def draw_crosscat(self, argin):
"""plot crosscat state
<generator> <modelno> [<options>]
Options
-f, --filename: the output filename. If not specified, tries to draw.
Example:
bayeslite> .ccstate mytable_cc 12 -f state_12.png
"""
parser = ArgumentParser(prog='.drawcc')
parser.add_argument('generator', type=str, help='Generator')
parser.add_argument('modelno', type=int, help='Model number to plot.')
parser.add_argument('-f', '--filename', type=str, default=None,
help='output filename')
parser.add_argument('-r', '--row-label-col', type=str, default=None,
help='The name of the column to use for row labels.')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message,))
return
figure = bdbcontrib.draw_crosscat(self._bdb, args.generator,
args.modelno, row_label_col=args.row_label_col)
if args.filename is None:
plt.show()
else:
figure.savefig(args.filename)
plt.close('all')
def histogram(self, argin):
"""plot histogram of one- or two-column table
<query> [<options>]
If two-column, subdivide the first column according to labels in
the second column.
Example (plots overlapping histograms of height for males and females):
bayeslite> .histogram SELECT height, sex FROM humans; --normed --bin 31
"""
parser = ArgumentParser(prog='.histogram')
parser.add_argument('bql', type=str, nargs='+', help='BQL query')
parser.add_argument('-b', '--bins', type=int, default=15,
help='number of bins')
parser.add_argument('--normed', action='store_true',
help='Normalize histograms?')
parser.add_argument('-f', '--filename', type=str, default=None,
help='output filename')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message,))
return
bql = " ".join(args.bql)
figure = bdbcontrib.histogram(self._bdb, bql, args.bins,
args.normed)
if args.filename is None:
plt.show()
else:
figure.savefig(args.filename)
plt.close('all')
def plot_crosscat_chain_diagnostics(self, argin):
"""plot crosscat diagnostic for all models of generator
<diagnostic> <generator> [output_filename]
Valid (crosscat) diagnostics are:
- logscore: log score of the model
- num_views: the number of views in the model
- column_crp_alpha: CRP alpha over columns
Example:
bayeslite> .chainplot logscore dha_cc scoreplot.png
"""
parser = ArgumentParser(prog='.chainplot')
parser.add_argument('diagnostic', type=str, help='Diagnostic name')
parser.add_argument('generator', type=str, help='Generator name.')
parser.add_argument('-f',
'--filename',
type=str,
default=None,
help='output filename')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message, ))
return
figure = bdbcontrib.plot_crosscat_chain_diagnostics(
self._bdb, args.diagnostic, args.generator)
if args.filename is None:
plt.show()
else:
figure.savefig(args.filename)
plt.close('all')
def plot_crosscat_chain_diagnostics(self, argin):
"""plot crosscat diagnostic for all models of generator
<diagnostic> <generator> [output_filename]
Valid (crosscat) diagnostics are:
- logscore: log score of the model
- num_views: the number of views in the model
- column_crp_alpha: CRP alpha over columns
Example:
bayeslite> .chainplot logscore dha_cc scoreplot.png
"""
parser = ArgumentParser(prog='.chainplot')
parser.add_argument('diagnostic', type=str, help='Diagnostic name')
parser.add_argument('generator', type=str, help='Generator name.')
parser.add_argument('-f', '--filename', type=str, default=None,
help='output filename')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message,))
return
figure = bdbcontrib.plot_crosscat_chain_diagnostics(self._bdb,
args.diagnostic, args.generator)
if args.filename is None:
plt.show()
else:
figure.savefig(args.filename)
plt.close('all')
def barplot(self, argin):
"""plot bar-plot of query giving categories and heights
<query> [<options>]
First column specifies names; second column specifies heights.
"""
parser = ArgumentParser(prog='.bar')
parser.add_argument('bql', type=str, nargs='+', help='BQL query')
parser.add_argument('-f',
'--filename',
type=str,
default=None,
help='output filename')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message, ))
return
bql = " ".join(args.bql)
figure = bdbcontrib.barplot(self._bdb, bql)
if args.filename is None:
plt.show()
else:
figure.savefig(args.filename)
plt.close('all')
def heatmap(self, argin):
"""plot clustered heatmap of pairwise matrix
<query> [<options>]
Options
-------
-f, --filename: str
Save as filename
--vmin: float
Minimun value of the colormap.
--vmax: float
Maximum value of the colormap.
Example:
bayeslite> .heatmap ESTIMATE DEPENDENCE PROBABILITY FROM PAIRWISE COLUMNS OF mytable
"""
parser = ArgumentParser(prog='.heatmap')
parser.add_argument('bql', type=str, nargs='+', help='PAIRWISE BQL query')
parser.add_argument('--vmin', type=float, default=None)
parser.add_argument('--vmax', type=float, default=None)
#XXX last-sort has been removed in favor of specifying row/col ordering.
parser.add_argument('--last-sort', action='store_true')
parser.add_argument('-f', '--filename', type=str, default=None,
help='output filename')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message,))
return
bql = " ".join(args.bql)
clustermap = bdbcontrib.heatmap(self._bdb, bql, vmin=args.vmin,
vmax=args.vmax)
if args.filename is None:
plt.show()
else:
clustermap.savefig(args.filename)
plt.close('all')
def barplot(self, argin):
"""plot bar-plot of query giving categories and heights
<query> [<options>]
First column specifies names; second column specifies heights.
"""
parser = ArgumentParser(prog='.bar')
parser.add_argument('bql', type=str, nargs='+', help='BQL query')
parser.add_argument('-f', '--filename', type=str, default=None,
help='output filename')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message,))
return
bql = " ".join(args.bql)
figure = bdbcontrib.barplot(self._bdb, bql)
if args.filename is None:
plt.show()
else:
figure.savefig(args.filename)
plt.close('all')
def estimate_kl_divergence(self, argin):
"""estimate KL divergence of generator from reference
<table> <generator-a> <generator-b> [--targets-cols <...>] [--given-cols <...>] [--n-samples <N>]
Examples:
bayeslite> .est_kl crosscat baxcat --targets height --givens age 1 nationality 17 --n-samples 1000
"""
parser = ArgumentParser(prog='.est_kl')
parser.add_argument('generator_a',
type=str,
help='Name of the reference generator.')
parser.add_argument('generator_b',
type=str,
help='Name of the approximating generator.')
parser.add_argument(
'--targets',
nargs='*',
help='Sequence of target columns to evaluate the log likelhood. '
'By default, all columns in <table> will be used.')
parser.add_argument(
'--givens',
nargs='*',
help='Sequence of columns and observed values to condition on. '
'The required format is [<col> <val>...].')
parser.add_argument(
'--n-samples',
type=int,
help='Number of rows in the dataset to use in the computation. '
'Defaults to all rows.')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message, ))
return
kl = bdbcontrib.estimate_kl_divergence(self._bdb,
args.generator_a,
args.generator_b,
targets=args.targets,
givens=args.givens,
n_samples=args.n_samples)
print kl
def mi_hist(self, argin):
"""plot histogram of mutual information over generator's models
<generator> <col1> <col2> [<options>]
Example:
bayeslite> .mihist sat_generator dry_mass launch_mass -n=1000
"""
parser = ArgumentParser(prog='.mihist')
parser.add_argument('generator', type=str, help='generator name')
parser.add_argument('col1', type=str, help='first column')
parser.add_argument('col2', type=str, help='second column')
parser.add_argument('-n', '--num-samples', type=int, default=1000)
parser.add_argument('-b', '--bins', type=int, default=15,
help='number of bins')
parser.add_argument('-f', '--filename', type=str, default=None,
help='output filename')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message,))
return
fig = bdbcontrib.mi_hist(self._bdb, args.generator, args.col1,
args.col2, num_samples=args.num_samples, bins=args.bins)
if args.filename is None:
plt.show()
else:
fig.savefig(args.filename)
plt.close('all')
def pairplot(self, argin):
"""plot array of plots for all pairs of columns
<query> [<options>]
- continuous-continuous as scatter (optional KDE contour)
- continuous-categorical as violinplot
- categorical-categorical as heatmap
Options:
-f, --filename: the output filename. If not specified, tries
to draw.
-g, --generator: the generator name. Providing the generator
name make help to plot more intelligently.
-s, --shortnames: Use column short names to label facets?
-m, --show-missing: Plot missing values in scatter plots as lines.
-t, --no-tril: Plot the entire square matrix, not just lower triangle.
--show-contour: Turn on contours.
--colorby: The name of a column to use as a marker variable for color.
Example:
bayeslite> .show SELECT foo, baz, quux + glorb FROM mytable
--filename myfile.png
bayeslite> .show ESTIMATE foo, baz FROM mytable_cc -g mytable_cc
"""
parser = ArgumentParser(prog='.show')
parser.add_argument('bql', type=str, nargs='+', help='BQL query')
parser.add_argument('-f', '--filename', type=str, default=None,
help='output filename')
parser.add_argument('-g', '--generator', type=str, default=None,
help='Query generator name')
parser.add_argument('-s', '--shortnames', action='store_true',
help='Use column short names?')
parser.add_argument('--no-contour', action='store_true',
help='Turn off countours (KDE).'),
parser.add_argument('--show-contour', action='store_true',
help='Turn on contours (KDE).')
parser.add_argument('-m', '--show-missing', action='store_true',
help='Plot missing values in scatterplot.')
parser.add_argument('--show-full', action='store_true')
parser.add_argument('--colorby', type=str, default=None,
help='Name of column to use as a dummy variable.')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message,))
return
bql = " ".join(args.bql)
figure = bdbcontrib.pairplot(self._bdb, bql,
generator_name=args.generator, show_contour=args.show_contour,
colorby=args.colorby, show_missing=args.show_missing,
show_full=args.show_full)
if args.filename is None:
plt.show()
else:
figure.savefig(args.filename)
plt.close('all')
def heatmap(self, argin):
"""plot clustered heatmap of pairwise matrix
<query> [<options>]
Options
-------
-f, --filename: str
Save as filename
--vmin: float
Minimun value of the colormap.
--vmax: float
Maximum value of the colormap.
Example:
bayeslite> .heatmap ESTIMATE DEPENDENCE PROBABILITY FROM PAIRWISE COLUMNS OF mytable
"""
parser = ArgumentParser(prog='.heatmap')
parser.add_argument('bql', type=str, nargs='+', help='PAIRWISE BQL query')
parser.add_argument('--vmin', type=float, default=None)
parser.add_argument('--vmax', type=float, default=None)
#XXX last-sort has been removed in favor of specifying row/col ordering.
parser.add_argument('--last-sort', action='store_true')
parser.add_argument('-f',
'--filename',
type=str,
default=None,
help='output filename')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message, ))
return
bql = " ".join(args.bql)
clustermap = bdbcontrib.heatmap(self._bdb,
bql,
vmin=args.vmin,
vmax=args.vmax)
if args.filename is None:
plt.show()
else:
clustermap.savefig(args.filename)
plt.close('all')
def mi_hist(self, argin):
"""plot histogram of mutual information over generator's models
<generator> <col1> <col2> [<options>]
Example:
bayeslite> .mihist sat_generator dry_mass launch_mass -n=1000
"""
parser = ArgumentParser(prog='.mihist')
parser.add_argument('generator', type=str, help='generator name')
parser.add_argument('col1', type=str, help='first column')
parser.add_argument('col2', type=str, help='second column')
parser.add_argument('-n', '--num-samples', type=int, default=1000)
parser.add_argument('-b',
'--bins',
type=int,
default=15,
help='number of bins')
parser.add_argument('-f',
'--filename',
type=str,
default=None,
help='output filename')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message, ))
return
fig = bdbcontrib.mi_hist(self._bdb,
args.generator,
args.col1,
args.col2,
num_samples=args.num_samples,
bins=args.bins)
if args.filename is None:
plt.show()
else:
fig.savefig(args.filename)
plt.close('all')
def histogram(self, argin):
"""plot histogram of one- or two-column table
<query> [<options>]
If two-column, subdivide the first column according to labels in
the second column.
Example (plots overlapping histograms of height for males and females):
bayeslite> .histogram SELECT height, sex FROM humans; --normed --bin 31
"""
parser = ArgumentParser(prog='.histogram')
parser.add_argument('bql', type=str, nargs='+', help='BQL query')
parser.add_argument('-b',
'--bins',
type=int,
default=15,
help='number of bins')
parser.add_argument('--normed',
action='store_true',
help='Normalize histograms?')
parser.add_argument('-f',
'--filename',
type=str,
default=None,
help='output filename')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message, ))
return
bql = " ".join(args.bql)
figure = bdbcontrib.histogram(self._bdb, bql, args.bins, args.normed)
if args.filename is None:
plt.show()
else:
figure.savefig(args.filename)
plt.close('all')
def draw_crosscat(self, argin):
"""plot crosscat state
<generator> <modelno> [<options>]
Options
-f, --filename: the output filename. If not specified, tries to draw.
Example:
bayeslite> .ccstate mytable_cc 12 -f state_12.png
"""
parser = ArgumentParser(prog='.drawcc')
parser.add_argument('generator', type=str, help='Generator')
parser.add_argument('modelno', type=int, help='Model number to plot.')
parser.add_argument('-f',
'--filename',
type=str,
default=None,
help='output filename')
parser.add_argument('-r',
'--row-label-col',
type=str,
default=None,
help='The name of the column to use for row labels.')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message, ))
return
figure = bdbcontrib.draw_crosscat(self._bdb,
args.generator,
args.modelno,
row_label_col=args.row_label_col)
if args.filename is None:
plt.show()
else:
figure.savefig(args.filename)
plt.close('all')
def pairplot(self, argin):
"""plot array of plots for all pairs of columns
<query> [<options>]
- continuous-continuous as scatter (optional KDE contour)
- continuous-categorical as violinplot
- categorical-categorical as heatmap
Options:
-f, --filename: the output filename. If not specified, tries
to draw.
-g, --generator: the generator name. Providing the generator
name make help to plot more intelligently.
-s, --shortnames: Use column short names to label facets?
-m, --show-missing: Plot missing values in scatter plots as lines.
-t, --no-tril: Plot the entire square matrix, not just lower triangle.
--show-contour: Turn on contours.
--colorby: The name of a column to use as a marker variable for color.
Example:
bayeslite> .show SELECT foo, baz, quux + glorb FROM mytable
--filename myfile.png
bayeslite> .show ESTIMATE foo, baz FROM mytable_cc -g mytable_cc
"""
parser = ArgumentParser(prog='.show')
parser.add_argument('bql', type=str, nargs='+', help='BQL query')
parser.add_argument('-f',
'--filename',
type=str,
default=None,
help='output filename')
parser.add_argument('-g',
'--generator',
type=str,
default=None,
help='Query generator name')
parser.add_argument('-s',
'--shortnames',
action='store_true',
help='Use column short names?')
parser.add_argument('--no-contour',
action='store_true',
help='Turn off countours (KDE).'),
parser.add_argument('--show-contour',
action='store_true',
help='Turn on contours (KDE).')
parser.add_argument('-m',
'--show-missing',
action='store_true',
help='Plot missing values in scatterplot.')
parser.add_argument('--show-full', action='store_true')
parser.add_argument('--colorby',
type=str,
default=None,
help='Name of column to use as a dummy variable.')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message, ))
return
bql = " ".join(args.bql)
figure = bdbcontrib.pairplot(self._bdb,
bql,
generator_name=args.generator,
show_contour=args.show_contour,
colorby=args.colorby,
show_missing=args.show_missing,
show_full=args.show_full)
if args.filename is None:
plt.show()
else:
figure.savefig(args.filename)
plt.close('all')
def estimate_kl_divergence(self, argin):
"""estimate KL divergence of generator from reference
<table> <generator-a> <generator-b> [--targets-cols <...>] [--given-cols <...>] [--n-samples <N>]
Examples:
bayeslite> .est_kl crosscat baxcat --targets height --givens age 1 nationality 17 --n-samples 1000
"""
parser = ArgumentParser(prog='.est_kl')
parser.add_argument('generator_a', type=str,
help='Name of the reference generator.')
parser.add_argument('generator_b', type=str,
help='Name of the approximating generator.')
parser.add_argument('--targets', nargs='*',
help='Sequence of target columns to evaluate the log likelhood. '
'By default, all columns in <table> will be used.')
parser.add_argument('--givens', nargs='*',
help='Sequence of columns and observed values to condition on. '
'The required format is [<col> <val>...].')
parser.add_argument('--n-samples', type=int,
help='Number of rows in the dataset to use in the computation. '
'Defaults to all rows.')
try:
args = parser.parse_args(shlex.split(argin))
except ArgparseError as e:
self.stdout.write('%s' % (e.message,))
return
kl = bdbcontrib.estimate_kl_divergence(self._bdb, args.generator_a,
args.generator_b, targets=args.targets, givens=args.givens,
n_samples=args.n_samples)
print kl