def _pack_column_data_source(kwargs, source):
"""
TBD
"""
if 'size' in kwargs and isinstance(kwargs['size'], list):
source.add(kwargs['size'], 'size')
del kwargs['size']
if 'radius' in kwargs and isinstance(kwargs['radius'], list):
source.add(kwargs['radius'], 'radius')
del kwargs['radius']
if 'c' in kwargs and isinstance(kwargs['c'], list):
source.add(kwargs['c'], 'color')
del kwargs['c']
if 'line_color' in kwargs and isinstance(kwargs['line_color'], list):
source.add(kwargs['line_color'], 'line_color')
del kwargs['line_color']
if 'color' in kwargs and isinstance(kwargs['color'], list):
p = palette(max(kwargs['color']))
source.add([p[i-1] for i in kwargs['color']], 'color')
del kwargs['color']
if 'marker' in kwargs and isinstance(kwargs['marker'], list):
source.add(kwargs['marker'], 'marker')
del kwargs['marker']
# TODO i hate this
for k,v in list(kwargs.items()):
if isinstance(v, list):
source.add(v, k)
del kwargs[k]
def _pack_column_data_source(kwargs, source):
"""
TBD
"""
if 'size' in kwargs and isinstance(kwargs['size'], list):
source.add(kwargs['size'], 'size')
del kwargs['size']
if 'color' in kwargs and isinstance(kwargs['color'], list):
p = palette(max(kwargs['color']))
source.add([p[i - 1] for i in kwargs['color']], 'color')
del kwargs['color']
if 'marker' in kwargs and isinstance(kwargs['marker'], list):
source.add(kwargs['marker'], 'marker')
del kwargs['marker']
def visualize(self, bucketing=True):
if bucketing:
er_n = len(self.all_episodes_returns)
er_buckets_number = min(int(np.ceil(0.1 * er_n)), 100)
er_bucket_size = int(er_n / er_buckets_number)
performance = (np.array(
bucketify(self.all_episodes_returns, er_buckets_number,
np.mean)) + 1) / 2.0
else:
performance = self.all_episodes_returns
mean_reward = np.mean(self.all_episodes_returns)
ep_act_times_flatten = [
item for sublist in self.episodes_actions_times for item in sublist
]
mean_action_time = np.mean(ep_act_times_flatten)
if bucketing:
at_n = len(ep_act_times_flatten)
at_buckets_number = min(int(np.ceil(0.1 * er_n)), 100)
at_bucket_size = int(at_n / at_buckets_number)
action_times = bucketify(ep_act_times_flatten, at_buckets_number,
np.mean)
else:
action_times = ep_act_times_flatten
# fig = plt.figure(constrained_layout=True)
# fig.suptitle(f"{self.__class__.__name__}", fontsize=16, y=1.1)
# spec = gridspec.GridSpec(ncols=1, nrows=2, figure=fig)
#
# ax0 = fig.add_subplot(spec[0, 0])
# if bucketing:
# ax0.set_xlabel(f"Episodes buckets ({er_buckets_number} buckets of size {er_bucket_size})")
# ax0.set_ylabel("Percentage of winnings")
# else:
# ax0.set_ylabel("Mean return")
# ax0.set_title(f"Performance")
# ax0.plot(performance)
#
# ax1 = fig.add_subplot(spec[1, 0])
# if bucketing:
# ax1.set_xlabel(f"Actions buckets ({at_buckets_number} buckets of size {at_bucket_size})")
# ax1.set_ylabel("Mean action time [s]")
# ax1.set_title(f"Action times")
# ax1.plot(action_times)
#
# plt.show()
# output_notebook(hide_banner=True)
performance_f = figure(
title='Performance',
x_axis_label=
f"Episodes buckets ({er_buckets_number} buckets of size {er_bucket_size})",
y_axis_label='Percentage of winnings',
plot_width=900,
plot_height=300,
toolbar_location=None,
y_range=(-0.1, 1.1))
performance_f.yaxis.bounds = (0, 1)
performance_f.yaxis.formatter = NumeralTickFormatter(format='0 %')
length = len(performance)
color_gradient_clipping = 8
colors = list(palette(int(1.3 * length))[:length])
color_mapper = LinearColorMapper(palette=colors,
low=min(performance),
high=max(performance))
performance_f.scatter(range(len(performance)),
performance,
color={
'field': 'y',
'transform': color_mapper
})
performance_f.line(range(len(performance)),
performance,
color='grey',
alpha=0.5)
action_times_f = figure(
title='Action times',
x_axis_label=
f"Actions buckets ({at_buckets_number} buckets of size {at_bucket_size})",
y_axis_label='Mean action time [ms]',
plot_width=900,
plot_height=300,
toolbar_location=None)
action_times_f.line(range(len(action_times)),
np.array(action_times) * 10**3,
color='blue')
show(
column(performance_f, action_times_f,
self.epsilon_strategy.get_figure()))
# print(f"Number of played episodes: {er_n}")
print(f"Mean reward: {'{0:0.6f}'.format(mean_reward)}")
# print(f"Number of taken actions: {at_n}")
print(f"Mean action time: {'{0:0.6f}'.format(mean_action_time)}")
def operating_hours(data_sets=[]):
########## BUILD FIGURES ################
if len(data_sets) < 1:
data_sets = _get_dummies()
series_count = len(data_sets)
colours = palette(series_count)
data_frames = []
for ds in data_sets:
df = ds.as_data_frame()
day_filter = (df['timestamp'].dt.dayofweek
== 5) | (df['timestamp'].dt.dayofweek == 6)
#df = df.drop(df[day_filter].index)
hour_filter = (df['timestamp'].dt.hour < 15) | (df['timestamp'].dt.hour
> 21)
#df = df.drop(df[hour_filter].index)
#df = df.drop(df[day_filter | hour_filter].index)
#df['temperature'][day_filter | hour_filter] = np.NaN
#df['humidity'][day_filter | hour_filter] = np.NaN
idx = df.ix[day_filter | hour_filter].index
#df.temperature[idx] = np.NaN
#df.humidity[idx] = np.NaN
df.loc[idx, 'temperature'] = np.NaN
df.loc[idx, 'humidity'] = np.NaN
#df.at[dates[5], 'E'] = 7
df['time'] = df['timestamp'].dt.time
data_frames.append(df)
all_data_temp = figure(responsive=True,
x_axis_label="Time of day",
x_axis_type="datetime",
y_axis_label="Temperature / C",
y_axis_type="linear",
**FIGURE_OPTIONS)
for (clr, ds, df) in zip(colours, data_sets, data_frames):
#my_plot = all_data_temp.scatter(df.time, df.temperature, color = clr, legend = ds.name, **SCATTER_OPTIONS)
my_plot = all_data_temp.line(df.time,
df.temperature,
color=clr,
legend=ds.name,
**LINE_OPTIONS)
all_data_humi = figure(x_range=all_data_temp.x_range,
responsive=True,
x_axis_label="Time of day",
x_axis_type="datetime",
y_axis_label="Relative humidity / \%",
y_axis_type="linear",
**FIGURE_OPTIONS)
for (clr, ds, df) in zip(colours, data_sets, data_frames):
my_plot = all_data_humi.scatter(df.time,
df.humidity,
color=clr,
legend=ds.name,
**SCATTER_OPTIONS)
#my_plot = all_data_humi.line(df.time, df.humidity, color = clr, legend = ds.name, **LINE_OPTIONS)
for p in [all_data_temp, all_data_humi]:
p.xaxis.formatter = DatetimeTickFormatter(formats=dict(
hours=["%k:%M"],
days=["%d. %m. %y"],
months=["%m %Y"],
years=["%Y"],
))
all_data = gridplot([all_data_temp, all_data_humi],
ncols=2,
plot_width=500,
plot_height=250,
sizing_mode='scale_width',
toolbar_options=dict(logo="grey"))
#toolbar_options=dict(logo="grey", location='above'), merge_tools=False)
########## RENDER PLOTS ################
resources = INLINE
js_resources = resources.render_js()
css_resources = resources.render_css()
plot_script, plot_divs = components(
{'Data fra kl. 15 - 22, uden loerdag': all_data})
return js_resources, css_resources, plot_script, plot_divs
def alldata(data_sets=[]):
########## BUILD FIGURES ################
if len(data_sets) < 1:
data_sets = _get_dummies()
series_count = len(data_sets)
colours = palette(series_count)
all_data_temp = figure(responsive=True,
x_axis_label="Days",
x_axis_type="datetime",
y_axis_label="Temperature / C",
y_axis_type="linear",
**FIGURE_OPTIONS)
for (clr, ds) in zip(colours, data_sets):
my_plot = all_data_temp.line(ds.time_series,
ds.temp_series,
color=clr,
legend=ds.name,
**LINE_OPTIONS)
all_data_humi = figure(x_range=all_data_temp.x_range,
responsive=True,
x_axis_label="Days",
x_axis_type="datetime",
y_axis_label="Relative humidity / \%",
y_axis_type="linear",
**FIGURE_OPTIONS)
for (clr, ds) in zip(colours, data_sets):
my_plot = all_data_humi.line(ds.time_series,
ds.humi_series,
color=clr,
legend=ds.name,
**LINE_OPTIONS)
for p in [all_data_temp, all_data_humi]:
p.xaxis.formatter = DatetimeTickFormatter(formats=dict(
hours=["%k:%M"],
days=["%d. %m. %y"],
months=["%m %Y"],
years=["%Y"],
))
all_data = gridplot([all_data_temp, all_data_humi],
ncols=2,
plot_width=500,
plot_height=250,
sizing_mode='scale_width',
toolbar_options=dict(logo="grey"))
#toolbar_options=dict(logo="grey", location='above'), merge_tools=False)
########## RENDER PLOTS ################
resources = INLINE
js_resources = resources.render_js()
css_resources = resources.render_css()
plot_script, plot_divs = components({'Oversigt over alt data': all_data})
return js_resources, css_resources, plot_script, plot_divs
def statistics(data_sets=[]):
########## BUILD FIGURES ################
if len(data_sets) < 1:
data_sets = _get_dummies()
series_count = len(data_sets)
colours = palette(series_count)
data_frame = pd.DataFrame()
data_frames = []
for ds in data_sets:
df = ds.as_data_frame()
day_filter = (df['timestamp'].dt.dayofweek
== 5) | (df['timestamp'].dt.dayofweek == 6)
#df = df.drop(df[day_filter].index)
hour_filter = (df['timestamp'].dt.hour < 15) | (df['timestamp'].dt.hour
> 21)
#df = df.drop(df[hour_filter].index)
#df = df.drop(df[day_filter | hour_filter].index)
#df['temperature'][day_filter | hour_filter] = np.NaN
#df['humidity'][day_filter | hour_filter] = np.NaN
idx = df.ix[day_filter | hour_filter].index
#df.temperature[idx] = np.NaN
#df.humidity[idx] = np.NaN
df.loc[idx, 'temperature'] = np.NaN
df.loc[idx, 'humidity'] = np.NaN
#df.at[dates[5], 'E'] = 7
df = df.drop(idx)
df['time'] = df['timestamp'].dt.time
df['box_label'] = [
"{1}-{2} - {0}".format(ds.name, tt, tt + 1)
for tt in df['timestamp'].dt.hour
]
df['box_label_merged'] = [
"kl. {1}-{2} - {0}".format(ds.name.split(',')[0], tt, tt + 1)
for tt in df['timestamp'].dt.hour
]
df.loc[:, 'colour'] = ds.name
df.loc[:, 'colour_merged'] = ds.name.split(',')[0]
data_frames.append(df)
data_frame = pd.concat([data_frame, df], ignore_index=True)
#data_frame = pd.DataFrame(columns=['timestamp', 'temperature', 'humidity', 'box_label'])
all_data_temp = BoxPlot(data_frame,
values='temperature',
label='box_label',
color='colour',
responsive=True,
xlabel="Time and place",
ylabel="Temperature / C",
legend=False)
all_data_humi = BoxPlot(data_frame,
values='humidity',
label='box_label',
color='colour',
responsive=True,
xlabel="Time and place",
ylabel="Relative humidity / \%",
legend=False)
all_data = gridplot([all_data_temp, all_data_humi],
ncols=2,
plot_width=500,
plot_height=400,
sizing_mode='scale_width',
toolbar_options=dict(logo="grey"))
#toolbar_options=dict(logo="grey", location='above'), merge_tools=False)
merged_data_temp = BoxPlot(data_frame,
values='temperature',
label='box_label_merged',
color='colour_merged',
responsive=True,
xlabel="Time and place",
ylabel="Temperature / C",
legend=False)
merged_data_humi = BoxPlot(data_frame,
values='humidity',
label='box_label_merged',
color='colour_merged',
responsive=True,
xlabel="Time and place",
ylabel="Relative humidity / \%",
legend=False)
merged_data = gridplot([merged_data_temp, merged_data_humi],
ncols=2,
plot_width=500,
plot_height=400,
sizing_mode='scale_width',
toolbar_options=dict(logo="grey"))
#toolbar_options=dict(logo="grey", location='above'), merge_tools=False)
########## RENDER PLOTS ################
resources = INLINE
js_resources = resources.render_js()
css_resources = resources.render_css()
plot_script, plot_divs = components({
'Data fra kl. 15 - 22, uden loerdag':
all_data,
'Data fra kl. 15 - 22, uden loerdag, reference og uden udsugning':
merged_data
})
return js_resources, css_resources, plot_script, plot_divs