cwindolf

def sliding_mean_and_stddev_interval(
    axis, xs, ys, n_neighbors=30, linewidth=1, color="k", ci_alpha=0.25
):
    """Plot moving means and standard deviations"""
    _1 = np.ones(2 * n_neighbors + 1)
    _ys = np.nan_to_num(ys)
    sliding_sum = correlate1d(_ys, _1, mode="constant")
    sliding_N = correlate1d(np.isfinite(ys).astype(float), _1, mode="constant")
    sliding_mean = sliding_sum / sliding_N
    sliding_sumsq = correlate1d(_ys**2, _1, mode="constant")

cwindolf

import torch
from torch import nn
from tqdm.auto import trange


class PPCA(nn.Module):
    def __init__(self, d, c):
        super().__init__()
        self.d = d
        self.c = c

cwindolf

import matplotlib.pyplot as plt
import contextlib

@contextlib.contextmanager
def subplots(*args, **kwargs):
    fig, axes = plt.subplots(*args, **kwargs)
    try:
        yield fig, axes
    finally:
        plt.show()

cwindolf

def runs_to_ranges(x, one_more=False):
    ranges = []
    cur = x[0]
    b = x[0]
    for a, b in zip(x, x[1:]):
        assert b > a
        if b - a == 1:
            continue
        else:
            ranges.append(range(cur, a + 1 + one_more))

cwindolf

%config InlineBackend.figure_format = 'retina'
import matplotlib.pyplot as plt
from matplotlib.markers import MarkerStyle
from matplotlib.transforms import offset_copy
from matplotlib.patches import Ellipse, Rectangle, ConnectionPatch
from matplotlib.lines import Line2D
from matplotlib.legend_handler import HandlerTuple
import contextlib

plt.rc("figure", dpi=300)

cwindolf

# https://stackoverflow.com/a/74706214/3979938
from matplotlib.transforms import offset_copy

def inline_xlabel(ax, label):
    t = offset_copy(
        ax.transAxes,
        y=-(ax.xaxis.get_tick_padding() + ax.xaxis.get_tick_space()), 
        fig=fig,
        units='points'
    )

cwindolf

try:
    import torch
    import torch.nn.functional as F
    HAVE_TORCH = True
except ImportError:
    HAVE_TORCH = False

def normxcorr1d(
    template,
    x,

cwindolf

def regline(x, y, ax=None, **kwargs):
    b = ((x - x.mean()) * (y - y.mean())).sum() / np.square(x - x.mean()).sum()
    a = y.mean() - b * x.mean()
    x0, x1 = ax.get_xlim()
    r = np.corrcoef(x, y)[0, 1]
    ax.plot([x0, x1], [a + b * x0, a + b * x1], lw=1)
    ax.text(
        0.1,
        0.9,
        f"$\\rho={r:.2f}$",

cwindolf

def tukey_scatter(x, y, iqrs=1.5, ax=None, **kwargs):
    ax = ax or plt.gca()
    x_25, x_75 = np.percentile(x, [25, 75])
    x_iqr = x_75 - x_25
    y_25, y_75 = np.percentile(x, [25, 75])
    y_iqr = y_75 - y_25
    inliers = np.flatnonzero(
        (x_25 - iqrs * x_iqr < x)
        & (x < x_75 + iqrs * x_iqr)
        & (y_25 - iqrs * y_iqr < y)

cwindolf

from pathlib import Path
import cmdstanpy
import ujson

def stanc(name, code, workdir=".stan"):
    Path(workdir).mkdir(exist_ok=True)
    path = Path(workdir) / f"{name}.stan"
    with open(path, "w") as f:
        f.write(code)
    model = cmdstanpy.CmdStanModel(stan_file=path, stanc_options={'warn-pedantic': True})