def fixed_point_visualization(simulation_name, rate_name, master_results_dir, query, filename=None):
"""
Visualizes all single values (or those whose names match ones in `value_names` argument)
present in the datastores of parameter search over a fixed set of parameters.
Parameters
----------
rate_name : str
The parameter full name that corresponds to the input rate.
simulation_name : str
The name of the simulation.
master_results_dir : str
The directory where the parameter search results are stored.
query : ParamFilterQuery
ParamFilterQuery filter query instance that will be applied to each datastore before records are retrieved.
value_names : list(str)
List of value names to visualize.
file_name : str
The file name into which to save the resulting figure. If None figure is just displayed.
"""
(parameters, datastores, n) = load_fixed_parameter_set_parameter_search(simulation_name, master_results_dir)
assert len(parameters) == 3
sorted_parameter_indexes = zip(*sorted(enumerate(parameters), key=lambda x: x[1]))[0]
# if value_names isNone lets set it to set of value_names in the first datastore
value_names = set(
[ads.value_name for ads in param_filter_query(datastores[0][1], identifier="SingleValue").get_analysis_result()]
)
# Lets first make sure that the value_names uniqly identify a SingleValue ADS in each DataStore and
# that they exist in each DataStore.
for (param_values, datastore) in datastores:
dsv = query.query(datastore)
for v in value_names:
assert len(param_filter_query(dsv, identifier="SingleValue", value_name=v).get_analysis_result()) == 1, (
"Error, %d ADS with value_name %s found for parameter combination:"
% (
len(param_filter_query(datastore, identifier="SingleValue").get_analysis_result()),
str([str(a) + ":" + str(b) + ", " for (a, b) in zip(parameters, param_values)]),
)
)
parameter_indexes = [0, 1, 2]
# parameter_indexes = parameter_indexes[]
values = {}
def fixed_point_visualization(simulation_name,rate_name,master_results_dir,query,filename=None):
"""
Visualizes all single values (or those whose names match ones in `value_names` argument)
present in the datastores of parameter search over a fixed set of parameters.
Parameters
----------
rate_name : str
The parameter full name that corresponds to the input rate.
simulation_name : str
The name of the simulation.
master_results_dir : str
The directory where the parameter search results are stored.
query : ParamFilterQuery
ParamFilterQuery filter query instance that will be applied to each datastore before records are retrieved.
value_names : list(str)
List of value names to visualize.
file_name : str
The file name into which to save the resulting figure. If None figure is just displayed.
"""
(parameters,datastores,n) = load_fixed_parameter_set_parameter_search(simulation_name,master_results_dir)
assert len(parameters) == 3
sorted_parameter_indexes = zip(*sorted(enumerate(parameters), key=lambda x: x[1]))[0]
# if value_names isNone lets set it to set of value_names in the first datastore
value_names = set([ads.value_name for ads in param_filter_query(datastores[0][1],identifier='SingleValue').get_analysis_result()])
# Lets first make sure that the value_names uniqly identify a SingleValue ADS in each DataStore and
# that they exist in each DataStore.
for (param_values,datastore) in datastores:
dsv = query.query(datastore)
for v in value_names:
assert len(param_filter_query(dsv,identifier='SingleValue',value_name=v).get_analysis_result()) == 1, "Error, %d ADS with value_name %s found for parameter combination:" % (len(param_filter_query(datastore,identifier='SingleValue').get_analysis_result()), str([str(a) + ':' + str(b) + ', ' for (a,b) in zip(parameters,param_values)]))
parameter_indexes = [0,1,2]
#parameter_indexes = parameter_indexes[]
values = {}
def single_value_visualization(simulation_name,
master_results_dir,
query,
value_names=None,
filename=None,
resolution=None,
treat_nan_as_zero=False,
ranges={},
cols=4):
"""
Visualizes all single values (or those whose names match ones in `value_names` argument)
present in the datastores of parameter search over a fixed set of parameters.
Parameters
----------
simulation_name : str
The name of the simulation.
master_results_dir : str
The directory where the parameter search results are stored.
query : ParamFilterQuery
ParamFilterQuery filter query instance that will be applied to each datastore before records are retrieved.
value_names : list(str)
List of value names to visualize.
file_name : str
The file name into which to save the resulting figure. If None figure is just displayed.
resolution : int
If not None data will be plotted on a interpolated grid of size (resolution,...,resolution)
ranges : dict
A dictionary with value names as keys, and tuples of (min,max) ranges as values indicating what range of values should be displayed.
cols : int
The number of columns in which to show plots, default is 4.
"""
(parameters, datastores, n) = load_fixed_parameter_set_parameter_search(
simulation_name,
master_results_dir,
filter=ParamFilterQuery(
ParameterSet({
'ads_unique': False,
'rec_unique': False,
'params': ParameterSet({'identifier': 'SingleValue'})
})))
# Lets first filter out stuff we were asked by user
datastores = [(a, query.query(b)) for a, b in datastores]
sorted_parameter_indexes = zip(
*sorted(enumerate(parameters), key=lambda x: x[1]))[0]
# if value_names isNone lets set it to set of value_names in the first datastore
if value_names == None:
value_names = [
ads.value_name for ads in
param_filter_query(datastores[0][1],
identifier='SingleValue').get_analysis_result()
]
value_names = set(sorted(value_names))
# Lets first make sure that the value_names uniqly identify a SingleValue ADS in each DataStore and
# that they exist in each DataStore.
for (param_values, datastore) in datastores:
for v in value_names:
assert len(
param_filter_query(datastore,
identifier='SingleValue',
value_name=v).get_analysis_result()
) == 1, "Error, %d ADS with value_name %s found for parameter combination:" % (
len(
param_filter_query(
datastore,
identifier='SingleValue').get_analysis_result()),
str([
str(a) + ':' + str(b) + ', '
for (a, b) in zip(parameters, param_values)
]))
rows = math.ceil(1.0 * len(value_names) / cols)
pylab.figure(figsize=(12 * cols, 6 * rows),
dpi=2000,
facecolor='w',
edgecolor='k')
for i, value_name in enumerate(value_names):
pylab.subplot(rows, cols, i + 1)
if len(parameters) == 1:
x = []
y = []
for (param_values, datastore) in datastores:
x.append(param_values[0])
y.append(
float(
param_filter_query(
datastore,
identifier='SingleValue',
value_name=value_name).get_analysis_result()
[0].value))
pylab.plot(x, y)
pylab.plot(x, y, marker='o')
pylab.xlabel(parameters[sorted_parameter_indexes[0]])
pylab.ylabel(value_name)
elif len(parameters) == 2:
x = []
y = []
z = []
for (param_values, datastore) in datastores:
x.append(param_values[sorted_parameter_indexes[0]])
y.append(param_values[sorted_parameter_indexes[1]])
z.append(
float(
param_filter_query(
datastore,
identifier='SingleValue',
value_name=value_name).get_analysis_result()
[0].value))
if treat_nan_as_zero:
z = numpy.nan_to_num(z)
if value_name in ranges:
vmin, vmax = ranges[value_name]
else:
vmin = min(z)
vmax = max(z)
if resolution != None:
xi = numpy.linspace(numpy.min(x), numpy.max(x), resolution)
yi = numpy.linspace(numpy.min(y), numpy.max(y), resolution)
gr = griddata((x, y),
z, (xi[None, :], yi[:, None]),
method='cubic')
pylab.imshow(gr,
interpolation='none',
vmin=vmin,
vmax=vmax,
aspect='auto',
cmap=cm.gray,
origin='lower',
extent=[
numpy.min(x),
numpy.max(x),
numpy.min(y),
numpy.max(y)
])
else:
pylab.scatter(x,
y,
marker='o',
s=300,
c=z,
cmap=cm.jet,
vmin=vmin,
vmax=vmax)
pylab.colorbar()
pylab.xlabel(parameters[sorted_parameter_indexes[0]])
pylab.ylabel(parameters[sorted_parameter_indexes[1]])
else:
raise ValueError("Currently cannot handle more than 2D data")
pylab.title(value_name)
if filename != None:
pylab.savefig(master_results_dir + '/' + filename, bbox_inches='tight')
def multi_curve_visualzition(simulation_name,
master_results_dir,
x_axis_parameter_name,
query,
filename=None,
value_name=None,
treat_nan_as_zero=False):
"""
Parameters
----------
x_axis_parameter_name : str
The parameter that will be varied along x axis.
simulation_name : str
The name of the simulation.
master_results_dir : str
The directory where the parameter search results are stored.
query : ParamFilterQuery
ParamFilterQuery filter query instance that will be applied to each datastore before records are retrieved.
value_name : list(str)
The value name to visualize
file_name : str
The file name into which to save the resulting figure. If None figure is just displayed.
"""
(parameters, datastores,
n) = load_fixed_parameter_set_parameter_search(simulation_name,
master_results_dir)
sorted_parameter_indexes = zip(
*sorted(enumerate(parameters), key=lambda x: x[1]))[0]
print sorted_parameter_indexes
# if value_names isNone lets set it to set of value_names in the first datastore
if value_name == None:
value_name = set([
ads.value_name for ads in
param_filter_query(datastores[0][1],
identifier='SingleValue').get_analysis_result()
])
assert len(value_name) == 1
value_name = list(value_name)[0]
# Lets first make sure that the value_names uniqly identify a SingleValue ADS in each DataStore and
# that they exist in each DataStore.
for (param_values, datastore) in datastores:
dsv = query.query(datastore)
assert len(
param_filter_query(dsv,
identifier='SingleValue',
value_name=value_name).get_analysis_result()
) == 1, "Error, %d ADS with value_name %s found for parameter combination:" % (
len(
param_filter_query(
datastore,
identifier='SingleValue').get_analysis_result()),
str([
str(a) + ':' + str(b) + ', '
for (a, b) in zip(parameters, param_values)
]))
x_axis_parameter_index = parameters.index(x_axis_parameter_name)
pylab.figure(figsize=(24, 12), dpi=2000, facecolor='w', edgecolor='k')
assert len(parameters) == 3
a = numpy.array([param_values for (param_values, datastore) in datastores])
a = a[:, [j for j in [0, 1, 2] if j != x_axis_parameter_index]]
mmax = numpy.max(a, axis=0)
mmin = numpy.min(a, axis=0)
d = {}
for (param_values, datastore) in datastores:
key = param_values[:]
key.pop(x_axis_parameter_index)
t = d.get(tuple(key), ([], []))
dsv = query.query(datastore)
t[0].append(param_values[x_axis_parameter_index])
t[1].append(
float(
param_filter_query(
dsv, identifier='SingleValue',
value_name=value_name).get_analysis_result()[0].value))
d[tuple(key)] = t
x = []
y = []
z = []
pylab.subplot(2, 2, 1)
for k in d.keys():
color = ((k[0] - mmin[0]) / (mmax[0] - mmin[0]),
(k[1] - mmin[1]) / (mmax[1] - mmin[1]), 0)
x.append(k[0])
y.append(k[1])
z.append(color)
pylab.plot(d[k][0], d[k][1])
pylab.plot(d[k][0], d[k][1], marker='o', color=color)
pylab.xlabel(parameters[sorted_parameter_indexes[0]])
pylab.ylabel(value_name)
pylab.subplot(2, 2, 2)
pylab.scatter(x, y, marker='o', s=100, c=z)
parameters.remove(x_axis_parameter_name)
pylab.xlabel(parameters[0])
pylab.ylabel(parameters[1])
# let's add ratio colored plot
x = []
y = []
z = []
pylab.subplot(2, 2, 3)
for k in d.keys():
a = (k[0] - mmin[0]) / (mmax[0] - mmin[0])
b = (k[1] - mmin[1]) / (mmax[1] - mmin[1])
color = [a, b, 0]
if numpy.sqrt(numpy.sum(numpy.power(color, 2))) != 0:
color = color / numpy.sqrt(numpy.sum(numpy.power(color, 2)))
color[2] = numpy.sqrt(a * a + b * b) / numpy.sqrt(2)
x.append(k[0])
y.append(k[1])
z.append(color)
pylab.plot(d[k][0], d[k][1])
pylab.plot(d[k][0], d[k][1], color=color)
pylab.xlabel(parameters[sorted_parameter_indexes[0]])
pylab.ylabel(value_name)
pylab.subplot(2, 2, 4)
pylab.scatter(x, y, marker='o', s=100, c=z)
pylab.xlabel(parameters[0])
pylab.ylabel(parameters[1])
if filename != None:
pylab.savefig(master_results_dir + '/' + filename)
def single_value_visualization(simulation_name,master_results_dir,query,value_names=None,filename=None,resolution=None,treat_nan_as_zero=False,ranges={},cols=4):
"""
Visualizes all single values (or those whose names match ones in `value_names` argument)
present in the datastores of parameter search over a fixed set of parameters.
Parameters
----------
simulation_name : str
The name of the simulation.
master_results_dir : str
The directory where the parameter search results are stored.
query : ParamFilterQuery
ParamFilterQuery filter query instance that will be applied to each datastore before records are retrieved.
value_names : list(str)
List of value names to visualize.
file_name : str
The file name into which to save the resulting figure. If None figure is just displayed.
resolution : int
If not None data will be plotted on a interpolated grid of size (resolution,...,resolution)
ranges : dict
A dictionary with value names as keys, and tuples of (min,max) ranges as values indicating what range of values should be displayed.
cols : int
The number of columns in which to show plots, default is 4.
"""
(parameters,datastores,n) = load_fixed_parameter_set_parameter_search(simulation_name,master_results_dir,filter=ParamFilterQuery(ParameterSet({'ads_unique' : False, 'rec_unique' : False, 'params' : ParameterSet({'identifier' : 'SingleValue'})})))
# Lets first filter out stuff we were asked by user
datastores = [(a,query.query(b)) for a,b in datastores]
sorted_parameter_indexes = zip(*sorted(enumerate(parameters), key=lambda x: x[1]))[0]
# if value_names isNone lets set it to set of value_names in the first datastore
if value_names == None:
value_names = [ads.value_name for ads in param_filter_query(datastores[0][1],identifier='SingleValue').get_analysis_result()]
value_names = set(sorted(value_names))
# Lets first make sure that the value_names uniqly identify a SingleValue ADS in each DataStore and
# that they exist in each DataStore.
for (param_values,datastore) in datastores:
for v in value_names:
assert len(param_filter_query(datastore,identifier='SingleValue',value_name=v).get_analysis_result()) == 1, "Error, %d ADS with value_name %s found for parameter combination:" % (len(param_filter_query(datastore,identifier='SingleValue').get_analysis_result()), str([str(a) + ':' + str(b) + ', ' for (a,b) in zip(parameters,param_values)]))
rows = math.ceil(1.0*len(value_names)/cols)
pylab.figure(figsize=(12*cols, 6*rows), dpi=2000, facecolor='w', edgecolor='k')
for i,value_name in enumerate(value_names):
pylab.subplot(rows,cols,i+1)
if len(parameters) == 1:
x = []
y = []
for (param_values,datastore) in datastores:
x.append(param_values[0])
y.append(float(param_filter_query(datastore,identifier='SingleValue',value_name=value_name).get_analysis_result()[0].value))
pylab.plot(x,y)
pylab.plot(x,y,marker='o')
pylab.xlabel(parameters[sorted_parameter_indexes[0]])
pylab.ylabel(value_name)
elif len(parameters) == 2:
x = []
y = []
z = []
for (param_values,datastore) in datastores:
x.append(param_values[sorted_parameter_indexes[0]])
y.append(param_values[sorted_parameter_indexes[1]])
z.append(float(param_filter_query(datastore,identifier='SingleValue',value_name=value_name).get_analysis_result()[0].value))
if treat_nan_as_zero:
z = numpy.nan_to_num(z)
if value_name in ranges:
vmin,vmax = ranges[value_name]
else:
vmin = min(z)
vmax = max(z)
if resolution != None:
xi = numpy.linspace(numpy.min(x),numpy.max(x),resolution)
yi = numpy.linspace(numpy.min(y),numpy.max(y),resolution)
gr = griddata((x,y),z,(xi[None, :], yi[:, None]),method='cubic')
pylab.imshow(gr,interpolation='none',vmin=vmin,vmax=vmax,aspect='auto',cmap=cm.gray,origin='lower',extent=[numpy.min(x),numpy.max(x),numpy.min(y),numpy.max(y)])
else:
pylab.scatter(x,y,marker='o',s=300,c=z,cmap=cm.jet,vmin=vmin,vmax=vmax)
pylab.colorbar()
pylab.xlabel(parameters[sorted_parameter_indexes[0]])
pylab.ylabel(parameters[sorted_parameter_indexes[1]])
else:
raise ValueError("Currently cannot handle more than 2D data")
pylab.title(value_name)
if filename != None:
pylab.savefig(master_results_dir+'/'+filename, bbox_inches='tight')
def multi_curve_visualzition(simulation_name,master_results_dir,x_axis_parameter_name,query,filename=None,value_name=None,treat_nan_as_zero=False):
"""
Parameters
----------
x_axis_parameter_name : str
The parameter that will be varied along x axis.
simulation_name : str
The name of the simulation.
master_results_dir : str
The directory where the parameter search results are stored.
query : ParamFilterQuery
ParamFilterQuery filter query instance that will be applied to each datastore before records are retrieved.
value_name : list(str)
The value name to visualize
file_name : str
The file name into which to save the resulting figure. If None figure is just displayed.
"""
(parameters,datastores,n) = load_fixed_parameter_set_parameter_search(simulation_name,master_results_dir)
sorted_parameter_indexes = zip(*sorted(enumerate(parameters), key=lambda x: x[1]))[0]
print sorted_parameter_indexes
# if value_names isNone lets set it to set of value_names in the first datastore
if value_name == None:
value_name = set([ads.value_name for ads in param_filter_query(datastores[0][1],identifier='SingleValue').get_analysis_result()])
assert len(value_name) == 1
value_name = list(value_name)[0]
# Lets first make sure that the value_names uniqly identify a SingleValue ADS in each DataStore and
# that they exist in each DataStore.
for (param_values,datastore) in datastores:
dsv = query.query(datastore)
assert len(param_filter_query(dsv,identifier='SingleValue',value_name=value_name).get_analysis_result()) == 1, "Error, %d ADS with value_name %s found for parameter combination:" % (len(param_filter_query(datastore,identifier='SingleValue').get_analysis_result()), str([str(a) + ':' + str(b) + ', ' for (a,b) in zip(parameters,param_values)]))
x_axis_parameter_index = parameters.index(x_axis_parameter_name)
pylab.figure(figsize=(24, 12), dpi=2000, facecolor='w', edgecolor='k')
#assert len(parameters) == 3, "We required there to be three changing parameters for this visualization, you provided %d" % (len(parameters))
a = numpy.array([param_values for (param_values,datastore) in datastores])
a = a[:,[j for j in [0,1,2] if j != x_axis_parameter_index]]
mmax = numpy.max(a,axis=0)
mmin = numpy.min(a,axis=0)
d = {}
for (param_values,datastore) in datastores:
key = param_values[:]
key.pop(x_axis_parameter_index)
t = d.get(tuple(key),([],[]))
dsv = query.query(datastore)
t[0].append(param_values[x_axis_parameter_index])
t[1].append(float(param_filter_query(dsv,identifier='SingleValue',value_name=value_name).get_analysis_result()[0].value))
d[tuple(key)] = t
x = []
y = []
z = []
pylab.subplot(2,2,1)
for k in d.keys():
color = ((k[0] - mmin[0]) /(mmax[0] - mmin[0]),
(k[1] - mmin[1]) /(mmax[1] - mmin[1]),
0)
x.append(k[0])
y.append(k[1])
z.append(color)
pylab.plot(d[k][0],d[k][1])
pylab.plot(d[k][0],d[k][1],marker='o',color=color)
pylab.xlabel(parameters[sorted_parameter_indexes[0]])
pylab.ylabel(value_name)
pylab.ylim(0,20)
pylab.subplot(2,2,2)
pylab.scatter(x,y,marker='o',s=100,c=z)
parameters.remove(x_axis_parameter_name)
pylab.xlabel(parameters[0])
pylab.ylabel(parameters[1])
# let's add ratio colored plot
x = []
y = []
z = []
pylab.subplot(2,2,3)
for k in d.keys():
a = (k[0] - mmin[0]) /(mmax[0] - mmin[0])
b = (k[1] - mmin[1]) /(mmax[1] - mmin[1])
color = [a,
b,
0]
if numpy.sqrt(numpy.sum(numpy.power(color,2))) != 0:
color = color / numpy.sqrt(numpy.sum(numpy.power(color,2)))
color[2] = numpy.sqrt(a*a+b*b)/numpy.sqrt(2)
x.append(k[0])
y.append(k[1])
z.append(color)
pylab.plot(d[k][0],d[k][1])
pylab.plot(d[k][0],d[k][1],color=color)
pylab.xlabel(parameters[sorted_parameter_indexes[0]])
pylab.ylabel(value_name)
pylab.subplot(2,2,4)
pylab.scatter(x,y,marker='o',s=100,c=z)
pylab.xlabel(parameters[0])
pylab.ylabel(parameters[1])
if filename != None:
pylab.savefig(master_results_dir+'/'+filename)
