def setUp(self):
self.dir = temp_path()
self.fn = self.dir / "test_read_table.csv"
self.files = [self.fn.with_suffix(s) for s in [".csv"]] # ".xlsx",
self.expect = pd.DataFrame(np.ones((2, 2)),
columns=["C1", "C2"],
index=["i0", "i1"])
for fn, ex in zip(self.files,
["to_csv"]): # make some csvs # "to_excel",
kw = dict() # engine="openpyxl"
getattr(self.expect, ex)(str(fn),
**subkwargs(kw, getattr(self.expect, ex)))
def add_to_axes(self, ax, label=False, **kwargs):
"""
Plot this point on an :class:`~matplotlib.axes.Axes`.
Parameters
----------
ax : :class:`~matplotlib.axes.Axes`.
Axes to plot the line on.
Returns
--------
:class:`matplotlib.collections.PathCollection`
PathCollection as plotted on the axes.
"""
return ax.scatter(self.x,
self.y,
label=[None, self.name][label],
**{
**self.kwargs,
**subkwargs(kwargs, ax.scatter, PathCollection)
})
def add_to_axes(self, ax, xs=None, label=False, **kwargs):
"""
Plot this line on an :class:`~matplotlib.axes.Axes`.
Parameters
----------
ax : :class:`~matplotlib.axes.Axes`.
Axes to plot the line on.
xs : :class:`numpy.ndarray`
X values at which to evaluate the line function.
Returns
--------
:class:`matplotlib.lines.Line2D`
Lines as plotted on the axes.
Todo
-----
* Update to draw lines based along points along their length
Notes
------
* If no x values are specified, the function will attempt to use the
validity limits of the line, or finally use the limits of the axes.
"""
if xs is None and self.xlim is not None:
xmin, xmax = self.xlim
elif xs is None and self.xlim is None: # use the axes limits
xmin, xmax = ax.get_xlim()
else:
xmin, xmax = np.nanmin(xs), np.nanmax(xs)
if xs is None:
linexs = np.logspace(np.log(xmin), np.log(xmax), 100, base=np.e)
else:
linexs = xs
xmin, xmax = max(xmin, np.nanmin(linexs)), min(xmax, np.nanmax(linexs))
ybounds = [self(xmin), self(xmax)]
ymin, ymax = min(*ybounds), max(*ybounds)
if self.ylim is not None:
ymin, ymax = max(self.ylim[0], ymin), min(self.ylim[1], ymax)
if not xmin > xmax:
if np.abs(self.slope) > 1.0: # more vertical than horizonal
lineys = np.logspace(np.log(ymin),
np.log(ymax),
xs.size,
base=np.e)
linexs = self.out_tfm(self.invfunc(self.in_tfm(lineys)))
else:
lineys = self.out_tfm(self.func(self.in_tfm(linexs)))
# fltr = np.ones(linexs.shape).astype(bool)
# fltr = (lineys >= ymin) & (lineys <= ymax)
# fltr = (linexs >= xmin) & (linexs <= xmax)
# linexs, lineys = linexs[fltr], lineys[fltr]
# append self-styling to the output, but let it be overridden
return ax.plot(linexs,
lineys,
label=[None, self.name][label],
**{
**self.kwargs,
**subkwargs(kwargs, ax.plot, Line2D)
})
def plot_mapping(X,
Y,
mapping=None,
ax=None,
cmap=None,
alpha=1.0,
s=10,
alpha_method="entropy",
**kwargs):
"""
Parameters
----------
X : :class:`numpy.ndarray`
Coordinates in multidimensional space.
Y : :class:`numpy.ndarray` | :class:`sklearn.base.BaseEstimator`
An array of targets, or a method to obtain such an array of targets
via :func:`Y.predict`. Transformers with probabilistic output
(via :func:`Y.predict_proba`) will have these probability estimates accounted
for via the alpha channel.
mapping : :class:`numpy.ndarray` | :class:`~sklearn.base.TransformerMixin`
Mapped points or transformer to create mapped points.
ax : :class:`matplotlib.axes.Axes`
Axes to plot on.
cmap : :class:`matplotlib.cm.ListedColormap`
Colormap to use for the classification visualisation (ideally this should be
a discrete colormap unless the classes are organised ).
alpha : :class:`float`
Coefficient for alpha.
alpha_method : :code:`'entropy' or 'kl_div'`
Method to map class probabilities to alpha. :code:`'entropy'` uses a measure of
entropy relative to null-scenario of equal distribution across classes, while
:code:`'kl_div'` calculates the information gain relative to the same
null-scenario.
Returns
-------
ax : :class:`~matplotlib.axes.Axes`
Axes on which the mapping is plotted.
tfm : :class:`~sklearn.base.BaseEstimator`
Fitted mapping transform.
Todo
------
* Option to generate colors for individual classes
This could be based on the distances between their centres in
multidimensional space (or low dimensional mapping of this space),
enabling a continuous (n-dimensional) colormap to be used
to show similar classes, in addition to classification confidence.
"""
X_ = X.copy() # avoid modifying input array
if mapping is None:
tfm = sklearn.manifold.MDS
tfm_kwargs = {k: v for k, v in kwargs.items() if inargs(k, tfm)}
tfm = tfm(n_components=2, metric=True, **tfm_kwargs)
mapped = tfm.fit_transform(X_)
elif isinstance(mapping, str):
if mapping.lower() == "mds":
cls = sklearn.manifold.MDS
kw = dict(n_components=2, metric=True)
elif mapping.lower() == "isomap":
# not necessarily consistent orientation, but consistent shape
cls = sklearn.manifold.Isomap
kw = dict(n_components=2)
elif mapping.lower() == "tsne":
# likely need to optimise!
cls = sklearn.manifold.TSNE
kw = dict(n_components=2)
else:
raise NotImplementedError
tfm = cls(**{**kw, **subkwargs(kwargs, cls)})
mapped = tfm.fit_transform(X_)
elif isinstance(
mapping,
(sklearn.base.TransformerMixin,
sklearn.base.BaseEstimator)): # manifold transforms can be either
tfm = mapping
mapped = tfm.fit_transform(X_)
else: # mapping is already performedata, expect a numpy.ndarray
mapped = mapping
tfm = None
assert mapped.shape[0] == X_.shape[0]
if ax is None:
fig, ax = plt.subplots(1, **kwargs)
if isinstance(Y, (np.ndarray, list)):
c = Y # need to encode alpha here
elif isinstance(Y, (sklearn.base.BaseEstimator)):
# need to split this into multiple methods depending on form of classifier
if hasattr(Y, "predict_proba"):
classes = Y.predict(X_)
cmap = cmap or DEFAULT_DISC_COLORMAP
c = cmap(classes)
ps = Y.predict_proba(X_)
a = alphas_from_multiclass_prob(ps,
method=alpha_method,
alpha=alpha)
c[:, -1] = a
else:
c = Y.predict(X)
ax.scatter(*mapped.T, c=c, s=s, edgecolors="none")
return ax, tfm, mapped