chebax 0.1.0.dev0__py3-none-any.whl

chebax/__init__.py

@@ -0,0 +1,11 @@
+"""chebax: differentiable Chebyshev approximants for special functions.
+
+Full float64 accuracy needs jax's x64 mode (jax.config.update("jax_enable_x64", True)).
+"""
+
+from chebax._src.generate import fit
+from chebax._src.recipes.besselj import besselj, besselj_dnu
+from chebax._src.series import ChebSeries, PiecewiseCheb
+
+__version__ = "0.1.0.dev0"
+__all__ = ["ChebSeries", "PiecewiseCheb", "besselj", "besselj_dnu", "fit", "__version__"]

chebax/_src/algorithms.py

@@ -0,0 +1,74 @@
+"""The textbook algorithms, in plain numpy, on [-1, 1].
+
+Coefficient convention is numpy.polynomial's: plain c[0], lowest order first,
+p(t) = sum(c[k] * T_k(t)). These are the reference implementations; the jax
+runtime in series.py must agree with them (tested against numpy.polynomial).
+Domain handling lives in the series classes, not here.
+"""
+
+import numpy as np
+
+
+def chebpts(n):
+    """First-kind Chebyshev points cos(pi*(2i+1)/(2n)), decreasing from ~1 to ~-1."""
+    i = np.arange(n)
+    return np.cos(np.pi * (2 * i + 1) / (2 * n))
+
+
+def chebfit_dct(samples):
+    """Interpolation coefficients from samples taken at chebpts(len(samples)).
+
+    Discrete orthogonality of cosines on the first-kind points; O(n^2), which
+    is fine at generator degrees.
+    """
+    n = len(samples)
+    i = np.arange(n)
+    j = np.arange(n)[:, None]
+    # reduce the angle with exact integer arithmetic; the raw product reaches
+    # hundreds of radians at large n and costs ~angle*eps of cosine accuracy
+    weights = np.cos(np.pi * ((j * (2 * i + 1)) % (4 * n)) / (2 * n))
+    c = (2.0 / n) * (weights @ np.asarray(samples, dtype=float))
+    c[0] /= 2
+    return c
+
+
+def chebval(t, c):
+    """Clenshaw evaluation of a plain Chebyshev series."""
+    t = np.asarray(t, dtype=float)
+    b1 = np.zeros_like(t)
+    b2 = np.zeros_like(t)
+    for ck in np.asarray(c, dtype=float)[:0:-1]:
+        b1, b2 = 2 * t * b1 - b2 + ck, b1
+    return t * b1 - b2 + c[0]
+
+
+def chebder(c):
+    """Coefficients of the derivative (numpy.polynomial's exact recurrence)."""
+    c = np.array(c, dtype=float)
+    n = c.size - 1
+    if n == 0:
+        return np.zeros(1)
+    der = np.zeros(n)
+    for j in range(n, 2, -1):
+        der[j - 1] = 2 * j * c[j]
+        c[j - 2] += j * c[j] / (j - 2)
+    if n > 1:
+        der[1] = 4 * c[2]
+    der[0] = c[1]
+    return der
+
+
+def chebint(c):
+    """Coefficients of the antiderivative that vanishes at t=0
+    (numpy.polynomial's k=0, lbnd=0 convention)."""
+    c = np.asarray(c, dtype=float)
+    n = c.size
+    out = np.zeros(n + 1)
+    out[1] = c[0]
+    if n > 1:
+        out[2] = c[1] / 4
+    for j in range(2, n):
+        out[j + 1] = c[j] / (2 * (j + 1))
+        out[j - 1] -= c[j] / (2 * (j - 1))
+    out[0] -= chebval(0.0, out)
+    return out

chebax/_src/generate.py

@@ -0,0 +1,106 @@
+"""Build-time fitting: sample at Chebyshev points, DCT to coefficients.
+
+The float64 path uses numpy. Passing dps switches sampling and the DCT to
+mpmath at that precision; mpmath is imported inside that branch only (the
+runtime-imports rule in CLAUDE.md). Adaptive fitting doubles the node count
+until the last two coefficients fall below tol relative to the largest, then
+truncates the converged tail.
+"""
+
+import numpy as np
+
+from chebax._src import algorithms
+from chebax._src.series import ChebSeries, PiecewiseCheb
+
+# ~2*eps: the practical floor of the float64 build path. Use dps= to go below.
+_DEFAULT_TOL = 5e-16
+
+
+def fit(f, domain=(-1.0, 1.0), *, deg=None, tol=None, breaks=None, dps=None, max_deg=256):
+    """Chebyshev interpolant of f, as a ChebSeries (or PiecewiseCheb if breaks given).
+
+    f:       callable on the domain. Vectorized over numpy arrays if possible;
+             with dps set it must accept mpmath floats instead.
+    domain:  (a, b). Ignored when breaks is given.
+    deg:     fixed degree; otherwise the degree adapts until tol is met.
+    tol:     target for the relative coefficient tail (default 5e-16).
+    breaks:  strictly increasing knots, endpoints included; one series per
+             segment, each fit independently, padded to a common degree.
+    dps:     mpmath working precision for sampling and the DCT. Derivatives of
+             a float64-built fit sit at a ~deg^2 * eps floor (differentiation
+             amplifies sample noise); build with dps when that matters.
+    max_deg: adaptive-path cap; exceeded means f is not smooth enough on the
+             interval (split the domain) or genuinely needs a higher degree.
+    """
+    if tol is None:
+        tol = _DEFAULT_TOL
+    if breaks is not None:
+        br = [float(t) for t in breaks]
+        cs = [_fit_interval(f, br[i], br[i + 1], deg, tol, max_deg, dps)
+              for i in range(len(br) - 1)]
+        n = max(c.size for c in cs)
+        coef = np.zeros((len(cs), n))
+        for i, c in enumerate(cs):
+            coef[i, : c.size] = c
+        return PiecewiseCheb(coef, br)
+    a, b = float(domain[0]), float(domain[1])
+    return ChebSeries(_fit_interval(f, a, b, deg, tol, max_deg, dps), (a, b))
+
+
+def _fit_interval(f, a, b, deg, tol, max_deg, dps):
+    dct = _dct_mp(f, a, b, dps) if dps is not None else _dct_np(f, a, b)
+    if deg is not None:
+        return dct(deg + 1)
+    return _adaptive(dct, tol, max_deg, (a, b))
+
+
+def _adaptive(dct, tol, max_deg, interval):
+    n = 17
+    while True:
+        c = dct(n)
+        scale = np.max(np.abs(c))
+        if scale == 0.0:
+            return np.zeros(1)
+        if np.max(np.abs(c[-2:])) <= tol * scale:
+            keep = np.nonzero(np.abs(c) > tol * scale)[0]
+            return c[: keep[-1] + 1] if keep.size else c[:1]
+        if n > max_deg:
+            raise ValueError(
+                f"no convergence to tol={tol:g} by degree {max_deg} on {interval}; "
+                "f may not be smooth there (split the domain) or needs a higher max_deg")
+        n = 2 * n - 1
+
+
+def _dct_np(f, a, b):
+    def dct(n):
+        xk = a + (b - a) * (algorithms.chebpts(n) + 1) / 2
+        return algorithms.chebfit_dct(_sample(f, xk))
+    return dct
+
+
+def _sample(f, xk):
+    # scalar-only callables (wrapped special functions) are common; fall back.
+    try:
+        y = np.asarray(f(xk), dtype=float)
+        if y.shape == xk.shape:
+            return y
+    except (TypeError, ValueError):
+        pass
+    return np.array([float(f(x)) for x in xk])
+
+
+def _dct_mp(f, a, b, dps):
+    import mpmath as mp
+
+    def dct(n):
+        with mp.workdps(dps):
+            mid = (mp.mpf(b) + mp.mpf(a)) / 2
+            half = (mp.mpf(b) - mp.mpf(a)) / 2
+            fk = [f(mid + half * mp.cos(mp.pi * (2 * i + 1) / (2 * n))) for i in range(n)]
+            c = []
+            for j in range(n):
+                s = mp.fsum(fk[i] * mp.cos(mp.pi * j * (2 * i + 1) / (2 * n)) for i in range(n))
+                c.append(2 * s / n)
+            c[0] /= 2
+            return np.array([float(x) for x in c])
+    return dct

chebax/_src/recipes/besselj.py

@@ -0,0 +1,138 @@
+"""J_v for real v in [0, 10] on x in [0, 8], from the baked nu-table.
+
+J_v(x) = (x/2)^v / Gamma(v+1) * g_v(x^2), where g_v(z) = 0F1(; v+1; -z/4) is
+entire, so g_v is one short Chebyshev series in z (the ../../bessel evidence,
+PROJECT.md section 2). The baked table stores each z-coefficient as a
+Chebyshev series in v; instantiation is 25 numpy Clenshaws, once per order,
+cached. dJ/dx comes from jax.grad through the series custom_jvp; dJ/dv comes
+from chebder along the v axis of the table (besselj_dnu).
+
+Out of range is the caller's problem in v1: x > 8 evaluates the polynomial
+outside its interval and diverges (tails are milestone M3); v outside [0, 10]
+raises at instantiation.
+"""
+
+import functools
+import math
+
+import jax
+import jax.numpy as jnp
+import numpy as np
+
+from chebax._src import algorithms
+from chebax._src.recipes import besselj_table as _tab
+from chebax._src.series import ChebSeries
+
+
+def _digamma(z):
+    """float64 psi(z) for z >= 1: recurrence up to z >= 15, then the Bernoulli
+    series through z^-12 (A&S 6.3.5 / 6.3.18). Truncation < 1e-16 there."""
+    acc = 0.0
+    while z < 15.0:
+        acc -= 1.0 / z
+        z += 1.0
+    zi = 1.0 / (z * z)
+    tail = 1 / 12 - zi * (1 / 120 - zi * (1 / 252 - zi * (1 / 240 - zi * (1 / 132 - zi * (691 / 32760)))))
+    return acc + math.log(z) - 0.5 / z - zi * tail
+
+
+def _coef_at(v):
+    """The z-coefficient vector at order v: Clenshaw in v across table rows."""
+    t = 2.0 * v / _tab.VMAX - 1.0
+    return np.array([algorithms.chebval(t, row) for row in _tab.TABLE])
+
+
+def _check_order(v):
+    v = float(v)
+    if not 0.0 <= v <= _tab.VMAX:
+        raise ValueError(f"besselj table covers v in [0, {_tab.VMAX}], got {v}")
+    return v
+
+
+@jax.tree_util.register_pytree_node_class
+class BesselJ:
+    """Callable J_v on [0, 8]. Build with besselj(v)."""
+
+    def __init__(self, v, g):
+        self.v = float(v)
+        self.g = g
+        self._inv_gamma = 1.0 / math.gamma(self.v + 1.0)
+
+    def __call__(self, x):
+        x = jnp.asarray(x)
+        return self._inv_gamma * jnp.power(x / 2, self.v) * self.g(x * x)
+
+    def astype(self, dtype):
+        return BesselJ(self.v, self.g.astype(dtype))
+
+    def __repr__(self):
+        return f"BesselJ(v={self.v}, dtype={self.g.coef.dtype})"
+
+    def tree_flatten(self):
+        return (self.g,), self.v
+
+    @classmethod
+    def tree_unflatten(cls, v, children):
+        obj = object.__new__(cls)
+        obj.v = v
+        obj.g = children[0]
+        obj._inv_gamma = 1.0 / math.gamma(v + 1.0)
+        return obj
+
+
+@jax.tree_util.register_pytree_node_class
+class BesselJdnu:
+    """Callable dJ_v/dv on (0, 8]. Build with besselj_dnu(v).
+
+    dJ/dv = pref * [(log(x/2) - psi(v+1)) g(z) + g_v(z)], g_v from chebder
+    along the v axis. The log makes x = 0 nan; for v = 0 that singularity
+    is real."""
+
+    def __init__(self, v, g, gnu):
+        self.v = float(v)
+        self.g = g
+        self.gnu = gnu
+        self._inv_gamma = 1.0 / math.gamma(self.v + 1.0)
+        self._psi = _digamma(self.v + 1.0)
+
+    def __call__(self, x):
+        x = jnp.asarray(x)
+        z = x * x
+        pref = self._inv_gamma * jnp.power(x / 2, self.v)
+        return pref * ((jnp.log(x / 2) - self._psi) * self.g(z) + self.gnu(z))
+
+    def astype(self, dtype):
+        return BesselJdnu(self.v, self.g.astype(dtype), self.gnu.astype(dtype))
+
+    def __repr__(self):
+        return f"BesselJdnu(v={self.v}, dtype={self.g.coef.dtype})"
+
+    def tree_flatten(self):
+        return (self.g, self.gnu), self.v
+
+    @classmethod
+    def tree_unflatten(cls, v, children):
+        obj = object.__new__(cls)
+        obj.v = v
+        obj.g, obj.gnu = children
+        obj._inv_gamma = 1.0 / math.gamma(v + 1.0)
+        obj._psi = _digamma(v + 1.0)
+        return obj
+
+
+@functools.lru_cache(maxsize=None)
+def besselj(v):
+    """J_v on [0, 8] for real v in [0, 10]. Cached per order; no mpmath."""
+    v = _check_order(v)
+    return BesselJ(v, ChebSeries(_coef_at(v), (0.0, _tab.ZMAX)))
+
+
+@functools.lru_cache(maxsize=None)
+def besselj_dnu(v):
+    """dJ_v/dv on (0, 8] for real v in [0, 10] (the order gradient)."""
+    v = _check_order(v)
+    t = 2.0 * v / _tab.VMAX - 1.0
+    gnu = np.array([algorithms.chebval(t, algorithms.chebder(row)) for row in _tab.TABLE])
+    gnu *= 2.0 / _tab.VMAX
+    return BesselJdnu(v, ChebSeries(_coef_at(v), (0.0, _tab.ZMAX)),
+                      ChebSeries(gnu, (0.0, _tab.ZMAX)))

chebax/_src/recipes/besselj_gen.py

@@ -0,0 +1,76 @@
+"""Generator for the besselj nu-table. Build-time only; imports mpmath inside.
+
+The table is the construction measured in experiments/02: fit the degree-24
+Chebyshev coefficients of g_v(z) = 0F1(; v+1; -z/4) on z in [0, 64] at 64
+Chebyshev nodes in v over [0, 10], then fit each coefficient as a Chebyshev
+series in v. BOTH stages stay in mpmath and round to float64 only at the end:
+rounding between the stages would put eps-level sample noise under the second
+fit, and differentiation along v would amplify it by ~deg_v^2 (the M1 floor).
+
+Regenerate with:  python -m chebax._src.recipes.besselj_gen
+The output must reproduce the checked-in table bit for bit (tested).
+"""
+
+import numpy as np
+
+NZ = 25      # z-degree 24, covers the worst case (v=0) with margin
+NV = 64      # v-nodes
+ZMAX = 64.0  # z = x^2, x in [0, 8]
+VMAX = 10.0
+DPS = 40
+
+
+def generate_table():
+    import mpmath as mp
+
+    with mp.workdps(DPS):
+        zmax = mp.mpf(ZMAX)
+        vmax = mp.mpf(VMAX)
+
+        def dct(samples):
+            n = len(samples)
+            c = []
+            for j in range(n):
+                s = mp.fsum(samples[i] * mp.cos(mp.pi * j * (2 * i + 1) / (2 * n))
+                            for i in range(n))
+                c.append(2 * s / n)
+            c[0] /= 2
+            return c
+
+        tz = [mp.cos(mp.pi * (2 * i + 1) / (2 * NZ)) for i in range(NZ)]
+        vnodes = [(mp.cos(mp.pi * (2 * i + 1) / (2 * NV)) + 1) / 2 * vmax for i in range(NV)]
+        rows = [dct([mp.hyp0f1(v + 1, -((t + 1) / 2 * zmax) / 4) for t in tz]) for v in vnodes]
+        table = [dct([rows[i][k] for i in range(NV)]) for k in range(NZ)]
+    return np.array([[float(x) for x in row] for row in table])
+
+
+def main():
+    import pathlib
+
+    import mpmath
+
+    table = generate_table()
+    lines = [
+        '"""Baked besselj nu-table. Generated by chebax._src.recipes.besselj_gen;',
+        'do not edit. Regenerate with:  python -m chebax._src.recipes.besselj_gen',
+        '"""',
+        "",
+        "import numpy as np",
+        "",
+        f"META = {{'nz': {NZ}, 'nv': {NV}, 'zmax': {ZMAX!r}, 'vmax': {VMAX!r}, "
+        f"'dps': {DPS}, 'mpmath': {mpmath.__version__!r}}}",
+        f"ZMAX = {ZMAX!r}",
+        f"VMAX = {VMAX!r}",
+        "",
+        "TABLE = np.array([",
+    ]
+    for row in table:
+        lines.append("    [" + ", ".join(repr(x) for x in row) + "],")
+    lines.append("])")
+    out = pathlib.Path(__file__).with_name("besselj_table.py")
+    out.write_text("\n".join(lines) + "\n")
+    print(f"wrote {out} ({table.shape[0]}x{table.shape[1]})")
+
+
+if __name__ == "__main__":
+    main()

chebax/_src/recipes/besselj_table.py

@@ -0,0 +1,37 @@
+"""Baked besselj nu-table. Generated by chebax._src.recipes.besselj_gen;
+do not edit. Regenerate with:  python -m chebax._src.recipes.besselj_gen
+"""
+
+import numpy as np
+
+META = {'nz': 25, 'nv': 64, 'zmax': 64.0, 'vmax': 10.0, 'dps': 40, 'mpmath': '1.4.0'}
+ZMAX = 64.0
+VMAX = 10.0
+
+TABLE = np.array([
+    [np.float64(0.3472316049235684), np.float64(0.20720848881772033), np.float64(-0.009396033994909556), np.float64(-0.012807849745354835), np.float64(0.008498663563727726), np.float64(-0.0038718580229806696), np.float64(0.0014745831630515842), np.float64(-0.0004689757458182797), np.float64(0.00010230159988173848), np.float64(1.0175341704064214e-05), np.float64(-3.2961364372376345e-05), np.float64(2.9131613650508137e-05), np.float64(-2.035877742827867e-05), np.float64(1.2870666269727138e-05), np.float64(-7.707226734199128e-06), np.float64(4.4655184091080755e-06), np.float64(-2.532257759004389e-06), np.float64(1.4150530342550788e-06), np.float64(-7.82618111481341e-07), np.float64(4.2962749583131126e-07), np.float64(-2.345639705080128e-07), np.float64(1.2754592234436162e-07), np.float64(-6.9142790243786e-08), np.float64(3.739588888596302e-08), np.float64(-2.018998753190061e-08), np.float64(1.0885849063045356e-08), np.float64(-5.863241558376156e-09), np.float64(3.1554791018305487e-09), np.float64(-1.697162858026686e-09), np.float64(9.123738658634458e-10), np.float64(-4.902976841216041e-10), np.float64(2.634028974686074e-10), np.float64(-1.414760301481393e-10), np.float64(7.597460720311407e-11), np.float64(-4.079380312439558e-11), np.float64(2.1901465858058404e-11), np.float64(-1.1757518012780338e-11), np.float64(6.311456347476991e-12), np.float64(-3.3878268110105655e-12), np.float64(1.8184249026263132e-12), np.float64(-9.760136513484117e-13), np.float64(5.238485390468316e-13), np.float64(-2.811559170928882e-13), np.float64(1.508975475344413e-13), np.float64(-8.098638593602279e-14), np.float64(4.3464816624318874e-14), np.float64(-2.332708921173419e-14), np.float64(1.2519321582100685e-14), np.float64(-6.718914569584727e-15), np.float64(3.60591875785447e-15), np.float64(-1.9352254794953177e-15), np.float64(1.038595094177753e-15), np.float64(-5.573912761163496e-16), np.float64(2.9913914880882207e-16), np.float64(-1.6054065584839236e-16), np.float64(8.615757426734226e-17), np.float64(-4.623716482603899e-17), np.float64(2.48114710674498e-17), np.float64(-1.3310302602605188e-17), np.float64(7.133228651036784e-18), np.float64(-3.809435014528638e-18), np.float64(2.009410704308605e-18), np.float64(-1.0131487183969746e-18), np.float64(4.221454185688748e-19)],
+    [np.float64(-0.35951080550020226), np.float64(-0.10993517564892244), np.float64(0.12828317520407462), np.float64(-0.06727211863790784), np.float64(0.02888483594978499), np.float64(-0.011156827117820452), np.float64(0.003935993426140997), np.float64(-0.0012270633972598227), np.float64(0.0002918978170821283), np.float64(-7.751000840748687e-06), np.float64(-5.5799749242161456e-05), np.float64(5.4029234610582617e-05), np.float64(-3.8677118005247395e-05), np.float64(2.4659076656973736e-05), np.float64(-1.4812350078470264e-05), np.float64(8.590465157045744e-06), np.float64(-4.871677025851566e-06), np.float64(2.721497359928708e-06), np.float64(-1.504506922561602e-06), np.float64(8.255407273089358e-07), np.float64(-4.5053064085641674e-07), np.float64(2.448887362639938e-07), np.float64(-1.3271275135415133e-07), np.float64(7.17589613084803e-08), np.float64(-3.873426459489578e-08), np.float64(2.0880749226456104e-08), np.float64(-1.1245033012078744e-08), np.float64(6.051172438960396e-09), np.float64(-3.2543096238927447e-09), np.float64(1.7493523314838905e-09), np.float64(-9.4002584505079e-10), np.float64(5.049880706026049e-10), np.float64(-2.712240083542699e-10), np.float64(1.45647038649702e-10), np.float64(-7.820200437954491e-11), np.float64(4.1984539264500465e-11), np.float64(-2.2538550396551682e-11), np.float64(1.2098604766441812e-11), np.float64(-6.494164406025384e-12), np.float64(3.4857368747660956e-12), np.float64(-1.8709099654856985e-12), np.float64(1.0041555219142151e-12), np.float64(-5.389408433754435e-13), np.float64(2.892510699105425e-13), np.float64(-1.5524011905385897e-13), np.float64(8.331613951016408e-14), np.float64(-4.4714802693483305e-14), np.float64(2.399778497314974e-14), np.float64(-1.2879208184704404e-14), np.float64(6.912031884344903e-15), np.float64(-3.709549522492538e-15), np.float64(1.990837047384409e-15), np.float64(-1.0684384152371041e-15), np.float64(5.734063355277588e-16), np.float64(-3.0773308668477996e-16), np.float64(1.6515151642179737e-16), np.float64(-8.862989887398692e-17), np.float64(4.755996613225916e-17), np.float64(-2.5513904398476898e-17), np.float64(1.3673355915863094e-17), np.float64(-7.302129419476395e-18), np.float64(3.851746134050904e-18), np.float64(-1.9420577229887513e-18), np.float64(8.091909359431358e-19)],
+    [np.float64(0.17493458997773245), np.float64(-0.13253057762059978), np.float64(0.020046721982906003), np.float64(0.013541703830843776), np.float64(-0.017005943806565716), np.float64(0.012746904629490116), np.float64(-0.008134429482334395), np.float64(0.004817110567615993), np.float64(-0.0027405772016556333), np.float64(0.0015237105598056595), np.float64(-0.0008356180714999071), np.float64(0.00045445569192335), np.float64(-0.0002458938834877242), np.float64(0.00013262522628759563), np.float64(-7.139223507320439e-05), np.float64(3.8383783282334266e-05), np.float64(-2.0621456546942925e-05), np.float64(1.1073682360302575e-05), np.float64(-5.944900727446621e-06), np.float64(3.1909920347015742e-06), np.float64(-1.7126373855806046e-06), np.float64(9.191407372364673e-07), np.float64(-4.932720598333328e-07), np.float64(2.6471918488446107e-07), np.float64(-1.420634734282566e-07), np.float64(7.623941898869778e-08), np.float64(-4.0914559579114945e-08), np.float64(2.1957240817891176e-08), np.float64(-1.178363626835784e-08), np.float64(6.323863578803849e-09), np.float64(-3.3938066485092345e-09), np.float64(1.8213480613433235e-09), np.float64(-9.77461939111963e-10), np.float64(5.245750633320443e-10), np.float64(-2.8152446966103703e-10), np.float64(1.5108635463431792e-10), np.float64(-8.108393373451389e-11), np.float64(4.3515578085083115e-11), np.float64(-2.3353662959543634e-11), np.float64(1.2533301891772357e-11), np.float64(-6.726299251777416e-12), np.float64(3.6098322562939646e-12), np.float64(-1.9373049030496233e-12), np.float64(1.039702357138009e-12), np.float64(-5.579819549596914e-13), np.float64(2.9945484110851925e-13), np.float64(-1.607098770749822e-13), np.float64(8.624895374502763e-14), np.float64(-4.628764975302285e-14), np.float64(2.4841421588931714e-14), np.float64(-1.3331768762075968e-14), np.float64(7.154826259343846e-15), np.float64(-3.839815446160314e-15), np.float64(2.0607313101210046e-15), np.float64(-1.1059399078169935e-15), np.float64(5.935245568000208e-16), np.float64(-3.185190655655259e-16), np.float64(1.7092121613706022e-16), np.float64(-9.169185971220959e-17), np.float64(4.913924992474557e-17), np.float64(-2.6242341263889748e-17), np.float64(1.3842369594346134e-17), np.float64(-6.97934596576309e-18), np.float64(2.908060806685867e-18)],
+    [np.float64(-0.08587799623640216), np.float64(0.12799729652324027), np.float64(-0.07604566186699718), np.float64(0.04051255844328909), np.float64(-0.020388186707383158), np.float64(0.009963749199746364), np.float64(-0.004805940552425734), np.float64(0.002311294355432297), np.float64(-0.0011154778325831467), np.float64(0.0005424209179406646), np.float64(-0.0002663582718268773), np.float64(0.00013221613823324912), np.float64(-6.634848780418864e-05), np.float64(3.3641660517215025e-05), np.float64(-1.721988096759028e-05), np.float64(8.888156892864288e-06), np.float64(-4.620833919501569e-06), np.float64(2.4169647467108065e-06), np.float64(-1.2706208703064759e-06), np.float64(6.707549965008845e-07), np.float64(-3.5528455033084093e-07), np.float64(1.8869885191799007e-07), np.float64(-1.0044026504765022e-07), np.float64(5.3555022294123164e-08), np.float64(-2.8595061008725996e-08), np.float64(1.5284652504852348e-08), np.float64(-8.17700359277463e-09), np.float64(4.3775150629995094e-09), np.float64(-2.3447317265467832e-09), np.float64(1.2564380894874824e-09), np.float64(-6.734917956184773e-10), np.float64(3.6110706585182053e-10), np.float64(-1.936546303119482e-10), np.float64(1.0386972050131279e-10), np.float64(-5.571912249175793e-11), np.float64(2.989248848456021e-11), np.float64(-1.6038109366000474e-11), np.float64(8.605386010648744e-12), np.float64(-4.617511382750228e-12), np.float64(2.4777734614033474e-12), np.float64(-1.3296206556233975e-12), np.float64(7.135160145557488e-13), np.float64(-3.829017785422217e-13), np.float64(2.0548355054916601e-13), np.float64(-1.1027357447483375e-13), np.float64(5.917926050304952e-14), np.float64(-3.1759269565743015e-14), np.float64(1.7044086895622824e-14), np.float64(-9.147001392251379e-15), np.float64(4.9089107512759685e-15), np.float64(-2.6344660586167327e-15), np.float64(1.4138420980903284e-15), np.float64(-7.587693934306406e-16), np.float64(4.072105040634307e-16), np.float64(-2.1853836188201092e-16), np.float64(1.1728263410741865e-16), np.float64(-6.294040944843031e-17), np.float64(3.3774537438803605e-17), np.float64(-1.8118559768254597e-17), np.float64(9.710039353053611e-18), np.float64(-5.185548690074643e-18), np.float64(2.7352833858906287e-18), np.float64(-1.3791338968079604e-18), np.float64(5.746389737708413e-19)],
+    [np.float64(0.02664997615543401), np.float64(-0.04607081009091971), np.float64(0.03329979958358112), np.float64(-0.021621313151958185), np.float64(0.013135383098642754), np.float64(-0.007648645094817189), np.float64(0.004332691674818518), np.float64(-0.0024102636895916925), np.float64(0.0013248494073225137), np.float64(-0.0007224571826930927), np.float64(0.0003918843705914111), np.float64(-0.000211822386596804), np.float64(0.00011422586942973025), np.float64(-6.150006430144133e-05), np.float64(3.307739597633335e-05), np.float64(-1.7777974632417388e-05), np.float64(9.550559099039305e-06), np.float64(-5.129059672456268e-06), np.float64(2.753935892460258e-06), np.float64(-1.4784509507113612e-06), np.float64(7.936282174112329e-07), np.float64(-4.2598850058054524e-07), np.float64(2.286432463509122e-07), np.float64(-1.2271702232953576e-07), np.float64(6.586298392216243e-08), np.float64(-3.534850626608474e-08), np.float64(1.8971244003400692e-08), np.float64(-1.0181623455195738e-08), np.float64(5.464314813829526e-09), np.float64(-2.9325980736138453e-09), np.float64(1.5738669778702437e-09), np.float64(-8.446611158873731e-10), np.float64(4.533109784672141e-10), np.float64(-2.4328168568937627e-10), np.float64(1.3056361831503393e-10), np.float64(-7.007040400556016e-11), np.float64(3.760510669815e-11), np.float64(-2.0181751081869436e-11), np.float64(1.0831054092341491e-11), np.float64(-5.8127613936952e-12), np.float64(3.1195660412113247e-12), np.float64(-1.6741941022132644e-12), np.float64(8.984985517841637e-13), np.float64(-4.82201899194368e-13), np.float64(2.5878578774563926e-13), np.float64(-1.3888389997837278e-13), np.float64(7.453553393378052e-14), np.float64(-4.000136543295361e-14), np.float64(2.146773638741104e-14), np.float64(-1.152119908021666e-14), np.float64(6.1831402237566066e-15), np.float64(-3.3183369611627655e-15), np.float64(1.7808682750367747e-15), np.float64(-9.557468116193857e-16), np.float64(5.129240361967787e-16), np.float64(-2.752708598628823e-16), np.float64(1.477260210113144e-16), np.float64(-7.927158785001395e-17), np.float64(4.2525788688688736e-17), np.float64(-2.279030452398354e-17), np.float64(1.2170941967745598e-17), np.float64(-6.419956114181353e-18), np.float64(3.2369526709826298e-18), np.float64(-1.3487302817262479e-18)],
+    [np.float64(-0.005087135154882115), np.float64(0.009231319871598372), np.float64(-0.007214742736723321), np.float64(0.005079333242792772), np.float64(-0.003324538800590909), np.float64(0.0020665492910572677), np.float64(-0.0012377770976329772), np.float64(0.0007215754806467639), np.float64(-0.00041232731676896445), np.float64(0.00023213077420045745), np.float64(-0.00012922936749154925), np.float64(7.133637354076514e-05), np.float64(-3.9125669506117025e-05), np.float64(2.135367673727717e-05), np.float64(-1.161023671099612e-05), np.float64(6.294258137358318e-06), np.float64(-3.4046391309286148e-06), np.float64(1.8384055101586473e-06), np.float64(-9.913451048002063e-07), np.float64(5.340139195038857e-07), np.float64(-2.8742570889405e-07), np.float64(1.5460458342243153e-07), np.float64(-8.311964954202854e-08), np.float64(4.467010281656685e-08), np.float64(-2.3999321036020276e-08), np.float64(1.2890757177898164e-08), np.float64(-6.922735959986884e-09), np.float64(3.7171874144042226e-09), np.float64(-1.995731694001275e-09), np.float64(1.0713997147460945e-09), np.float64(-5.751364797290803e-10), np.float64(3.087214758453297e-10), np.float64(-1.6570835573495207e-10), np.float64(8.894214607109217e-11), np.float64(-4.7737486976589604e-11), np.float64(2.5621390593370092e-11), np.float64(-1.3751147786224555e-11), np.float64(7.380228469091434e-12), np.float64(-3.960923361601902e-12), np.float64(2.1257870923407283e-12), np.float64(-1.1408814084421404e-12), np.float64(6.122928857896239e-13), np.float64(-3.286066415899361e-13), np.float64(1.7635680321878023e-13), np.float64(-9.464705955247418e-14), np.float64(5.079503601774928e-14), np.float64(-2.726056311933158e-14), np.float64(1.46301205916765e-14), np.float64(-7.851644413090116e-15), np.float64(4.2137915904097855e-15), np.float64(-2.2614409450138036e-15), np.float64(1.2136606019646734e-15), np.float64(-6.513419466923985e-16), np.float64(3.4955899731492454e-16), np.float64(-1.8759919607690832e-16), np.float64(1.0067888384765225e-16), np.float64(-5.403004556168253e-17), np.float64(2.899319205398206e-17), np.float64(-1.555360131363475e-17), np.float64(8.335445193244637e-18), np.float64(-4.451464668990943e-18), np.float64(2.3480689705746716e-18), np.float64(-1.1839004246233666e-18), np.float64(4.932918701381847e-19)],
+    [np.float64(0.0006454768915704566), np.float64(-0.001196557823052477), np.float64(0.0009743792976015991), np.float64(-0.0007182723791055646), np.float64(0.0004913855618565048), np.float64(-0.00031785851300131137), np.float64(0.00019709451445398056), np.float64(-0.00011833795186386364), np.float64(6.931649199948305e-05), np.float64(-3.983425136617894e-05), np.float64(2.2554822579444065e-05), np.float64(-1.2624233932137905e-05), np.float64(7.002427393753256e-06), np.float64(-3.856731957576026e-06), np.float64(2.1124147148702638e-06), np.float64(-1.1519800786585705e-06), np.float64(6.26068821784011e-07), np.float64(-3.393349881021513e-07), np.float64(1.8353322425944934e-07), np.float64(-9.910073759775523e-08), np.float64(5.344043026280049e-08), np.float64(-2.878827403830849e-08), np.float64(1.549564787604991e-08), np.float64(-8.335423128964618e-09), np.float64(4.481554107734462e-09), np.float64(-2.408570311231296e-09), np.float64(1.2940657360740398e-09), np.float64(-6.951018675452154e-10), np.float64(3.7330025505393243e-10), np.float64(-2.0044885585396323e-10), np.float64(1.0762130948125615e-10), np.float64(-5.7776774042196446e-11), np.float64(3.10153873484607e-11), np.float64(-1.6648563139761596e-11), np.float64(8.93628903798372e-12), np.float64(-4.796480606989915e-12), np.float64(2.5744025370921635e-12), np.float64(-1.3817231450284998e-12), np.float64(7.415806955501312e-13), np.float64(-3.980064996926552e-13), np.float64(2.136079929317948e-13), np.float64(-1.146413727328362e-13), np.float64(6.1526547952793e-14), np.float64(-3.302034452137726e-14), np.float64(1.7721439289116274e-14), np.float64(-9.510756925884227e-15), np.float64(5.104229051614507e-15), np.float64(-2.7393304846997964e-15), np.float64(1.4701379256651749e-15), np.float64(-7.8898953283560215e-16), np.float64(4.234323289793327e-16), np.float64(-2.272461135499193e-16), np.float64(1.219575269888295e-16), np.float64(-6.545160809586843e-17), np.float64(3.5126188254535527e-17), np.float64(-1.885118312605212e-17), np.float64(1.0116625787870063e-17), np.float64(-5.428708066260295e-18), np.float64(2.912269075540812e-18), np.float64(-1.5607358259924936e-18), np.float64(8.334961131628966e-19), np.float64(-4.396545123564983e-19), np.float64(2.216745770977177e-19), np.float64(-9.236441561551737e-20)],
+    [np.float64(-5.8263640583020745e-05), np.float64(0.00010925923075492961), np.float64(-9.123741251078639e-05), np.float64(6.931608487892682e-05), np.float64(-4.8882525993969596e-05), np.float64(3.2526497383165673e-05), np.float64(-2.068363439126073e-05), np.float64(1.2693912222264529e-05), np.float64(-7.576019411099099e-06), np.float64(4.423051483063637e-06), np.float64(-2.5376705001100126e-06), np.float64(1.4359806658326162e-06), np.float64(-8.037090772498739e-07), np.float64(4.459286486345224e-07), np.float64(-2.45711547878236e-07), np.float64(1.3464732915327152e-07), np.float64(-7.346391092308619e-08), np.float64(3.994351780315994e-08), np.float64(-2.1658403636405718e-08), np.float64(1.1718244890457482e-08), np.float64(-6.3292450645482494e-09), np.float64(3.413902150089093e-09), np.float64(-1.8394341690281483e-09), np.float64(9.902614090709034e-10), np.float64(-5.327528146098352e-10), np.float64(2.8646627140552237e-10), np.float64(-1.5397197663253353e-10), np.float64(8.273107402482365e-11), np.float64(-4.444108415437361e-11), np.float64(2.3867857143269338e-11), np.float64(-1.2816626909983403e-11), np.float64(6.88145554445073e-12), np.float64(-3.694406694790904e-12), np.float64(1.983243391337467e-12), np.float64(-1.0645875592330342e-12), np.float64(5.714344486755375e-13), np.float64(-3.067153460636525e-13), np.float64(1.6462360122961657e-13), np.float64(-8.835658012313857e-14), np.float64(4.7421799044222216e-14), np.float64(-2.545136872757527e-14), np.float64(1.3659649358525406e-14), np.float64(-7.331017752544793e-15), np.float64(3.934468767929941e-15), np.float64(-2.111570911079779e-15), np.float64(1.1332441003346509e-15), np.float64(-6.081908404156161e-16), np.float64(3.2640379286346643e-16), np.float64(-1.7517400791248512e-16), np.float64(9.401203818027454e-17), np.float64(-5.0454132358403696e-17), np.float64(2.7077563733731896e-17), np.float64(-1.4531887917395646e-17), np.float64(7.798911510599094e-18), np.float64(-4.185476277643659e-18), np.float64(2.2462217098795033e-18), np.float64(-1.2054516886831321e-18), np.float64(6.468606043765875e-19), np.float64(-3.470130312503041e-19), np.float64(1.859703636383252e-19), np.float64(-9.931571250152056e-20), np.float64(5.2387292210130713e-20), np.float64(-2.641376614960953e-20), np.float64(1.1005737275632698e-20)],
+    [np.float64(3.935847969419735e-06), np.float64(-7.433881003177129e-06), np.float64(6.314114479412722e-06), np.float64(-4.901387053355243e-06), np.float64(3.535274491680246e-06), np.float64(-2.4037567306895095e-06), np.float64(1.5590311751673978e-06), np.float64(-9.737063768912605e-07), np.float64(5.900412080376554e-07), np.float64(-3.4899723926250594e-07), np.float64(2.0245333043394305e-07), np.float64(-1.1562570667312025e-07), np.float64(6.521468634815601e-08), np.float64(-3.6414274704812005e-08), np.float64(2.0169622843671146e-08), np.float64(-1.109993120832778e-08), np.float64(6.077153191597591e-09), np.float64(-3.3135148327829746e-09), np.float64(1.8007330719043295e-09), np.float64(-9.760512747510242e-10), np.float64(5.279491077321003e-10), np.float64(-2.850979816152915e-10), np.float64(1.5375446699962987e-10), np.float64(-8.283453602861427e-11), np.float64(4.459021290082801e-11), np.float64(-2.398760587048307e-11), np.float64(1.2897717181051754e-11), np.float64(-6.932092290264462e-12), np.float64(3.724590025905736e-12), np.float64(-2.0007119512547552e-12), np.float64(1.074498245437665e-12), np.float64(-5.769791374166761e-13), np.float64(3.097862176881213e-13), np.float64(-1.6631174746160116e-13), np.float64(8.927944312345977e-14), np.float64(-4.7924179905418415e-14), np.float64(2.5723972894131825e-14), np.float64(-1.380720658290714e-14), np.float64(7.410736886080131e-15), np.float64(-3.9774744400440834e-15), np.float64(2.134744498463106e-15), np.float64(-1.1457201041236475e-15), np.float64(6.149029294835923e-16), np.float64(-3.3001295198909894e-16), np.float64(1.7711387446062008e-16), np.float64(-9.505434425738515e-17), np.float64(5.101402893872972e-17), np.float64(-2.7378264900102814e-17), np.float64(1.4693361211893217e-17), np.float64(-7.885614737177926e-18), np.float64(4.2320354566587584e-18), np.float64(-2.271237285464429e-18), np.float64(1.2189201298527662e-18), np.float64(-6.541651855633563e-19), np.float64(3.510738611178167e-19), np.float64(-1.8841104973281002e-19), np.float64(1.0111222483215186e-19), np.float64(-5.4258107738749e-20), np.float64(2.9107157209290003e-20), np.float64(-1.5599037404785643e-20), np.float64(8.330519050619816e-21), np.float64(-4.3942026435427916e-21), np.float64(2.2155649207855827e-21), np.float64(-9.231521948897386e-22)],
+    [np.float64(-2.066773130456935e-07), np.float64(3.922900445677778e-07), np.float64(-3.373268628322731e-07), np.float64(2.661412580128811e-07), np.float64(-1.953648792664808e-07), np.float64(1.3515228385562235e-07), np.float64(-8.908336427750861e-08), np.float64(5.645434791338712e-08), np.float64(-3.465259057964316e-08), np.float64(2.0726289990857152e-08), np.float64(-1.2138810244333745e-08), np.float64(6.989154272517958e-09), np.float64(-3.968914744448334e-09), np.float64(2.2287530873587615e-09), np.float64(-1.240302397659498e-09), np.float64(6.852165650963067e-10), np.float64(-3.7633986907255745e-10), np.float64(2.057245182664736e-10), np.float64(-1.1203444276008558e-10), np.float64(6.082822579800853e-11), np.float64(-3.2946680711187e-11), np.float64(1.7810829689764437e-11), np.float64(-9.613765855913988e-12), np.float64(5.182945731401315e-12), np.float64(-2.7915330662408695e-12), np.float64(1.5023770723415316e-12), np.float64(-8.080800854935325e-13), np.float64(4.344342884813923e-13), np.float64(-2.334702563613698e-13), np.float64(1.25432836788431e-13), np.float64(-6.737369447204443e-14), np.float64(3.618181852489746e-14), np.float64(-1.942800729610653e-14), np.float64(1.043079224260111e-14), np.float64(-5.599741692817673e-15), np.float64(3.005997420250989e-15), np.float64(-1.61356180878918e-15), np.float64(8.660921319541102e-16), np.float64(-4.648662985875928e-16), np.float64(2.495058024230174e-16), np.float64(-1.33913486360369e-16), np.float64(7.187220894856479e-17), np.float64(-3.857377927968413e-17), np.float64(2.0702323833471116e-17), np.float64(-1.111073123350492e-17), np.float64(5.962982751739445e-18), np.float64(-3.2002389092665953e-18), np.float64(1.717511419873389e-18), np.float64(-9.217552073463666e-19), np.float64(4.946870864420069e-19), np.float64(-2.654879201103163e-19), np.float64(1.4248145074145205e-19), np.float64(-7.646652321255015e-20), np.float64(4.103776857048703e-20), np.float64(-2.2023937968218123e-20), np.float64(1.181960548843363e-20), np.float64(-6.343082819589929e-21), np.float64(3.403779669997966e-21), np.float64(-1.825982669863739e-21), np.float64(9.78576334715906e-22), np.float64(-5.225995238412465e-22), np.float64(2.7566209859118827e-22), np.float64(-1.3898933545360154e-22), np.float64(5.79122339077916e-23)],
+    [np.float64(8.686503590764639e-09), np.float64(-1.65473383773328e-08), np.float64(1.436297104024318e-08), np.float64(-1.1477650061935758e-08), np.float64(8.545735734449281e-09), np.float64(-5.996895398705995e-09), np.float64(4.006769217225344e-09), np.float64(-2.5709942873109855e-09), np.float64(1.5957903137197333e-09), np.float64(-9.638514964105788e-10), np.float64(5.693034083832903e-10), np.float64(-3.3017290333152663e-10), np.float64(1.8865033340673577e-10), np.float64(-1.0648549423310168e-10), np.float64(5.951496734516843e-11), np.float64(-3.299706105855277e-11), np.float64(1.8176169950047973e-11), np.float64(-9.959820041291788e-12), np.float64(5.434586636259126e-12), np.float64(-2.955348188377066e-12), np.float64(1.6027689647939799e-12), np.float64(-8.673418310650257e-13), np.float64(4.685520234722645e-13), np.float64(-2.52770908395652e-13), np.float64(1.3621383779622755e-13), np.float64(-7.333962383515107e-14), np.float64(3.946007898357147e-14), np.float64(-2.121981181620983e-14), np.float64(1.1406148867425603e-14), np.float64(-6.1290026846446705e-15), np.float64(3.292493998913348e-15), np.float64(-1.7683538979350812e-15), np.float64(9.49602414668502e-16), np.float64(-5.098685027437213e-16), np.float64(2.7373502339253325e-16), np.float64(-1.4694941187717948e-16), np.float64(7.888203458470995e-17), np.float64(-4.234157301715805e-17), np.float64(2.2726843116453342e-17), np.float64(-1.21982659739002e-17), np.float64(6.547066935972056e-18), np.float64(-3.513883743466329e-18), np.float64(1.8859130620734142e-18), np.float64(-1.0121642406073693e-18), np.float64(5.432208443677431e-19), np.float64(-2.915404497323382e-19), np.float64(1.5646558511988773e-19), np.float64(-8.397247820010338e-20), np.float64(4.506648114324224e-20), np.float64(-2.4186285811956072e-20), np.float64(1.2980272483168877e-20), np.float64(-6.966228485649673e-21), np.float64(3.73861718923897e-21), np.float64(-2.0064281031012134e-21), np.float64(1.0767998928777274e-21), np.float64(-5.778872839668759e-22), np.float64(3.1012775820143866e-22), np.float64(-1.6641856652296907e-22), np.float64(8.927647822442783e-23), np.float64(-4.7844845392442827e-23), np.float64(2.5551093650505325e-23), np.float64(-1.347775540767547e-23), np.float64(6.795509304866681e-24), np.float64(-2.8314628209420286e-24)],
+    [np.float64(-2.990345478876676e-10), np.float64(5.71233848719464e-10), np.float64(-4.995187719435948e-10), np.float64(4.033291211769605e-10), np.float64(-3.038540338668844e-10), np.float64(2.1581425812477414e-10), np.float64(-1.4588437413085387e-10), np.float64(9.462846621897066e-11), np.float64(-5.931417889974814e-11), np.float64(3.613911353904689e-11), np.float64(-2.1509025881667172e-11), np.float64(1.2556725207741481e-11), np.float64(-7.214967848464867e-12), np.float64(4.091987343575556e-12), np.float64(-2.2961768375541066e-12), np.float64(1.2773220937382121e-12), np.float64(-7.055445195435752e-13), np.float64(3.8748789857515953e-13), np.float64(-2.1182569107340982e-13), np.float64(1.1536560517811403e-13), np.float64(-6.264259666797539e-14), np.float64(3.3932639840954545e-14), np.float64(-1.8345503084699946e-14), np.float64(9.90318393977599e-15), np.float64(-5.339362356133946e-15), np.float64(2.8759572609812404e-15), np.float64(-1.5478941101289816e-15), np.float64(8.325982729921944e-16), np.float64(-4.476314988820801e-16), np.float64(2.405695638739485e-16), np.float64(-1.2924999605685865e-16), np.float64(6.942531210609342e-17), np.float64(-3.728416512376491e-17), np.float64(2.0020158802573773e-17), np.float64(-1.0748818108929984e-17), np.float64(5.770516821981004e-18), np.float64(-3.0976898450311078e-18), np.float64(1.6627884451454886e-18), np.float64(-8.925182660909434e-19), np.float64(4.790516771180387e-19), np.float64(-2.571200609613425e-19), np.float64(1.3800042488821573e-19), np.float64(-7.406580097419567e-20), np.float64(3.975112392023773e-20), np.float64(-2.1334216787945772e-20), np.float64(1.144986985810027e-20), np.float64(-6.144997325473445e-21), np.float64(3.297924673542651e-21), np.float64(-1.7699382273013215e-21), np.float64(9.498919083003859e-22), np.float64(-5.0978757394367295e-22), np.float64(2.7359205813718107e-22), np.float64(-1.4683075534683792e-22), np.float64(7.880065774295856e-23), np.float64(-4.22903622505459e-23), np.float64(2.2696017935564084e-23), np.float64(-1.2179999386546807e-23), np.float64(6.5359463822771706e-24), np.float64(-3.506257555217285e-24), np.float64(1.879065709443117e-24), np.float64(-1.0034976093571413e-24), np.float64(5.293274873249532e-25), np.float64(-2.6688791443824415e-25), np.float64(1.112033248388646e-25)],
+    [np.float64(8.590913565501381e-12), np.float64(-1.644747244344771e-11), np.float64(1.4469771653023546e-11), np.float64(-1.17844033378004e-11), np.float64(8.96677770680088e-12), np.float64(-6.435116137299384e-12), np.float64(4.394394784525721e-12), np.float64(-2.8778475636047527e-12), np.float64(1.8197393664590925e-12), np.float64(-1.1174802489107638e-12), np.float64(6.697195285079766e-13), np.float64(-3.9334057611250915e-13), np.float64(2.2718618553574766e-13), np.float64(-1.2942068908024125e-13), np.float64(7.289509240807865e-14), np.float64(-4.0677477416306326e-14), np.float64(2.252734993914674e-14), np.float64(-1.23987980370397e-14), np.float64(6.789986471957273e-15), np.float64(-3.703350837544549e-15), np.float64(2.0132562053050933e-15), np.float64(-1.091592054367329e-15), np.float64(5.90616535804845e-16), np.float64(-3.190204156187325e-16), np.float64(1.7208669323732956e-16), np.float64(-9.272809664110933e-17), np.float64(4.9923649294176413e-17), np.float64(-2.6860168721642793e-17), np.float64(1.444373823973649e-17), np.float64(-7.763677770407289e-18), np.float64(4.171679860813903e-18), np.float64(-2.2409934260347286e-18), np.float64(1.2035953798137423e-18), np.float64(-6.46323380469939e-19), np.float64(3.4702736500696527e-19), np.float64(-1.8630905515037655e-19), np.float64(1.0001610706674565e-19), np.float64(-5.3688223611849094e-20), np.float64(2.8818210992911103e-20), np.float64(-1.546815236298492e-20), np.float64(8.302270523196498e-21), np.float64(-4.4559995494906615e-21), np.float64(2.391582765028338e-21), np.float64(-1.2835696409529957e-21), np.float64(6.88887899681626e-22), np.float64(-3.697207587406423e-22), np.float64(1.9842487064595156e-22), np.float64(-1.0649176346859133e-22), np.float64(5.715234902318239e-23), np.float64(-3.067260793736177e-23), np.float64(1.6461377830941049e-23), np.float64(-8.834475186724073e-24), np.float64(4.74126861622721e-24), np.float64(-2.5445299877659388e-24), np.float64(1.3655867527694875e-24), np.float64(-7.328712422035599e-25), np.float64(3.9330129668053335e-25), np.float64(-2.1105063778619116e-25), np.float64(1.132197162729119e-25), np.float64(-6.067646371141272e-26), np.float64(3.240370522569741e-26), np.float64(-1.70923904495467e-26), np.float64(8.618015798273004e-27), np.float64(-3.5908408939843676e-27)],
+    [np.float64(-2.0917254877820944e-13), np.float64(4.0119923996513445e-13), np.float64(-3.54740772706991e-13), np.float64(2.9102066080556357e-13), np.float64(-2.233428263991212e-13), np.float64(1.6174167092080764e-13), np.float64(-1.114456861221764e-13), np.float64(7.361103948580295e-14), np.float64(-4.691539004850211e-14), np.float64(2.90168229731649e-14), np.float64(-1.750101775108679e-14), np.float64(1.033617173760745e-14), np.float64(-5.998883534475675e-15), np.float64(3.4315549237745648e-15), np.float64(-1.9396071153036413e-15), np.float64(1.0855680321354046e-15), np.float64(-6.026839976939976e-16), np.float64(3.3239382481927193e-16), np.float64(-1.823390285035146e-16), np.float64(9.958876368355027e-17), np.float64(-5.420107292565327e-17), np.float64(2.9415059140008093e-17), np.float64(-1.5927172971077508e-17), np.float64(8.608197174057999e-18), np.float64(-4.6456887603566085e-18), np.float64(2.504271314126552e-18), np.float64(-1.3486824394693845e-18), np.float64(7.258020503301345e-19), np.float64(-3.903671477306522e-19), np.float64(2.0985912561042288e-19), np.float64(-1.1277792203800966e-19), np.float64(6.058922050070497e-20), np.float64(-3.254379240454546e-20), np.float64(1.747686017900872e-20), np.float64(-9.384208597045561e-21), np.float64(5.038298888968273e-21), np.float64(-2.704783533930433e-21), np.float64(1.4519494695311658e-21), np.float64(-7.793764122411127e-22), np.float64(4.1833556023327908e-22), np.float64(-2.245370478042823e-22), np.float64(1.2051469884720668e-22), np.float64(-6.46819701844077e-23), np.float64(3.471519177019611e-23), np.float64(-1.8631614803556507e-23), np.float64(9.999474866371489e-24), np.float64(-5.366616649775894e-24), np.float64(2.8801914484827184e-24), np.float64(-1.5457530220275515e-24), np.float64(8.295780575244021e-25), np.float64(-4.4521845091616415e-25), np.float64(2.389395567778271e-25), np.float64(-1.2823368301292542e-25), np.float64(6.882010473777031e-26), np.float64(-3.693407342011202e-26), np.float64(1.982146347575685e-26), np.float64(-1.0637351514473825e-26), np.float64(5.708143427674616e-27), np.float64(-3.0621772765051276e-27), np.float64(1.6410755497614014e-27), np.float64(-8.764013176832831e-28), np.float64(4.622864684014418e-28), np.float64(-2.3308572781915197e-28), np.float64(9.711908241723077e-29)],
+    [np.float64(4.37271797638155e-15), np.float64(-8.399934282987266e-15), np.float64(7.45909871906253e-15), np.float64(-6.157789084734837e-15), np.float64(4.7612000842121434e-15), np.float64(-3.475643388042952e-15), np.float64(2.414102348741187e-15), np.float64(-1.6068661533418328e-15), np.float64(1.0315055309059747e-15), np.float64(-6.42171285654316e-16), np.float64(3.895945520715493e-16), np.float64(-2.3129129548980095e-16), np.float64(1.3484407147327527e-16), np.float64(-7.743627752648789e-17), np.float64(4.391496954891062e-17), np.float64(-2.4647880130848035e-17), np.float64(1.371639556438039e-17), np.float64(-7.579843507525494e-18), np.float64(4.164823640259028e-18), np.float64(-2.277778148928511e-18), np.float64(1.2410434616395257e-18), np.float64(-6.741212742092268e-19), np.float64(3.6527706331932163e-19), np.float64(-1.9753789693479664e-19), np.float64(1.0665787658002077e-19), np.float64(-5.75159268878208e-20), np.float64(3.0984706010117974e-20), np.float64(-1.6678622190725495e-20), np.float64(8.972183261562506e-21), np.float64(-4.824124150843235e-21), np.float64(2.5927864673295993e-21), np.float64(-1.3930899717983376e-21), np.float64(7.483149694945755e-22), np.float64(-4.0188822720625393e-22), np.float64(2.158040983778784e-22), np.float64(-1.1586755136645104e-22), np.float64(6.220465573985535e-23), np.float64(-3.3392711305857547e-23), np.float64(1.7924833798266255e-23), np.float64(-9.621409665660362e-24), np.float64(5.164243060744281e-24), np.float64(-2.7718032931355025e-24), np.float64(1.4876766405790612e-24), np.float64(-7.984489605732115e-25), np.float64(4.285286679165083e-25), np.float64(-2.2998948576335345e-25), np.float64(1.2343333515634984e-25), np.float64(-6.624515326989634e-26), np.float64(3.555277731678655e-26), np.float64(-1.90805642789767e-26), np.float64(1.0240178491260118e-26), np.float64(-5.495696628632166e-27), np.float64(2.9494230325443997e-27), np.float64(-1.5828890689717295e-27), np.float64(8.494983015680867e-28), np.float64(-4.55901630506514e-28), np.float64(2.44663416188043e-28), np.float64(-1.312896457363867e-28), np.float64(7.043134587327039e-29), np.float64(-3.7745421439485837e-29), np.float64(2.0157597128585743e-29), np.float64(-1.0632782926923721e-29), np.float64(5.361069872887366e-30), np.float64(-2.2337798461361527e-30)],
+    [np.float64(-7.935632040439955e-17), np.float64(1.526435903155495e-16), np.float64(-1.3604843303401677e-16), np.float64(1.1293000975009147e-16), np.float64(-8.789449844174412e-17), np.float64(6.462084903445617e-17), np.float64(-4.5208992116843394e-17), np.float64(3.030302158471025e-17), np.float64(-1.9580938562461715e-17), np.float64(1.2264115230578804e-17), np.float64(-7.481079654960666e-18), np.float64(4.462853130372727e-18), np.float64(-2.6129341200320116e-18), np.float64(1.5060501134781304e-18), np.float64(-8.568002493182415e-19), np.float64(4.821849247014204e-19), np.float64(-2.6894191263237196e-19), np.float64(1.4890276489917927e-19), np.float64(-8.194530070723535e-20), np.float64(4.487499939979815e-20), np.float64(-2.4476218068321635e-20), np.float64(1.3306824079081323e-20), np.float64(-7.215504984579717e-21), np.float64(3.904305646239164e-21), np.float64(-2.1090502616211055e-21), np.float64(1.137740269070055e-21), np.float64(-6.130998292528465e-22), np.float64(3.3010092311470984e-22), np.float64(-1.7760961115210712e-22), np.float64(9.551063126678094e-23), np.float64(-5.133947530771461e-23), np.float64(2.758700335617055e-23), np.float64(-1.4819786004566229e-23), np.float64(7.959543050764641e-24), np.float64(-4.2742755593561686e-24), np.float64(2.294987822029945e-24), np.float64(-1.2321223007102458e-24), np.float64(6.614428622328464e-25), np.float64(-3.550613729748472e-25), np.float64(1.9058687774991901e-25), np.float64(-1.0229765323901798e-25), np.float64(5.490666882937583e-26), np.float64(-2.94695924674207e-26), np.float64(1.5816668683248306e-26), np.float64(-8.488862888137602e-27), np.float64(4.555950698189287e-27), np.float64(-2.445144845065232e-27), np.float64(1.3122817625196163e-27), np.float64(-7.042829715656596e-28), np.float64(3.779770188395352e-28), np.float64(-2.028533199098862e-28), np.float64(1.0886735377813066e-28), np.float64(-5.84268181556023e-29), np.float64(3.135637476121578e-29), np.float64(-1.6828214390558695e-29), np.float64(9.031227618805109e-30), np.float64(-4.846684761072936e-30), np.float64(2.6007960005002855e-30), np.float64(-1.3952173165083745e-30), np.float64(7.477220613917413e-31), np.float64(-3.993141612107969e-31), np.float64(2.1063130844374033e-31), np.float64(-1.0620072092601593e-31), np.float64(4.4250316110806344e-32)],
+    [np.float64(1.2622270804228821e-18), np.float64(-2.4307027824061497e-18), np.float64(2.1734483110712334e-18), np.float64(-1.8128503773390334e-18), np.float64(1.4192597872602558e-18), np.float64(-1.0501528845020997e-18), np.float64(7.395107712452798e-19), np.float64(-4.988634930502404e-19), np.float64(3.2430882944998793e-19), np.float64(-2.0426470611144622e-19), np.float64(1.252357502373818e-19), np.float64(-7.504961490056292e-20), np.float64(4.411668061171623e-20), np.float64(-2.55169521852711e-20), np.float64(1.456047426702165e-20), np.float64(-8.215343943826839e-21), np.float64(4.592148640352777e-21), np.float64(-2.5471497203161706e-21), np.float64(1.4039066059953617e-21), np.float64(-7.697813883936654e-22), np.float64(4.2029968575875798e-22), np.float64(-2.286962300240691e-22), np.float64(1.2409448906147024e-22), np.float64(-6.718520942901782e-23), np.float64(3.630895916699866e-23), np.float64(-1.9594238226153695e-23), np.float64(1.056193158148494e-23), np.float64(-5.688008800706701e-24), np.float64(3.060981721387274e-24), np.float64(-1.6463045014635582e-24), np.float64(8.850356705033623e-25), np.float64(-4.756134975883518e-25), np.float64(2.555191327582551e-25), np.float64(-1.3724444245837776e-25), np.float64(7.370364524772181e-26), np.float64(-3.957513864611427e-26), np.float64(2.12475157834543e-26), np.float64(-1.1406603212146765e-26), np.float64(6.123151676067352e-27), np.float64(-3.286779312970819e-27), np.float64(1.7642003301825313e-27), np.float64(-9.469149981283258e-28), np.float64(5.082330567182625e-28), np.float64(-2.727759360910647e-28), np.float64(1.464004214060611e-28), np.float64(-7.857297758888237e-29), np.float64(4.216963812097017e-29), np.float64(-2.2632015334803788e-29), np.float64(1.2146299654247699e-29), np.float64(-6.51872592157271e-30), np.float64(3.4984837079398576e-30), np.float64(-1.877568211140763e-30), np.float64(1.007651937301147e-30), np.float64(-5.4078464471594105e-31), np.float64(2.9022625353042113e-31), np.float64(-1.5575628075938452e-31), np.float64(8.358794491813536e-32), np.float64(-4.485441917511448e-32), np.float64(2.4062506512830572e-32), np.float64(-1.289553149349858e-32), np.float64(6.886741776959089e-33), np.float64(-3.632637310533064e-33), np.float64(1.831582974032926e-33), np.float64(-7.631598644530059e-34)],
+    [np.float64(-1.7743158467838407e-20), np.float64(3.420286330609023e-20), np.float64(-3.067038856198645e-20), np.float64(2.5692067946168564e-20), np.float64(-2.0220258761728095e-20), np.float64(1.5048535999612423e-20), np.float64(-1.0660505956208897e-20), np.float64(7.233807472613522e-21), np.float64(-4.729087457809713e-21), np.float64(2.9941690271643058e-21), np.float64(-1.844497586884073e-21), np.float64(1.1100811575400281e-21), np.float64(-6.55019261284426e-22), np.float64(3.8011959234326146e-22), np.float64(-2.1752764753389963e-22), np.float64(1.230367272016543e-22), np.float64(-6.891832160315404e-23), np.float64(3.829484382625652e-23), np.float64(-2.1138098977440058e-23), np.float64(1.1604554503665097e-23), np.float64(-6.342499632918669e-24), np.float64(3.453995333182322e-24), np.float64(-1.8754702535932605e-24), np.float64(1.0159466678814138e-24), np.float64(-5.492938351871759e-25), np.float64(2.9653454173023227e-25), np.float64(-1.598878323093271e-25), np.float64(8.612563371224175e-26), np.float64(-4.635672504787648e-26), np.float64(2.4935937048847458e-26), np.float64(-1.3406849096620568e-26), np.float64(7.205432637858599e-27), np.float64(-3.871336388707621e-27), np.float64(2.079491760995103e-27), np.float64(-1.1167888926000866e-27), np.float64(5.996806978826498e-28), np.float64(-3.219719164551738e-28), np.float64(1.7285254083244931e-28), np.float64(-9.279017573101752e-29), np.float64(4.980849881429162e-29), np.float64(-2.673532458943502e-29), np.float64(1.435001021642128e-29), np.float64(-7.702061635916391e-30), np.float64(4.133827812150164e-30), np.float64(-2.2186582805402808e-30), np.float64(1.1907557380407787e-30), np.float64(-6.390729393609863e-31), np.float64(3.4298460842736933e-31), np.float64(-1.8407552954236103e-31), np.float64(9.879052872286356e-32), np.float64(-5.301916111511764e-32), np.float64(2.845437425198371e-32), np.float64(-1.5270881608348352e-32), np.float64(8.19555017661304e-33), np.float64(-4.3983583620869714e-33), np.float64(2.3604760813140595e-33), np.float64(-1.2667700234169104e-33), np.float64(6.797660034629898e-34), np.float64(-3.6466587595951325e-34), np.float64(1.954310443866262e-34), np.float64(-1.0436818717017954e-34), np.float64(5.505241903922718e-35), np.float64(-2.775754065542364e-35), np.float64(1.1565646663364855e-35)],
+    [np.float64(2.220440354026113e-22), np.float64(-4.284093072889805e-22), np.float64(3.851422684995113e-22), np.float64(-3.2387495080890593e-22), np.float64(2.561145353060588e-22), np.float64(-1.916175294174983e-22), np.float64(1.3648882435598838e-22), np.float64(-9.31201457290597e-23), np.float64(6.11949815706122e-23), np.float64(-3.8934227043195247e-23), np.float64(2.4092156866684253e-23), np.float64(-1.4558134456978065e-23), np.float64(8.621164942549044e-24), np.float64(-5.018855765374548e-24), np.float64(2.8800015254792378e-24), np.float64(-1.6328301077394272e-24), np.float64(9.164695111930425e-25), np.float64(-5.101121673996131e-25), np.float64(2.8197733021376777e-25), np.float64(-1.549871939205536e-25), np.float64(8.479250515568508e-26), np.float64(-4.621385971161094e-26), np.float64(2.511018036644936e-26), np.float64(-1.3609616391133093e-26), np.float64(7.361567773875064e-27), np.float64(-3.975526311966163e-27), np.float64(2.144165518514979e-27), np.float64(-1.1552454554825247e-27), np.float64(6.219186607929203e-28), np.float64(-3.3458729584644458e-28), np.float64(1.7991213532983867e-28), np.float64(-9.670154852761602e-29), np.float64(5.195958101478797e-29), np.float64(-2.791172768548237e-29), np.float64(1.4990639500004403e-29), np.float64(-8.04979020509953e-30), np.float64(4.3220981459020486e-30), np.float64(-2.3203951635173078e-30), np.float64(1.2456490199337413e-30), np.float64(-6.686564627154231e-31), np.float64(3.5891341286459296e-31), np.float64(-1.9264603489051031e-31), np.float64(1.033993259885907e-31), np.float64(-5.549646740396038e-32), np.float64(2.978551603045632e-32), np.float64(-1.5985961789610935e-32), np.float64(8.579610942948494e-33), np.float64(-4.604607777055709e-33), np.float64(2.471238902634209e-33), np.float64(-1.3262778062255814e-33), np.float64(7.117909317734511e-34), np.float64(-3.820049205903705e-34), np.float64(2.050143576475669e-34), np.float64(-1.1002679831327113e-34), np.float64(5.904880647227472e-35), np.float64(-3.1689854089648865e-35), np.float64(1.7006639764667074e-35), np.float64(-9.125995109166033e-36), np.float64(4.895713300396078e-36), np.float64(-2.623701693228091e-36), np.float64(1.401164344231653e-36), np.float64(-7.390900500592235e-37), np.float64(3.726507600783862e-37), np.float64(-1.5527126046793397e-37)],
+    [np.float64(-2.4898866187092057e-24), np.float64(4.8078089519029016e-24), np.float64(-4.332150486133127e-24), np.float64(3.65574346440805e-24), np.float64(-2.9034615224020295e-24), np.float64(2.182823558967432e-24), np.float64(-1.562693507608604e-24), np.float64(1.0715381851569129e-24), np.float64(-7.076039827310665e-25), np.float64(4.5226363517836645e-25), np.float64(-2.8103917274529997e-25), np.float64(1.7047376034719912e-25), np.float64(-1.0129839068486468e-25), np.float64(5.914967115488602e-26), np.float64(-3.403180832568896e-26), np.float64(1.9338491746287164e-26), np.float64(-1.0875434331469607e-26), np.float64(6.06335712371769e-27), np.float64(-3.356347359375836e-27), np.float64(1.8469464035866093e-27), np.float64(-1.0114296404842306e-27), np.float64(5.516918469830958e-28), np.float64(-2.9995623642257626e-28), np.float64(1.6266170131891608e-28), np.float64(-8.802326616416963e-29), np.float64(4.7552512369222433e-29), np.float64(-2.565424220893397e-29), np.float64(1.3825250728054222e-29), np.float64(-7.444071940258095e-30), np.float64(4.005426574840928e-30), np.float64(-2.154018195146306e-30), np.float64(1.1578750028117504e-30), np.float64(-6.2219289181967085e-31), np.float64(3.3424949321824504e-31), np.float64(-1.7952446748169191e-31), np.float64(9.640584942587675e-32), np.float64(-5.176372597785765e-32), np.float64(2.779088420559582e-32), np.float64(-1.4919132844162697e-32), np.float64(8.008603879573938e-33), np.float64(-4.2988082200352053e-33), np.float64(2.3073957061311328e-33), np.float64(-1.2384616323885246e-33), np.float64(6.647103155689095e-34), np.float64(-3.567581945605956e-34), np.float64(1.9147362996352241e-34), np.float64(-1.0276349561617865e-34), np.float64(5.515244367145052e-35), np.float64(-2.959971377052536e-35), np.float64(1.5885752561170446e-35), np.float64(-8.525623372734316e-36), np.float64(4.575546536917628e-36), np.float64(-2.455605447224785e-36), np.float64(1.3178709986548752e-36), np.float64(-7.072706883203856e-37), np.float64(3.795725240388029e-37), np.float64(-2.037010716031168e-37), np.float64(1.0930858307220684e-37), np.float64(-5.863957669383695e-38), np.float64(3.142609810917475e-38), np.float64(-1.6782704020285772e-38), np.float64(8.852602731892978e-39), np.float64(-4.463537253521267e-39), np.float64(1.8599737513253373e-39)],
+    [np.float64(2.5163348323717328e-26), np.float64(-4.8623931866788436e-26), np.float64(4.3904251071448363e-26), np.float64(-3.716707706785965e-26), np.float64(2.963622267839849e-26), np.float64(-2.2380029530254343e-26), np.float64(1.6097092886374927e-26), np.float64(-1.1089670551578083e-26), np.float64(7.356603890555823e-27), np.float64(-4.7222306209254274e-27), np.float64(2.9461395954783243e-27), np.float64(-1.7935796996388216e-27), np.float64(1.0692547496595266e-27), np.float64(-6.261605370480257e-28), np.float64(3.611752054234792e-28), np.float64(-2.0568805287010388e-28), np.float64(1.158916493469814e-28), np.float64(-6.471664394171452e-29), np.float64(3.587224035123187e-29), np.float64(-1.976237799078799e-29), np.float64(1.0832557065197134e-29), np.float64(-5.913315630573359e-30), np.float64(3.21714790324172e-30), np.float64(-1.745524998526197e-30), np.float64(9.449814152858455e-31), np.float64(-5.106801339904001e-31), np.float64(2.7558483902201826e-31), np.float64(-1.485477219675856e-31), np.float64(7.999835182529335e-32), np.float64(-4.305078608663631e-32), np.float64(2.3154258960950717e-32), np.float64(-1.2447503836543606e-32), np.float64(6.689237237154279e-33), np.float64(-3.593741255413533e-33), np.float64(1.9302743660938274e-33), np.float64(-1.036606780500056e-33), np.float64(5.566064278029638e-34), np.float64(-2.9883710557114616e-34), np.float64(1.6042911476089225e-34), np.float64(-8.61196531537633e-35), np.float64(4.622725790058559e-35), np.float64(-2.481279870752339e-35), np.float64(1.3318001576209487e-35), np.float64(-7.148108632172917e-36), np.float64(3.836493318193483e-36), np.float64(-2.0590686689564007e-36), np.float64(1.1051004476947904e-36), np.float64(-5.9310066388082915e-37), np.float64(3.1831106179568515e-37), np.float64(-1.7083313664471296e-37), np.float64(9.168343723499573e-38), np.float64(-4.920494899718123e-38), np.float64(2.6407519989862766e-38), np.float64(-1.4172014183588892e-38), np.float64(7.605844552351107e-39), np.float64(-4.08212800537476e-39), np.float64(2.190548320342365e-39), np.float64(-1.1751134515718221e-39), np.float64(6.307162859069174e-40), np.float64(-3.3805552042081823e-40), np.float64(1.8030515185447593e-40), np.float64(-9.528394975537816e-41), np.float64(4.800178908179549e-41), np.float64(-2.0040578395599972e-41)],
+    [np.float64(-2.3039162295045917e-28), np.float64(4.4548561644608075e-28), np.float64(-4.030049705379764e-28), np.float64(3.4215655521745766e-28), np.float64(-2.7382642454279363e-28), np.float64(2.076362509824024e-28), np.float64(-1.4999584410160358e-28), np.float64(1.037898196734664e-28), np.float64(-6.914640167853199e-29), np.float64(4.4565873095634185e-29), np.float64(-2.790929814325887e-29), np.float64(1.7049693207752094e-29), np.float64(-1.0195985884832378e-29), np.float64(5.987378150281066e-30), np.float64(-3.461997694829631e-30), np.float64(1.975778889663319e-30), np.float64(-1.1152557651433727e-30), np.float64(6.237570417798332e-31), np.float64(-3.4620335847158534e-31), np.float64(1.9093852489370272e-31), np.float64(-1.0475793881504456e-31), np.float64(5.722946130679766e-32), np.float64(-3.1155407180938453e-32), np.float64(1.6912685466472032e-32), np.float64(-9.159931384329944e-33), np.float64(4.951830438319712e-33), np.float64(-2.672953591751783e-33), np.float64(1.4411128189620413e-33), np.float64(-7.762289489882217e-34), np.float64(4.177834420086499e-34), np.float64(-2.2472421954047727e-34), np.float64(1.2082034851385438e-34), np.float64(-6.493295949833683e-35), np.float64(3.4886690385243327e-35), np.float64(-1.873920920177088e-35), np.float64(1.0063787551275328e-35), np.float64(-5.403903647619717e-36), np.float64(2.901371368500479e-36), np.float64(-1.5576127438409848e-36), np.float64(8.361503811564656e-37), np.float64(-4.488331844034851e-37), np.float64(2.4091647313040135e-37), np.float64(-1.2930981350969868e-37), np.float64(6.940444894140826e-38), np.float64(-3.725045666430735e-38), np.float64(1.9992478203269394e-38), np.float64(-1.073024512134337e-38), np.float64(5.758675031092922e-39), np.float64(-3.090445049083221e-39), np.float64(1.6585738209074663e-39), np.float64(-8.90286473679398e-40), np.float64(4.7779769345748295e-40), np.float64(-2.5643084276302838e-40), np.float64(1.375994719647321e-40), np.float64(-7.356505010928658e-41), np.float64(3.9721636986774256e-41), np.float64(-2.1530281969161777e-41), np.float64(1.1008561858716536e-41), np.float64(-6.179567286233701e-42), np.float64(3.571850271765166e-42), np.float64(-1.558318033583675e-42), np.float64(1.000851064156167e-42), np.float64(-4.520424498600341e-43), np.float64(1.198442877609194e-43)],
+    [np.float64(1.920071671140516e-30), np.float64(-3.714884153679332e-30), np.float64(3.3664601000125984e-30), np.float64(-2.865822195887202e-30), np.float64(2.3012581716345456e-30), np.float64(-1.7516827745379046e-30), np.float64(1.2705583369755341e-30), np.float64(-8.827867076543274e-31), np.float64(5.904986611092253e-31), np.float64(-3.82049389949628e-31), np.float64(2.4011763424403077e-31), np.float64(-1.471709886000766e-31), np.float64(8.827319049249059e-32), np.float64(-5.197463829196592e-32), np.float64(3.012317607943603e-32), np.float64(-1.7226692503047024e-32), np.float64(9.741072742151314e-33), np.float64(-5.4563973803200004e-33), np.float64(3.03235858165363e-33), np.float64(-1.6742232101681503e-33), np.float64(9.19390484807342e-34), np.float64(-5.026424267808094e-34), np.float64(2.7380547827168267e-34), np.float64(-1.4871062671314106e-34), np.float64(8.057524879204305e-35), np.float64(-4.3573383028597396e-35), np.float64(2.3526913168898699e-35), np.float64(-1.268722027041269e-35), np.float64(6.834936950164844e-36), np.float64(-3.6792285111791394e-36), np.float64(1.979264805658882e-36), np.float64(-1.0642234356083949e-36), np.float64(5.719901052264712e-37), np.float64(-3.073318189805369e-37), np.float64(1.650887675428633e-37), np.float64(-8.866298557329596e-38), np.float64(4.76101757897262e-38), np.float64(-2.556266925647616e-38), np.float64(1.3723820867023308e-38), np.float64(-7.367372155594241e-39), np.float64(3.9546502270123034e-39), np.float64(-2.1227462209521985e-39), np.float64(1.1392299716664255e-39), np.float64(-6.116415690807875e-40), np.float64(3.2827483513720865e-40), np.float64(-1.7599101332831905e-40), np.float64(9.46737559726821e-41), np.float64(-5.097930060255915e-41), np.float64(2.6870630131937604e-41), np.float64(-1.4657705287218863e-41), np.float64(7.781199041280875e-42), np.float64(-4.220087252625221e-42), np.float64(2.2258027364840935e-42), np.float64(-1.2089872594585966e-42), np.float64(4.1227309829771716e-43), np.float64(-5.098067461249564e-43), np.float64(3.019808402126974e-43), np.float64(-6.196751855064298e-44), np.float64(1.0530840159013826e-43), np.float64(2.4971742363649453e-43), np.float64(2.003855188771927e-43), np.float64(-4.445263964441535e-43), np.float64(5.829179837909701e-44), np.float64(1.1222931918135973e-43)],
+    [np.float64(-1.4627704846251397e-32), np.float64(2.831676774997611e-32), np.float64(-2.5701812726138254e-32), np.float64(2.1933727161899244e-32), np.float64(-1.766803795592527e-32), np.float64(1.3496703017019418e-32), np.float64(-9.826922272076901e-33), np.float64(6.854217219548874e-33), np.float64(-4.60226905291296e-33), np.float64(2.9885042750040527e-33), np.float64(-1.8846961305046318e-33), np.float64(1.1587932378186544e-33), np.float64(-6.970322664187511e-34), np.float64(4.114589189819142e-34), np.float64(-2.3901161533769367e-34), np.float64(1.3695594917721695e-34), np.float64(-7.757677776863584e-35), np.float64(4.3518243180931164e-35), np.float64(-2.421536000168178e-35), np.float64(1.338393335375388e-35), np.float64(-7.356237510951628e-36), np.float64(4.02471759687906e-36), np.float64(-2.1937302646306843e-36), np.float64(1.1920669287243695e-36), np.float64(-6.461569077052656e-37), np.float64(3.495440091905975e-37), np.float64(-1.887826797532628e-37), np.float64(1.0182575970772986e-37), np.float64(-5.486596736880453e-38), np.float64(2.953838383821007e-38), np.float64(-1.589196899020225e-38), np.float64(8.545715757295949e-39), np.float64(-4.593447372770047e-39), np.float64(2.468132891446256e-39), np.float64(-1.3259269423320973e-39), np.float64(7.1166932841988e-40), np.float64(-3.822231078783727e-40), np.float64(2.0506004345985726e-40), np.float64(-1.1038901551459538e-40), np.float64(5.922642819344176e-41), np.float64(-3.1987884348747365e-41), np.float64(1.6698296325729472e-41), np.float64(-9.385130509196511e-42), np.float64(4.7860700150506506e-42), np.float64(-2.7614277963080914e-42), np.float64(1.1631454570568529e-42), np.float64(-9.956131908903346e-43), np.float64(3.1623367533503366e-43), np.float64(-1.2823022466266157e-43), np.float64(1.4540240986950343e-45), np.float64(-2.837709307569982e-44), np.float64(-2.7656852637152316e-43), np.float64(-2.6375410159495027e-43), np.float64(1.3407590695180913e-43), np.float64(-1.2338716201634958e-43), np.float64(-4.2403972505012315e-43), np.float64(-1.1432552786600802e-43), np.float64(-6.907165113413791e-44), np.float64(-1.2021626549206828e-43), np.float64(-1.6964271961843138e-43), np.float64(-3.27313112068647e-44), np.float64(6.972267482373026e-44), np.float64(-1.343471406468362e-44), np.float64(-2.0382270102184443e-43)],
+    [np.float64(1.0226308004562628e-34), np.float64(-1.9806502199303824e-34), np.float64(1.8003942745282216e-34), np.float64(-1.5399672329708754e-34), np.float64(1.244094828150271e-34), np.float64(-9.535454116736428e-35), np.float64(6.967565428706013e-35), np.float64(-4.877580825416555e-35), np.float64(3.286860684746725e-35), np.float64(-2.1417339271005783e-35), np.float64(1.355086241917144e-35), np.float64(-8.356779575965551e-36), np.float64(5.0405329755787365e-36), np.float64(-2.9827705529031454e-36), np.float64(1.7364486886105617e-36), np.float64(-9.969137747781492e-37), np.float64(5.656303511608179e-37), np.float64(-3.1775766550090603e-37), np.float64(1.7703044451577096e-37), np.float64(-9.794677254406567e-38), np.float64(5.388164740161099e-38), np.float64(-2.950078704549593e-38), np.float64(1.6089705901551178e-38), np.float64(-8.747400018636538e-39), np.float64(4.7433728224101326e-39), np.float64(-2.5668390473469682e-39), np.float64(1.3864892062971537e-39), np.float64(-7.483486835523648e-40), np.float64(4.029147514162724e-40), np.float64(-2.1717401993838258e-40), np.float64(1.1681334793301566e-40), np.float64(-6.280591519546902e-41), np.float64(3.3879801069938614e-41), np.float64(-1.8008639537847828e-41), np.float64(9.918608260676095e-42), np.float64(-4.911377393016602e-42), np.float64(2.744808562754028e-42), np.float64(-1.4281305294140673e-42), np.float64(9.44599572346965e-43), np.float64(-5.733291211029036e-43), np.float64(3.1732384496461817e-43), np.float64(-4.754497253427022e-44), np.float64(2.4494264735431823e-43), np.float64(-9.338872255746089e-44), np.float64(1.2840626134721953e-43), np.float64(2.618437056294341e-43), np.float64(1.2175329192450237e-43), np.float64(-7.928885810552037e-44), np.float64(2.279641714986866e-44), np.float64(9.051965735610348e-44), np.float64(5.041747598734484e-44), np.float64(-6.578704878574725e-44), np.float64(1.7788972623692624e-44), np.float64(-1.9907179201441002e-44), np.float64(1.0471345915891166e-44), np.float64(1.4263283064039205e-43), np.float64(2.8347486435717912e-43), np.float64(1.4771379269524987e-43), np.float64(-8.339188567067874e-44), np.float64(-7.774719623048242e-44), np.float64(-1.9764544187863247e-43), np.float64(-1.6055891608010742e-43), np.float64(-1.0288461140146938e-44), np.float64(1.5741197673474639e-43)],
+])

chebax/_src/series.py

@@ -0,0 +1,195 @@
+"""Runtime evaluation in jax: single and segmented Chebyshev series.
+
+Only two algorithms run per call: Clenshaw, and (for gradients) one more
+Clenshaw on the derivative coefficients. Differentiation and integration act
+on coefficients through matrices built from the numpy references in
+algorithms.py, so they work on traced values and agree with the references
+exactly. Gradients come from a custom_jvp bound to the derivative series:
+jax.grad(p)(x) and p.deriv()(x) are the same computation.
+
+Series are pytrees: coefficients are leaves, domains and breakpoints are
+static. Full float64 accuracy needs jax's x64 mode.
+"""
+
+import functools
+
+import jax
+import jax.numpy as jnp
+import numpy as np
+
+from chebax._src import algorithms
+
+
+@functools.lru_cache(maxsize=None)
+def _der_matrix(n):
+    """chebder as an (n-1, n) matrix (n=1: zero map), for traced coefficients."""
+    if n == 1:
+        return np.zeros((1, 1))
+    return np.array([algorithms.chebder(e) for e in np.eye(n)]).T
+
+
+@functools.lru_cache(maxsize=None)
+def _int_matrix(n):
+    """chebint as an (n+1, n) matrix, for traced coefficients."""
+    return np.array([algorithms.chebint(e) for e in np.eye(n)]).T
+
+
+def _clenshaw(t, coef, coef_last_axis=False):
+    """Clenshaw with static degree. coef_last_axis: per-element coefficient
+    vectors of shape t.shape + (n,), used by the segmented series."""
+    n = coef.shape[-1] if coef_last_axis else coef.shape[0]
+    ck = (lambda k: coef[..., k]) if coef_last_axis else (lambda k: coef[k])
+    b1 = jnp.zeros(jnp.shape(t), jnp.result_type(t, coef))
+    b2 = b1
+    two_t = 2 * t
+    for k in range(n - 1, 0, -1):
+        b1, b2 = two_t * b1 - b2 + ck(k), b1
+    return t * b1 - b2 + ck(0)
+
+
+def _dcoef(coef, domain):
+    a, b = domain
+    d = jnp.asarray(_der_matrix(coef.shape[0]), dtype=coef.dtype)
+    return (2.0 / (b - a)) * (d @ coef)
+
+
+def _chebval_impl(x, coef, domain):
+    a, b = domain
+    t = (2 * x - (b + a)) / (b - a)
+    return _clenshaw(t, coef)
+
+
+# The jvp rule calls the plain impl so the tangent computation is made of
+# ordinary transposable primitives (reverse mode works), and so the value
+# path and the deriv() path are literally the same ops.
+_chebval = jax.custom_jvp(_chebval_impl, nondiff_argnums=(2,))
+
+
+@_chebval.defjvp
+def _chebval_jvp(domain, primals, tangents):
+    x, coef = primals
+    xdot, cdot = tangents
+    y = _chebval_impl(x, coef, domain)
+    dy = _chebval_impl(x, _dcoef(coef, domain), domain) * xdot + _chebval_impl(x, cdot, domain)
+    return y, dy
+
+
+@jax.tree_util.register_pytree_node_class
+class ChebSeries:
+    """sum(c[k] T_k(t)), t the affine image of x taking [a, b] to [-1, 1].
+
+    Plain-c0, lowest-first coefficients (the numpy.polynomial convention).
+    """
+
+    def __init__(self, coef, domain=(-1.0, 1.0)):
+        self.coef = jnp.asarray(coef)
+        self.domain = (float(domain[0]), float(domain[1]))
+
+    @property
+    def degree(self):
+        return self.coef.shape[0] - 1
+
+    def __call__(self, x):
+        return _chebval(jnp.asarray(x), self.coef, self.domain)
+
+    def deriv(self):
+        return ChebSeries(_dcoef(self.coef, self.domain), self.domain)
+
+    def integ(self):
+        """Antiderivative vanishing at the domain midpoint."""
+        a, b = self.domain
+        m = jnp.asarray(_int_matrix(self.coef.shape[0]), dtype=self.coef.dtype)
+        return ChebSeries(((b - a) / 2.0) * (m @ self.coef), self.domain)
+
+    def astype(self, dtype):
+        """Round the coefficients; error floor grows to ~eps(dtype)*sum|c|."""
+        return ChebSeries(self.coef.astype(dtype), self.domain)
+
+    def __repr__(self):
+        return f"ChebSeries(degree={self.degree}, domain={self.domain}, dtype={self.coef.dtype})"
+
+    def tree_flatten(self):
+        return (self.coef,), self.domain
+
+    @classmethod
+    def tree_unflatten(cls, domain, children):
+        obj = object.__new__(cls)
+        obj.coef = children[0]
+        obj.domain = domain
+        return obj
+
+
+def _pw_dcoef(coef, breaks):
+    d = jnp.asarray(_der_matrix(coef.shape[1]), dtype=coef.dtype)
+    widths = np.diff(np.asarray(breaks))
+    return (coef @ d.T) * jnp.asarray(2.0 / widths, dtype=coef.dtype)[:, None]
+
+
+def _pw_chebval_impl(x, coef, breaks):
+    dtype = jnp.result_type(x, coef)
+    br = np.asarray(breaks)
+    idx = jnp.clip(jnp.searchsorted(jnp.asarray(br), x, side="right") - 1, 0, len(breaks) - 2)
+    a = jnp.asarray(br[:-1], dtype)[idx]
+    b = jnp.asarray(br[1:], dtype)[idx]
+    t = (2 * x.astype(dtype) - (b + a)) / (b - a)
+    return _clenshaw(t, coef[idx], coef_last_axis=True)
+
+
+_pw_chebval = jax.custom_jvp(_pw_chebval_impl, nondiff_argnums=(2,))
+
+
+@_pw_chebval.defjvp
+def _pw_chebval_jvp(breaks, primals, tangents):
+    x, coef = primals
+    xdot, cdot = tangents
+    y = _pw_chebval_impl(x, coef, breaks)
+    dy = (_pw_chebval_impl(x, _pw_dcoef(coef, breaks), breaks) * xdot
+          + _pw_chebval_impl(x, cdot, breaks))
+    return y, dy
+
+
+@jax.tree_util.register_pytree_node_class
+class PiecewiseCheb:
+    """Segmented series: row i of coef is a ChebSeries on [breaks[i], breaks[i+1]],
+    all padded to one degree, evaluated by gather + one Clenshaw (no branches).
+    x outside [breaks[0], breaks[-1]] uses the nearest end segment's polynomial.
+    """
+
+    def __init__(self, coef, breaks):
+        self.coef = jnp.asarray(coef)
+        self.breaks = tuple(float(t) for t in breaks)
+        if self.coef.ndim != 2 or self.coef.shape[0] != len(self.breaks) - 1:
+            raise ValueError("coef must have shape (len(breaks) - 1, degree + 1)")
+        if not all(u < v for u, v in zip(self.breaks, self.breaks[1:])):
+            raise ValueError("breaks must be strictly increasing")
+
+    @property
+    def degree(self):
+        return self.coef.shape[1] - 1
+
+    @property
+    def domain(self):
+        return (self.breaks[0], self.breaks[-1])
+
+    def __call__(self, x):
+        return _pw_chebval(jnp.asarray(x), self.coef, self.breaks)
+
+    def deriv(self):
+        return PiecewiseCheb(_pw_dcoef(self.coef, self.breaks), self.breaks)
+
+    def astype(self, dtype):
+        return PiecewiseCheb(self.coef.astype(dtype), self.breaks)
+
+    def __repr__(self):
+        return (f"PiecewiseCheb(segments={len(self.breaks) - 1}, degree={self.degree}, "
+                f"domain={self.domain}, dtype={self.coef.dtype})")
+
+    def tree_flatten(self):
+        return (self.coef,), self.breaks
+
+    @classmethod
+    def tree_unflatten(cls, breaks, children):
+        obj = object.__new__(cls)
+        obj.coef = children[0]
+        obj.breaks = breaks
+        return obj

chebax-0.1.0.dev0.dist-info/METADATA

@@ -0,0 +1,35 @@
+Metadata-Version: 2.4
+Name: chebax
+Version: 0.1.0.dev0
+Summary: Differentiable Chebyshev approximants for special functions, aimed at JAX and GPUs
+Author-email: Andres Hernandez <andres.hernandez.deml@gmail.com>
+License: BSD-3-Clause
+Project-URL: Repository, https://github.com/andreshernandez-spec/chebax
+Requires-Python: >=3.11
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: jax>=0.4.30
+Requires-Dist: numpy>=1.26
+Provides-Extra: gen
+Requires-Dist: mpmath>=1.3; extra == "gen"
+Provides-Extra: test
+Requires-Dist: pytest; extra == "test"
+Dynamic: license-file
+
+# chebax
+
+Differentiable Chebyshev approximants for special functions, aimed at JAX and GPUs.
+A build-time generator (mpmath precision) produces fixed-degree branchless polynomial
+kernels with exact derivative series, including gradients with respect to function
+parameters (e.g. the Bessel order). Prebuilt recipes give out-of-the-box special
+functions; a bake step emits self-contained pure-JAX modules and xsf-style C++ headers.
+
+Early development: the generic core (fit, jax evaluation, exact derivative series,
+segmentation) and the first recipe are in — `besselj(v)` for any real order in
+[0, 10] on x in [0, 8], with dJ/dx via jax.grad and the order gradient via
+`besselj_dnu(v)`, no mpmath at use time. Tails (x > 8), more recipes, and bake
+emitters are not yet. Read
+`PROJECT.md` for the plan and evidence, `CLAUDE.md` for how to work here. Grown out
+of a private research project (the `../bessel/` references in `PROJECT.md` point
+there); the two load-bearing measurements are reproduced here in `experiments/`.
+BSD-3-Clause.

chebax-0.1.0.dev0.dist-info/RECORD

@@ -0,0 +1,14 @@
+chebax/__init__.py,sha256=VqcRUk0Z8AIqOjyDjt-aDXkAX2FIYS1tlcV1J7ZggHQ,439
+chebax/_src/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+chebax/_src/algorithms.py,sha256=zXedVYHoOeEt-4sbJj7skGL3K8ufAJgseaP8ViNhKZI,2261
+chebax/_src/generate.py,sha256=UTW2D-zNqdrTGCJvgnEp2pkvsdz4UD0VMxd0I3yBbUI,4070
+chebax/_src/series.py,sha256=3CKnEjs54VmoHvDcPi0Hb6QBOJxDBKB6EgKM5ZCkcsA,6550
+chebax/_src/recipes/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+chebax/_src/recipes/besselj.py,sha256=9TUm4GSxoPYAsG48MN_Ue2XlClbz7m8VDj-0JhkH5cQ,4411
+chebax/_src/recipes/besselj_gen.py,sha256=qVEBxBkcg9l6A-OV8R3B4uqcrxD2gvrph-nA3xfxMio,2595
+chebax/_src/recipes/besselj_table.py,sha256=UHGvgde1RGjX2AXIZkJlp0uF0wVnjy5BxwqxwWvJH4E,57784
+chebax-0.1.0.dev0.dist-info/licenses/LICENSE,sha256=iy1BeCaZsuO9-d8beVVU73ylzIH4isCFVCvVWsdZk94,1503
+chebax-0.1.0.dev0.dist-info/METADATA,sha256=Hh9VGgtN71Wr5cyGGCld23fnj6ypUEPAr3EMaOgQMgs,1630
+chebax-0.1.0.dev0.dist-info/WHEEL,sha256=K260EYznzXsJYBQGqmI8VTxEdiZYNvDZwW9cBh9-_MA,91
+chebax-0.1.0.dev0.dist-info/top_level.txt,sha256=sqcHOEHa4Axx8YQek95toY12wgM9Sbv_My__N4rcsxw,7
+chebax-0.1.0.dev0.dist-info/RECORD,,

chebax-0.1.0.dev0.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (83.0.0)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

chebax-0.1.0.dev0.dist-info/licenses/LICENSE

@@ -0,0 +1,28 @@
+BSD 3-Clause License
+
+Copyright (c) 2026, Andres Hernandez
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+1. Redistributions of source code must retain the above copyright notice, this
+   list of conditions and the following disclaimer.
+
+2. Redistributions in binary form must reproduce the above copyright notice,
+   this list of conditions and the following disclaimer in the documentation
+   and/or other materials provided with the distribution.
+
+3. Neither the name of the copyright holder nor the names of its
+   contributors may be used to endorse or promote products derived from
+   this software without specific prior written permission.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

chebax-0.1.0.dev0.dist-info/top_level.txt

@@ -0,0 +1 @@
+chebax