def _update_experiment_comparison_plot(experiment_table_data,
experiment_table_selected_rows,
hidden_columns, anchor, data):
"""The callback to render a new experiment comparison plot.
Triggered when new rows in the experiment table are selected.
or deselected.
"""
if experiment_table_selected_rows is None or len(
experiment_table_selected_rows) == 0:
return [html.Div()]
commit_hash = data["commit_hash"]
selected_experiment_ids = [
experiment_table_data[row]["id"]
for row in experiment_table_selected_rows
]
anchor_data, dimensions = app._rubicon_model.get_dimensions(
commit_hash, selected_experiment_ids, hidden_columns, anchor)
return [
dcc.Graph(figure=go.Figure(
go.Parcoords(
line=dict(
color=anchor_data, colorscale="plasma", showscale=True),
dimensions=dimensions,
)))
]
def updateParCord(begin_year, end_year, selected_region):
selected_year = []
if not isinstance(selected_region, list):
selected_region = [selected_region]
for i in range(int(begin_year), int(end_year) + 1):
selected_year.append(int(i))
filtered_df = df.loc[df['iyear'].isin(selected_year)]
df_by_continent = filtered_df[df['country_txt'].isin(selected_region)]
# print(selected_region)
country_val = []
for i in range(len(selected_region)):
for j in range(len(country)):
if selected_region[i] == country[j]:
country_val.append(country_no[j])
# print(country_val)
figure = go.Figure(data=go.Parcoords(
line=dict(
color=df_by_continent['country'],
colorscale='thermal',
),
dimensions=list([
{
'label': 'Weapon Type',
'range': [0, len(weaptype_no) + 1],
'tickvals': weaptype_no,
'ticktext': weaptype,
'values': df_by_continent['weaptype1'].tolist()
},
{
'label': 'Country',
'range': [min(country_val),
max(country_val) + 1],
'tickvals': country_val,
'ticktext': selected_region,
'values': df_by_continent['country'].tolist()
},
{
'label': 'Attack Type',
'range': [0, len(attack_type_no)],
'tickvals': attack_type_no,
'ticktext': attack_type,
'values': df_by_continent['attacktype1'].tolist()
},
{
'label': 'Success',
'range': [0, 1],
'tickvals': [0, 1],
'ticktext': ['No Success', 'Success'],
'values': df_by_continent['success'].tolist()
},
])))
figure.update_layout(title='Parallel Plot for terrorism data',
plot_bgcolor='#FFFFE0',
paper_bgcolor='#FFFFE0',
height=450)
return figure
def plot_parallel_coordinates(expsdf, colslabels, categcols, tickslabels,
outdir):
dimensions = []
for col in categcols: # categorical columns
colname = colslabels[col]
plotcol = dict(label=colname,
values=expsdf[col],
tickvals=list(range(len(tickslabels[col]))),
ticktext=tickslabels[col])
dimensions.append(plotcol)
dimensions.append(dict(
label='Convergence time',
values=expsdf['t'],
))
dimensions.append(dict(
label='Infected ratio',
values=expsdf['inf'],
))
fig = go.Figure(
data=go.Parcoords(line_color='blue', dimensions=dimensions))
plotpath = pjoin(outdir, 'parallel.html')
plotly.offline.plot(fig, filename=plotpath, auto_open=False)
def traffic_parallel_coords(df, max_value=50000, range=[0.8, 1]):
df = df[df.target < max_value]
dimensions = []
for col in [
'traff_m5', 'traff_m4', 'traff_m3', 'traff_m2', 'traff_m1',
'target'
]:
dimensions.append(
dict(range=[0, max_value],
tickvals=[0, 20, max_value * 0.5, 120, max_value],
label=col,
values=df[col]))
if col == 'target':
dimensions[-1]['constraintrange'] = [
max_value * range[0], max_value * range[1]
]
fig = go.Figure(
data=go.Parcoords(line=dict(color=df['target'],
colorscale=px.colors.sequential.Rainbow,
showscale=True,
cmin=0,
cmax=max_value),
dimensions=list(dimensions)))
# fig.update_layout(template='plotly_dark', paper_bgcolor='rgba(0,0,0,0)',
# plot_bgcolor='rgba(0,0,0,0)', font=dict(color="white", size=18, ), width=1100, height=600)
return fig
def draw_plotly_go_parallel_coordinates_matrix(dataDF):
# Converting the professor attribute into int datatype
dataDF['professor'] = pd.to_numeric(dataDF['professor'])
# Calculate the min, max and range to be used while defining the range
dimensions = [
"lecture", "participants", "professional expertise", "motivation",
"clear presentation", "overall impression"
]
min_max_range = {}
for col in dimensions:
min_max_range[col] = [
dataDF[col].min(), dataDF[col].max(),
np.ptp(dataDF[col])
]
# Reference : https://plotly.com/python/parallel-coordinates-plot/
# Usage : Advanced Parallel Coordinates Plot
figure = go.Figure(data=go.Parcoords(
line=dict(color=dataDF['professor'],
colorsrc="professor",
colorscale='Rainbow',
showscale=False),
dimensions=list([
dict(range=[(dataDF['professor'].min() -
1), (dataDF['professor'].max() + 1)],
label="Professor Number",
values=dataDF['professor']),
dict(range=[(dataDF['lecture'].min() -
1), (dataDF['lecture'].max() + 1)],
label="Lecture Number",
values=dataDF['lecture']),
dict(range=[(dataDF['participants'].min() -
1), (dataDF['participants'].max() + 1)],
label="Participants",
values=dataDF['participants']),
dict(range=[6, 0],
tickvals=[1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6],
label='Professional Expertise',
values=dataDF['professional expertise']),
dict(range=[6, 0],
tickvals=[1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6],
label='Motivation',
values=dataDF['motivation']),
dict(range=[6, 0],
tickvals=[1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6],
label='Clean Presentation',
values=dataDF['clear presentation']),
dict(range=[6, 0],
tickvals=[1, 1.5, 2, 2.5, 3, 3.5, 4, 4.5, 5, 5.5, 6],
label='Overall impression',
values=dataDF['overall impression'])
])))
figure.update_layout(title="Parallel Co-ordinates of Weierstrass-Prize")
figure.show()
def create_graphs(dff, title):
return [{
'data': [{
'x': dff['YearRemodel'],
'name': 'Year Remodel',
'type': 'histogram'
}, {
'x': dff['YearBuilt'],
'name': 'Year Built',
'type': "histogram"
}],
'layout': {
'plot_bgcolor': colors['background'],
'paper_bgcolor': colors['background'],
'title': title + " Neighborhood(s)",
'font': {
'color': colors['text']
}
},
}, {
'data': [
go.Parcoords(line=dict(color=dff['SalePrice'],
colorscale='Electric',
showscale=True,
cmin=df['SalePrice'].min(),
cmax=df['SalePrice'].max()),
dimensions=list([
dict(label='SalePrice',
values=dff['SalePrice'].unique()),
dict(label='"OverallCond"',
values=df["OverallCond"].unique()),
dict(label="GrLivArea",
values=df["GrLivArea"].unique()),
dict(label="TotRmsAbvGrd",
values=df["TotRmsAbvGrd"].unique()),
dict(label="YrSold", values=df["YrSold"].unique())
]))
],
'layout':
dict(title=title + " Neighborhood",
plot_bgcolor=colors['background'],
paper_bgcolor=colors['background'])
}, {
'data': [
go.Heatmap(x=dff.columns.values[8:],
y=dff['Neighborhood'].unique(),
z=dff[[
"GrLivArea", "GarageArea", "TotRmsAbvGrd",
"TotalBsmtSF", "FullBath"
]].values,
colorscale='Electric',
colorbar={"title": "Square feet"},
showscale=True)
],
"layout":
go.Layout(title=title + " Neighborhood",
plot_bgcolor=colors['background'],
paper_bgcolor=colors['background'])
}]
def parallel_coordinates(save_dir, metric="loss"):
# DO NOT CHANGE ANYTHING IN THIS FUNCTION
# This function take a folder path and loads all model files,
# extracts hyperparamaters and the given score, and creates a
# parallell coordination plot to display the scores for each hyperparamater
model_files = glob(save_dir + "/*.pt")
print("models found:", model_files)
rows = []
for model_file in model_files:
model_dict = torch.load(model_file)
score_dict = model_dict["scores"]
score_dict["loss"] = model_dict["loss"]
score_dict["loss"] = score_dict["loss"].cpu().detach().numpy()
score_dict["loss"] = score_dict["loss"].item()
row = {"model_name": model_dict["model_name"]}
row.update(model_dict["hyperparamaters"])
row[metric] = score_dict[metric]
rows.append(row)
df = pd.DataFrame(rows)
y_dims = []
for cname, cdata in df.iteritems():
values = list(cdata.values)
add_text_ticks = False
if isinstance(values[0], str):
ticktexts = sorted(set(values))
ticktext2id = {v: i for i, v in enumerate(ticktexts)}
tickvals = list(ticktext2id.values())
values = [ticktext2id[v] for v in values]
add_text_ticks = True
max_v = np.max(values)
min_v = np.min(values)
y_dict = {"values": values, "label": cname, "range": [min_v, max_v]}
if add_text_ticks:
y_dict["ticktext"] = ticktexts
y_dict["tickvals"] = tickvals
y_dims.append(y_dict)
fig = go.Figure(data=go.Parcoords(
line=dict(color=df[metric],
colorscale='Electric',
showscale=True,
cmin=df[metric].min(),
cmax=df[metric].max()),
dimensions=y_dims,
))
fig.update_layout(margin=dict(l=120, r=30, b=20, t=40))
fig.show()
def create_parcoords_fig(data=None, columns=None):
parcoords = go.Figure(
data=go.Parcoords(line_color='red',
dimensions=list([
dict(range=[data[col].min(), data[col].max()],
label=col,
values=data[col]) for col in columns
if isinstance(data[col].iloc[0], numbers.Number)
])))
return parcoords
def _value_path_plot_candidates(
self,
zs: np.ndarray,
names: Optional[List[str]] = None,
labels: Optional[List[str]] = None) -> Figure:
"""Plots multiple solution candidates as an parallel axis plot.
Args:
zs (np.ndarray): A 2D array with each candidate on its' rows.
names (List[str], optional): A list of the objective names.
"""
if zs.ndim == 1:
zs = zs.reshape(1, -1)
zs_original = self.scaler.inverse_transform(zs)
zs_original = np.where(self.is_max, -zs_original, zs_original)
nadir_original = self.scaler.inverse_transform(
self.nadir.reshape(1, -1))[0]
nadir_original = np.where(self.is_max, -nadir_original, nadir_original)
ideal_original = self.scaler.inverse_transform(
self.ideal.reshape(1, -1))[0]
ideal_original = np.where(self.is_max, -ideal_original, ideal_original)
if names is None:
names = ["Obj {}".format(i + 1) for i in range(zs.shape[1])]
names = [
"{} ({})".format(name, val)
for (val,
name) in zip(["⇧" if m is True else "⇩"
for m in self.is_max], names)
]
rows = [list(row) for row in zs_original.T]
fig = go.Figure(data=go.Parcoords(dimensions=list(
# stupid hack to "label" each solution on the first axis
[
dict(
range=[1, len(zs_original)],
label="Candidate",
values=list(range(1,
len(zs_original) + 1)),
tickvals=list(range(1,
len(zs_original) + 1)),
)
] + [
dict(range=[low, up], label=name, values=vals)
for (low, up, name,
vals) in zip(nadir_original, ideal_original, names, rows)
])))
return fig
def plot_optimization(fname='report_sim2_5DoF'):
"""
make parallel coordinates plot for the hyperparameter optimization results
"""
df = pd.read_csv(f'./tests/optimization_logs/{fname}.csv')
df['value'] = -df['value']
df['params_lr'] = np.log10(df['params_lr'])
df['params_buffer_size'] = np.log10(df['params_buffer_size'])
df['params_net_arch'] = df['params_net_arch'].replace(
['small', 'medium', 'big'], [16, 32, 64])
data = [
go.Parcoords(
line=dict(color=df['value'],
showscale=True,
colorscale='thermal',
colorbar={'title': 'Return'},
cmin=-1000,
cmax=-500),
dimensions=list([
dict(range=[np.log10(1e-4), np.log10(1e-2)],
tickvals=[-4., -3., -2],
ticktext=['1E-4', '1E-3', '1E-2'],
label='Learning Rate',
values=df['params_lr']),
dict(tickvals=[256, 512],
range=[200, 550],
label='Minibatch Size',
values=df['params_batch_size']),
dict(tickvals=[1, 50, 100],
range=[0, 100],
label='Train Frequency',
values=df['params_train_freq']),
dict(tickvals=[np.log10(5e4),
np.log10(1e6),
np.log10(2e6)],
ticktext=['5E4', '1E6', '2E6'],
range=[np.log10(2.8e4), np.log10(5e6)],
label='Buffer Size',
values=df['params_buffer_size']),
dict(label='Network Width',
values=df['params_net_arch'],
tickvals=[16, 32, 64]) # , range = [1,3],)
]))
]
fig = go.FigureWidget(data=data)
fig.update_layout(template="plotly")
fig.write_image("./figures/fig_par_coords2.pdf")
return
def parameter_graph(self):
if self.performance_metric_selector and len(
self.hyperparameter_selector) > 0:
df = self.results[(
self.results['pipeline_name'].isin(self.pipeline_run_selector))
& (self.results['slice_name'] == '')]
# merge
extra_df = pd.merge(self.hparam_info,
df,
on='pipeline_name',
how='left')
dimensions = ['pipeline_name'] + self.hyperparameter_selector + \
[self.performance_metric_selector]
new_dims = []
for d in dimensions:
try:
new_dims.append({
'label': d,
'values': pd.to_numeric(extra_df[d])
})
except:
u = sorted(
list(extra_df[d].apply(lambda x: str(x)).unique()))
mapping = {v: i for i, v in enumerate(u)}
new_dims.append({
'label':
d,
'tickvals': [mapping[x] for x in u],
'ticktext':
u,
'values':
extra_df[d].apply(lambda x: str(x)).map(mapping)
})
final_col = pd.to_numeric(
extra_df[self.performance_metric_selector])
fig = go.Figure(data=go.Parcoords(line=dict(color=final_col,
colorscale='inferno',
showscale=True,
cmin=min(final_col),
cmax=max(final_col)),
dimensions=new_dims))
else:
fig = px.scatter(pd.DataFrame(),
marginal_y='rug',
width=1100,
title='Hyperparameter Comparison')
return fig
def adjacency_parallel_coordinate_plot(rules: List):
'''
Visualizes the antecedents and consequents of each rule by drawing lines
representing each rule across identical vertical axis representing the
potential items in the set
Has the advantage of making it easier to visualize compound rules over
scatterplots
'''
#These two structures track the rules and entities therein based on the number of antecedents/consequents involved
#Allows us to visualize each number separately in a parallel coordinate graph
#Note these are indexed 0->2 axis on (no association rule can have less then 2)
unique_entities_by_axis_count = []
rules_by_axis_count = []
for rule in rules:
axis_required = len(rule.lhs) + 1
#Filter out rules with multiple consequents
#TODO consider allowing multiple consequents
if len(rule.rhs) == 1:
#If the rules_by_axis_count list lacks a slot for the current number of antacedents, add them
while len(rules_by_axis_count) < axis_required - 1:
rules_by_axis_count.append([])
unique_entities_by_axis_count.append(set())
#Add the rule and the entities found in its atecedents and consequents to their respective structure
rules_by_axis_count[axis_required - 2].append(rule)
unique_entities_by_axis_count[axis_required -
2] = unique_entities_by_axis_count[
axis_required - 2].union(
set(rule.lhs), set(rule.rhs))
axis_counter = 2
for rules, unique_entities in zip(rules_by_axis_count,
unique_entities_by_axis_count):
unique_entities = list(unique_entities)
unique_entities = _parallel_coord_axis_optimizer(
rules, unique_entities, axis_counter)
line_color = list(map(lambda rule: round(rule.confidence, 2), rules))
dimensions = _paracoord_builder(rules, unique_entities, axis_counter)
fig = go.Figure(data=go.Parcoords(line=dict(
color=line_color, colorscale=[[0, 'white'], [1, 'red']]),
dimensions=dimensions))
fig.update_layout(plot_bgcolor='white', paper_bgcolor='white')
fig.show()
axis_counter += 1
def update_figure(value):
dimensions = []
for axe in value:
dimensions.append(
dict(range=[df[axe].min(), df[axe].max()],
tickvals=[],
label=axe,
values=df[axe]), )
fig = go.Figure(data=go.Parcoords(line=dict(
color=df['Year'],
colorscale='earth',
),
dimensions=dimensions))
return fig
def __create_parallel_coordinates(df_differences, k):
fig_parallel = go.Figure(data=go.Parcoords(
line=dict(color=df_differences['Cluster'],
colorscale=[CLUSTER_COLORS[i] for i in range(k)]
),
dimensions=list([
dict(range=[df_differences[col].min(), df_differences[col].max()],
label=col, values=df_differences[col]) for col in MAIN_COLS])
)
)
fig_parallel.update_layout(
plot_bgcolor='white',
paper_bgcolor='white'
)
return fig_parallel
def Technologies_Parallel_Cordianate(Range):
df = technologies_data[technologies_data["Range Type"] == Range]
b = df["Range Type"]
new_list1 = []
for i in b:
if i not in new_list1:
new_list1.append(i)
a = df["Value"]
new_list = []
for i in a:
if i not in new_list:
new_list.append(i)
fig = go.Figure(data=go.Parcoords(dimensions=list([
dict(tickvals=df["Value"],
ticktext=df.index,
label="Technology",
values=df["Value"]),
dict(range=[min(df['Delay (ms)']),
max(df['Delay (ms)'])],
label='Delay (ms)',
values=df['Delay (ms)']),
dict(range=[min(df['Max Range (m)']),
max(df['Max Range (m)'])],
label='Max Range (m)',
values=df['Max Range (m)']),
dict(range=[min(df['Data Rate (Mb/s)']),
max(df['Data Rate (Mb/s)'])],
label='Data Rate (Mb/s)',
values=df['Data Rate (Mb/s)']),
dict(range=[
min(df['Frequency Band(s) (GHz)']),
max(df['Frequency Band(s) (GHz)'])
],
label='Frequency Band(s) (GHz)',
values=df['Frequency Band(s) (GHz)']),
dict(range=[
min(df['Frequency Band(s) (GHz)']),
max(df['Frequency Band(s) (GHz)'])
],
label='Frequency Band(s) (GHz)',
values=df['Frequency Band(s) (GHz)']),
])))
return fig.show()
def parallel_coordinates(df1,cont_feat):
columns=list(df1.columns)
labelencoder = LabelEncoder()
df = df1.copy()
for i in columns:
if i not in cont_feat:
df[i]=labelencoder.fit_transform(df[i])
new_list=[]
for i in columns:
if i in cont_feat:
x=dict(range = [min(df[i].values.tolist()),max(df[i].values.tolist())],
label = i, values = df[i].values.tolist())
new_list.append(x)
else:
x=dict(range = [min(df[i].values.tolist()),max(df[i].values.tolist())],
tickvals = list(df[i].unique()),
label = i, values = df[i].values.tolist(),
ticktext = list(df1[i].unique()))
new_list.append(x)
fig = go.Figure(
data=go.Parcoords(
line=dict(
color = df[explore.outcome_name],
colorscale = [[0, 'red'], [0.5, 'red'], [0.5, 'blue'], [1, 'blue']],
showscale = True,
colorbar=dict(
lenmode='pixels',
len=75,
tickmode='array',
tickvals=[0.25, 0.75],
ticktext=['0', '1']
)
),
dimensions = list(new_list)
)
)
fig.show()
def change_plot(change):
if only_stable.v_model:
mask = dimensions[0]['values'] == 1
else:
mask = slice(dimensions[0]['values'].size)
new_data = []
for i in items.v_model:
d = dimensions[i]
new_data.append(
dict(values=d['values'][mask], label=d['label']))
fig.children = [
go.FigureWidget(
go.Parcoords(
line=dict(color=dimensions[color.v_model]
['values'][mask]),
dimensions=new_data,
))
]
def parallel_plot(year):
fig = go.Figure(data=go.Parcoords(
line=dict(color=new_combined_hap['Year']),
dimensions=list([
dict(range=[0, 8],
constraintrange=[4, 8],
label='Happiness Score',
values=new_combined_hap['Happiness Score']),
dict(range=[0, 8],
label='Economy (GDP per Capita)',
values=new_combined_hap['Economy (GDP per Capita)']),
dict(range=[0, 8],
label='Family',
values=new_combined_hap['Family']),
dict(range=[0, 8],
label='Health (Life Expectancy)',
values=new_combined_hap['Health (Life Expectancy)'])
])))
fig.update_layout(plot_bgcolor='white', paper_bgcolor='white')
print('parallel done')
return fig
def parallel_coordinates_plot(data, variables, group_var=None):
"""
Create a parallel coordinates plot based on a given dataframe and a list of variable names
Args:
data: dataframe containing the data
variables: list of variable names
group_var: variable name to set the color scales for the created lines
Returns:
"""
# Remove na values
plot_data_vars = variables.copy() + [group_var]
print(plot_data_vars)
plot_data = data[plot_data_vars].copy()
plot_data = plot_data.dropna(how="any")
# create dimensions for parallel
dimensions = []
for var in variables:
label = "<br>".join(textwrap.wrap(var, 30))
print(label)
dim = dict(range=[plot_data[var].min(), plot_data[var].max()],
label=label,
values=plot_data[var])
dimensions.append(dim)
fig = go.Figure(
data=go.Parcoords(line=dict(color=plot_data[group_var],
colorscale=px.colors.diverging.Geyser,
showscale=True),
dimensions=dimensions))
fig.show()
fig.update_layout(width=1200)
fig.write_image("parallel_coordinates.png")
def main():
year = 2014
elo_rating = elo(year)
massey_rating = massey(year)
premier_rating = premier(year)
teams = elo_rating.keys()
ranking_df = pd.DataFrame({
"Team": [],
"Premier League Rating": [],
"Massey Rating": [],
"Elo Rating": []
})
for team in teams:
df_current = pd.Series(
{
"Team": team,
"Premier League Rating": premier_rating[team],
"Massey Rating": massey_rating[team],
"Elo Rating": elo_rating[team]
},
index=None)
ranking_df = ranking_df.append(df_current, ignore_index=True)
# Parallel Plot
fig = go.Figure(
data=go.Parcoords(line=dict(color="#002366"),
dimensions=list([
dict(range=[22, 95],
constraintrange=[4, 8],
label='Premier League',
values=ranking_df['Premier League Rating']),
dict(range=[-1.1, 1.6],
label='Masseys Method',
values=ranking_df['Massey Rating']),
dict(range=[1117, 1300],
label='Elo',
values=ranking_df['Elo Rating'])
])))
fig.update_layout(plot_bgcolor='red', paper_bgcolor='white')
fig.show()
def return_parallel_plot_fig( GEOID_to_plot = "ALL_SEASONS", height = 800):
# 3) Plot:
fig = go.Figure(
data = go.Parcoords(
# LINE COLOR DEF:
line = dict(
color = parallelDF[colorCol], # each lines' color val
colorscale = ["purple", "lightcoral", "red", "firebrick","maroon"], # [counties0->4]
colorbar = dict(
tickvals = [0,1,2,3,4], # [counties0->4]
ticktext = countyColorOrder, # [counties0->4]
title = {'text': "NYC Counties"},
x = -0.5, # ******delete this to turn the colorscale bar to normal position
ticks = "outside",
),
showscale = True, # show the scale
cmin = parallelDF[colorCol].min(),
cmax = parallelDF[colorCol].max(),
),
dimensions = making_dimension_list_for_parallel(GEOID_to_plot),
labelangle = -45,
)
)
fig.update_layout(
title = "Focused look on Season Gas Leak Reports and Total Crime of all GEOID in NYC",
autosize=False,
width = widthPlot,#1000,
height = height,#700,
margin=dict(
l=10,
r=100,
b=10,
t=200,
pad=4
),
)
return fig
def change_plot(self, change):
if self.only_stable.v_model:
mask = self.results[0]['values'] == 1
else:
mask = slice(self.results[0]['values'].size)
new_data = []
for i in self.items.v_model:
d = self.results[i]
new_data.append(dict(values=d['values'][mask], label=d['label']))
self.fig.children = [
go.FigureWidget(
go.Parcoords(
line=dict(color=self.results[self.color.v_model]['values']
[mask]),
dimensions=new_data,
))
]
self.plotly_plot.children = [
self.color, self.items, self.only_stable, self.fig
]
def repr(self) -> go.Figure:
fig = go.Figure()
# rearange colors
colors = []
for idx, ens in enumerate(self._colors):
colors.append([idx / len(self._colors), self._colors[ens]])
colors.append([(idx + 1) / len(self._colors), self._colors[ens]])
# create a single dataframe, where all ensemble parameters are stacked
data_df = pd.concat(list(self.data.values()))
# drop a column with ids
ensemble_ids = data_df["ensemble_id"]
data_df = data_df.drop("ensemble_id", axis=1)
# create parallel coordinates dimension list
dimensions = []
for parameter in data_df:
dimensions.append(dict(label=parameter, values=data_df[parameter]))
fig.add_trace(
go.Parcoords(
line=dict(
color=ensemble_ids,
colorscale=colors,
showscale=True,
colorbar=dict(
tickvals=list(range(0, len(self.data))),
ticktext=list(self.data.keys()),
len=0.2 * len(self.data),
title="Ensemble",
),
),
dimensions=dimensions,
labelangle=45,
labelside="top",
))
fig.update_layout(clickmode="event+select")
fig.update_layout(uirevision=True)
return fig
def pc_imputed(df, imp_arr, df_excl, save_fig=True, dataset=None):
# Requires column named imputed
subset = list(df.columns)
for item in list(df_excl.columns):
subset.remove(item)
df = df[subset]
df = vis_encode_cols(df.copy())
imp_arr_num = np.zeros(len(imp_arr))
imp_arr_num[imp_arr] = 1
df['imputed'] = imp_arr_num
dim = []
for col in df.columns[:-1]:
col_dict = {}
col_dict['range'] = [min(df[col]), max(df[col])]
col_dict['label'] = col
col_dict['values'] = df[col]
dim.append(col_dict)
fig = go.Figure(
data=go.Parcoords(line=dict(color=df['imputed'],
colorscale="Bluered",
colorbar={
'tickvals': [0, 1],
'ticktext': ["not imputed", "imputed"]
}),
dimensions=dim))
fig.update_layout(title="PC plot visualizing the imputed datapoints")
if save_fig:
html_path = "vis/" + dataset[:-4] + "_parallel_coordinates.html"
fig.write_html(html_path)
fig.show()
def update_parallel_figure(selected_color):
week_text = [i for i in week_dict]
week_vals = [week_dict[i] for i in week_text]
pd_text = [i for i in pd_dict]
pd_vals = [pd_dict[i] for i in pd_text]
res_text = [i for i in res_dict]
res_vals = [res_dict[i] for i in res_text]
color_text = [i for i in color_dict]
color_vals = [color_dict[i] for i in color_text]
fig = go.Figure(data=go.Parcoords(
line=dict(color=df['Color'],
colorscale=selected_color,
showscale=True,
cmin=min(color_code),
cmax=max(color_code),
colorbar=dict(tickvals=color_vals, ticktext=color_text)),
dimensions=list([
dict(tickvals=week_vals,
ticktext=week_text,
label='Day of Week',
values=df['Week']),
dict(tickvals=pd_vals,
ticktext=pd_text,
label='Police Department District',
values=df['Pd']),
dict(tickvals=res_vals,
ticktext=res_text,
label='Resolution',
values=df['Res'])
])))
return fig
def run_parametric_study():
from pathos.multiprocessing import ProcessingPool
pool = pp.ProcessPool()
output = pool.map(run_simulation, args)
dimensions = [
dict(values=np.asarray([o[0] for o in output],
dtype=np.float64),
label='stability')
]
dimensions.extend([
dict(values=sampling[:, ik], label=f'{k}')
for ik, k in enumerate(design_space.keys())
])
for i, r in enumerate(self.responses.widget.v_model):
if output[0][i + 1] is not None:
dimensions.append(
dict(values=np.asarray([o[i + 1] for o in output],
dtype=np.float64),
label=r))
for isamp in range(len(sampling)):
tmp_design = {
f'{k}': sampling[isamp, ik]
for ik, k in enumerate(design_space.keys())
}
tmp_responses = {
r: output[isamp][ir + 1]
for ir, r in enumerate(self.responses.widget.v_model)
}
tmp_responses['stability'] = output[isamp][0]
simu_path = os.path.join(path, f'simu_{isamp}')
save_param_study_for_simu(simu_path, 'param_study.json',
tmp_design, tmp_responses)
fig = v.Row(children=[
go.FigureWidget(data=go.Parcoords(
line=dict(color=dimensions[0]['values']),
dimensions=dimensions,
)),
],
align='center',
justify='center')
def change_plot(change):
if only_stable.v_model:
mask = dimensions[0]['values'] == 1
else:
mask = slice(dimensions[0]['values'].size)
new_data = []
for i in items.v_model:
d = dimensions[i]
new_data.append(
dict(values=d['values'][mask], label=d['label']))
fig.children = [
go.FigureWidget(
go.Parcoords(
line=dict(color=dimensions[color.v_model]
['values'][mask]),
dimensions=new_data,
))
]
color = v.Select(label='color',
items=[{
'text': v['label'],
'value': i
} for i, v in enumerate(dimensions)],
v_model=0)
items = v.Select(label='Show items',
items=[{
'text': v['label'],
'value': i
} for i, v in enumerate(dimensions)],
v_model=[i for i in range(len(dimensions))],
multiple=True)
only_stable = v.Switch(label='Show only stable results',
v_model=False)
color.observe(change_plot, 'v_model')
items.observe(change_plot, 'v_model')
only_stable.observe(change_plot, 'v_model')
self.plotly_plot.children = [color, items, only_stable, fig]
self.stop_simulation(None)
)
fig_parallel2 = go.Figure(data=go.Parcoords(
line=dict(color=df2['colorVal'],
colorscale='Electric',
showscale=True,
cmin=-4000,
cmax=-100),
dimensions=list([
dict(range=[32000, 227900],
constraintrange=[100000, 150000],
label="Block Height",
values=df2['blockHeight']),
dict(range=[0, 700000], label='Block Width', values=df2['blockWidth']),
dict(tickvals=[0, 0.5, 1, 2, 3],
ticktext=['A', 'AB', 'B', 'Y', 'Z'],
label='Cyclinder Material',
values=df2['cycMaterial']),
dict(range=[-1, 4],
tickvals=[0, 1, 2, 3],
label='Block Material',
values=df2['blockMaterial']),
dict(range=[134, 3154],
visible=True,
label='Total Weight',
values=df2['totalWeight']),
dict(range=[9, 19984],
label='Assembly Penalty Wt',
values=df2['assemblyPW']),
dict(
range=[49000, 568000], label='Height st Width', values=df2['HstW'])
])))
def get_parallel_coordinate_plot(study=None,coloring="blues",params= None,objective_value="Loss",logLoss=True):
layout = go.Layout(title="Parallel Coordinate Plot",)
trials = [trial for trial in study.trials if trial.state == TrialState.COMPLETE]
if len(trials) == 0:
_logger.warning("Your study does not have any completed trials.")
return go.Figure(data=[], layout=layout)
all_params = {p_name for t in trials for p_name in t.params.keys()}
if params is not None:
for input_p_name in params:
if input_p_name not in all_params:
raise ValueError("Parameter {} does not exist in your study.".format(input_p_name))
all_params = set(params)
sorted_params = sorted(list(all_params))
if logLoss==True:
for t in trials:
t.value=-1*np.log(t.value)
dims = [
{
"label": objective_value,
"values": tuple([t.value for t in trials]),
"range": (round(min([t.value for t in trials]),3), round(max([t.value for t in trials]),3)),
}
] # type: List[Dict[str, Any]]
for p_name in sorted_params:
values = []
for t in trials:
if p_name in t.params:
values.append(t.params[p_name])
is_categorical = False
try:
tuple(map(float, values))
except (TypeError, ValueError):
vocab = defaultdict(lambda: len(vocab)) # type: DefaultDict[str, int]
values = [vocab[v] for v in values]
is_categorical = True
dim = {
"label": name_shortner(p_name),
"values": tuple(values),
"range": ( round(min(values),2), round(max(values),2)),
}
if is_categorical:
dim["tickvals"] = list(range(len(vocab)))
ticktext=list(sorted(vocab.items(), key=lambda x: x[1]))
dim["ticktext"] = [x[0] for x in ticktext]
if p_name in (["Time Steps", "Latent Size", "Frequency Bins"]):
dim["tickvals"]= list(set(values))
dims.append(dim)
traces = [
go.Parcoords(
dimensions=dims,
tickfont=plotly.graph_objs.parcoords.Tickfont(color="black",size=18,family="Times New Roman"),
rangefont=plotly.graph_objs.parcoords.Rangefont(color="white",size=1,family="Times New Roman"),
line={
"color": dims[0]["values"],
"colorscale": coloring,
"colorbar": {"title": objective_value},
"showscale": True,
# "reversescale":True,
},
)
]
figure = go.Figure(data=traces, layout=layout)
return figure
# plotting Parallel Coordinates for the data frame
fig2 = go.Figure(data=go.Parcoords(
line=dict(color=num,
colorscale='HSV',
showscale=False,
cmin=0,
cmax=len(countries_in_europe)),
dimensions=list([
dict(range=[0, len(countries_in_europe)],
tickvals=c,
ticktext=countries_in_europe,
label="countries",
values=num),
dict(range=[
0, max(recent_deaths_data_frame['hospital_beds_per_thousand'])
],
label="Hospitals beds per 1000",
values=recent_deaths_data_frame['hospital_beds_per_thousand']),
dict(range=[0, max(recent_deaths_data_frame['median_age'])],
label='Median Age',
values=recent_deaths_data_frame['median_age']),
dict(range=[0, max(recent_deaths_data_frame['population'])],
label='Population',
values=recent_deaths_data_frame['population']),
dict(range=[0, max(recent_deaths_data_frame['life_expectancy'])],
label='Life expectancy',
values=covid19_data_frame['life_expectancy']),
dict(range=[0, max(recent_deaths_data_frame['covid19_death_rate'])],
label='COVID-19 Death rate',
values=recent_deaths_data_frame['covid19_death_rate']),
])),
layout=go.Layout(autosize=True,
))
for col in df:
if col == "Geoid" or col == "CountyName" or col == "CensusTract_2010_NAME":
continue
parallelBarForCol = dict(range=[df[col].min(), df[col].max()],
label=col,
values=df[col])
dimList.append(parallelBarForCol)
# 3) Plot:
colorCol = "TotalCrime"
figParallel = go.Figure(data=go.Parcoords(
# LINE COLOR DEF:
line=dict(
color=df[colorCol], #each lines' color val
colorscale="Electric",
showscale=True, #show the scale
cmin=df[colorCol].min(),
cmax=df[colorCol].max()),
dimensions=dimList))
figParallel.update_layout(
autosize=False,
width=5000,
height=1000,
)
figParallel.update_yaxes(
tickangle=45,
# tickfont=dict(family='Rockwell', color='crimson', size=14)
)
