import matplotlib.pyplot as plt
import numpy as np
from scipy.sparse.csgraph import reverse_cuthill_mckee
[docs]
class SparseMatrixGrapher:
def __init__(self, matrix):
[docs]
def graph(
self,
filename=None,
marker_string="c.",
mew=0.5,
ms=1,
alpha=0.8,
width=None,
height=None,
dpi=300,
):
tm = self.matrix.tocoo()
y, x = self.matrix.shape
plt.figure(figsize=(width or x / 1000, height or y / 1000))
ax = plt.axes([0, 0, 1, 1])
# Start from top left corner, not bottom left corner
ax.xaxis.set_ticks_position("none")
ax.yaxis.set_ticks_position("none")
ax.xaxis.set_ticklabels([])
ax.yaxis.set_ticklabels([])
ax.plot(
tm.shape[1] - tm.col, tm.row, marker_string, mew=mew, ms=ms, alpha=alpha
)
ax.set_ylim((0, tm.shape[0]))
ax.set_xlim((0, tm.shape[1]))
plt.box(False)
if filename:
plt.savefig(filename, dpi=dpi)
[docs]
def magnitude_graph(self, filename=None, dpi=600, width=None, height=None):
def get_relative_scores(data):
return np.abs(data) / np.max(np.abs(data))
def get_colors(distances):
cmap = plt.get_cmap("Dark2")
return cmap(distances)
def unroll(data):
return [list(row) for row in data]
nm = reverse_cuthill_mckee(self.matrix)
ro = self.matrix[nm, :][:, nm]
as_coo = ro.tocoo()
y, x = as_coo.shape
colors = unroll(get_colors(get_relative_scores(as_coo.data)))
plt.figure(figsize=(width or x / 1000, height or y / 1000))
ax = plt.axes([0, 0, 1, 1])
plt.scatter(
list(as_coo.shape[1] - as_coo.col),
list(as_coo.row),
s=10,
c=colors,
marker=".",
edgecolors="None",
)
ax.xaxis.set_ticks_position("none")
ax.yaxis.set_ticks_position("none")
ax.xaxis.set_ticklabels([])
ax.yaxis.set_ticklabels([])
ax.set_ylim((0, self.matrix.shape[0]))
ax.set_xlim((0, self.matrix.shape[1]))
plt.box(False)
if filename:
plt.savefig(filename, dpi=dpi)
[docs]
def ordered_graph(self, filename=None, dpi=600, width=None, height=None):
def get_distances(xs, ys):
z = np.abs(xs - ys) / 2
return np.sqrt(2 * z ** 2) / MAX_DIST
def get_colors(distances):
cmap = plt.get_cmap("Dark2")
return cmap(distances)
def unroll(data):
return [list(row) for row in data]
nm = reverse_cuthill_mckee(self.matrix)
ro = self.matrix[nm, :][:, nm]
as_coo = ro.tocoo()
y, x = as_coo.shape
MAX_DIST = np.sqrt(2 * (x / 2.0) ** 2)
colors = unroll(get_colors(get_distances(as_coo.col, as_coo.row)))
plt.figure(figsize=(width or x / 1000, height or y / 1000))
ax = plt.axes([0, 0, 1, 1])
plt.scatter(
list(as_coo.shape[1] - as_coo.col),
list(as_coo.row),
s=10,
c=colors,
marker=".",
edgecolors="None",
)
ax.xaxis.set_ticks_position("none")
ax.yaxis.set_ticks_position("none")
ax.xaxis.set_ticklabels([])
ax.yaxis.set_ticklabels([])
ax.set_ylim((0, self.matrix.shape[0]))
ax.set_xlim((0, self.matrix.shape[1]))
plt.box(False)
if filename:
plt.savefig(filename, dpi=dpi)
