and np.any((x < t[k]) | (x > t[-k])): 

raise ValueError('Out of bounds w/ x = %s.' % x) 

if x.size != w.size: 

raise ValueError('Incompatible weights.') 

 

# number of coefficients 

n = t.size - k - 1 

 

# construct A.T @ A and rhs with A the collocation matrix, and 

# rhs = A.T @ y for solving the LSQ problem ``A.T @ A @ c = A.T @ y`` 

lower = True 

extradim = prod(y.shape[1:]) 

ab = np.zeros((k+1, n), dtype=np.float_, order='F') 

rhs = np.zeros((n, extradim), dtype=y.dtype, order='F') 

_bspl._norm_eq_lsq(x, t, k, 

y.reshape(-1, extradim), 

w, 

ab, rhs) 

rhs = rhs.reshape((n,) + y.shape[1:]) 

 

# have observation matrix & rhs, can solve the LSQ problem 

cho_decomp = cholesky_banded(ab, overwrite_ab=True, lower=lower, 

check_finite=check_finite) 

c = cho_solve_banded((cho_decomp, lower), rhs, overwrite_b=True, 

check_finite=check_finite) 

 

c = np.ascontiguousarray(c) 

return BSpline.construct_fast(t, c, k, axis=axis)