def process(self, inputs):
"""
Plot the P & L graph from the `strategy_returns` column.
`label` in the `conf` defines the stock symbol name
Arguments
-------
inputs: list
list of input dataframes.
Returns
-------
Figure
"""
input_df = inputs[0]
if isinstance(input_df, dask_cudf.DataFrame):
input_df = input_df.compute() # get the computed value
label = 'stock'
if 'label' in self.conf:
label = self.conf['label']
num_points = self.conf['points']
stride = max(len(input_df) // num_points, 1)
date_co = DateScale()
linear_co = LinearScale()
yax = Axis(label='Cumulative return', scale=linear_co,
orientation='vertical')
xax = Axis(label='Time', scale=date_co, orientation='horizontal')
panzoom_main = PanZoom(scales={'x': [date_co]})
line = Lines(x=input_df['datetime'][::stride],
y=(input_df['strategy_returns'].cumsum())[::stride],
scales={'x': date_co, 'y': linear_co},
colors=['blue'], labels=[label], display_legend=True)
new_fig = Figure(marks=[line], axes=[yax, xax], title='P & L',
interaction=panzoom_main)
return new_fig
def __init__(self, viewer, **kwargs):
super().__init__(viewer, **kwargs)
self.interact = PanZoom(scales={
'x': [self.viewer.scale_x],
'y': [self.viewer.scale_y]
})
def process(self, inputs):
"""
Plot the lines from the input dataframe. The plotted lines are the
columns in the input dataframe which are specified in the `lines` of
node's `conf`
The plot title is defined in the `title` of the node's `conf`
Arguments
-------
inputs: list
list of input dataframes.
Returns
-------
Figure
"""
input_df = inputs[self.INPUT_PORT_NAME]
num_points = self.conf['points']
stride = max(len(input_df) // num_points, 1)
date_co = DateScale()
linear_co = LinearScale()
yax = Axis(label='', scale=linear_co, orientation='vertical')
xax = Axis(label='Time', scale=date_co, orientation='horizontal')
panzoom_main = PanZoom(scales={'x': [date_co]})
lines = []
for line in self.conf['lines']:
col_name = line['column']
label_name = line['label']
color = line['color']
if (isinstance(input_df, cudf.DataFrame)
or isinstance(input_df, dask_cudf.DataFrame)):
line = Lines(x=input_df['datetime'][::stride].to_array(),
y=input_df[col_name][::stride].to_array(),
scales={
'x': date_co,
'y': linear_co
},
colors=[color],
labels=[label_name],
display_legend=True)
else:
line = Lines(x=input_df['datetime'][::stride],
y=input_df[col_name][::stride],
scales={
'x': date_co,
'y': linear_co
},
colors=[color],
labels=[label_name],
display_legend=True)
lines.append(line)
new_fig = Figure(marks=lines,
axes=[yax, xax],
title=self.conf['title'],
interaction=panzoom_main)
return {self.OUTPUT_PORT_NAME: new_fig}
def process(self, inputs):
"""
Plot the ROC curve
Arguments
-------
inputs: list
list of input dataframes.
Returns
-------
Figure
"""
input_df = inputs[self.INPUT_PORT_NAME]
if isinstance(input_df, dask_cudf.DataFrame):
input_df = input_df.compute() # get the computed value
label_col = input_df[self.conf['label']].values
pred_col = input_df[self.conf['prediction']].values
if isinstance(input_df, cudf.DataFrame):
fpr, tpr, _ = metrics.roc_curve(label_col.get(),
pred_col.get())
else:
fpr, tpr, _ = metrics.roc_curve(label_col,
pred_col)
auc_value = metrics.auc(fpr, tpr)
out = {}
if self.outport_connected(self.OUTPUT_PORT_NAME):
linear_x = LinearScale()
linear_y = LinearScale()
yax = Axis(label='True Positive Rate', scale=linear_x,
orientation='vertical')
xax = Axis(label='False Positive Rate', scale=linear_y,
orientation='horizontal')
panzoom_main = PanZoom(scales={'x': [linear_x]})
curve_label = 'ROC (area = {:.2f})'.format(auc_value)
line = Lines(x=fpr, y=tpr,
scales={'x': linear_x, 'y': linear_y},
colors=['blue'], labels=[curve_label],
display_legend=True)
new_fig = Figure(marks=[line], axes=[yax, xax], title='ROC Curve',
interaction=panzoom_main)
out.update({self.OUTPUT_PORT_NAME: new_fig})
if self.outport_connected(self.OUTPUT_VALUE_NAME):
out.update({self.OUTPUT_VALUE_NAME: float(auc_value)})
return out
def __init__(self,
volume=None,
default_directory=os.getcwd(),
title='',
enhancement_steps=1000,
**kwargs):
def on_chosen_path_change(old_path, new_path):
self.dataset = FolderDataset(new_path)
# TODO: If the path doesn't contain images, display a warning
# A widget for changing the image folder
self.pathchooser = PathChooser(
chosen_path_desc='Image folder:',
default_directory=default_directory,
on_chosen_path_change=on_chosen_path_change,
)
self.pathchooser.layout.margin = '0 0 10px 0'
# The number of increments of the min/max slider
self.enhancement_steps = enhancement_steps
self.scales = {
'x': LinearScale(),
'y': LinearScale(),
}
# The currently displayed image will be in bytes at `self.image_plot.image.value`
self.image_plot = BQImage(
image=IPyImage(),
scales=self.scales,
)
self.figure = Figure(
marks=[self.image_plot],
padding_x=0,
padding_y=0,
animation_duration=1000,
fig_margin={
'top': 0,
'right': 0,
'bottom': 0,
'left': 0,
},
layout=Layout(
grid_area='figure',
margin='0',
width='320px',
height='320px',
),
)
# Custom toolbar
toolbar_width = '100%'
toolbar_margin = '0px 0 2px 0'
self.pan_zoom = PanZoom(scales={
'x': [self.scales['x']],
'y': [self.scales['y']],
}, )
self.save_button = Button(
description='Save Image',
tooltip='Save Image',
icon='save',
layout=Layout(
width=toolbar_width,
# flex='1 1 auto',
margin=toolbar_margin,
),
)
self.save_button.on_click(self.save_current_image)
self.hide_button = Button(
description='Hide Image',
tooltip='Hide Image',
icon='eye-slash',
layout=Layout(
width=toolbar_width,
# flex='1 1 auto',
margin=toolbar_margin,
))
self.hide_button.on_click(self.hide_current_image)
self.pan_zoom_toggle_button = ToggleButton(
description='Pan / Zoom',
tooltip='Pan/Zoom',
icon='arrows',
layout=Layout(
width=toolbar_width,
# flex='1 1 auto',
margin=toolbar_margin,
),
)
self.pan_zoom_toggle_button.observe(self.on_pan_zoom_toggle,
names='value')
self.reset_pan_zoom_button = Button(
description='Undo Zoom',
tooltip='Reset pan/zoom',
icon='refresh',
layout=Layout(
width=toolbar_width,
# flex='1 1 auto',
margin=toolbar_margin,
),
)
self.reset_pan_zoom_button.on_click(self.reset_pan_zoom)
self.reset_enhancements_button = Button(
description='Un-Enhance',
tooltip='Reset enhancements',
icon='ban',
layout=Layout(
width=toolbar_width,
# flex='1 1 auto',
margin=toolbar_margin,
),
)
self.reset_enhancements_button.on_click(self.reset_enhancements)
self.mini_map = IPyImage(layout=Layout(
grid_area='mini-map',
margin='0',
))
self.mini_map.width = 180
self.mini_map.height = 180
# PERFORMANCE CONCERN
# Ideally instead of four observations, this would observe 'scales' on `self.pan_zoom`
# However, it doesn't fire updates
# Ref: https://github.com/bloomberg/bqplot/issues/800
self.image_plot.scales['x'].observe(self.on_pan_zoom_change('x_min'),
names='min')
self.image_plot.scales['x'].observe(self.on_pan_zoom_change('x_max'),
names='max')
self.image_plot.scales['y'].observe(self.on_pan_zoom_change('y_min'),
names='min')
self.image_plot.scales['y'].observe(self.on_pan_zoom_change('y_max'),
names='max')
self.plane_toggle = ToggleButtons(
options=['yz', 'xz', 'xy'],
description='',
disabled=False,
button_style='',
tooltips=[
'Step in x direction', 'Step in y direction',
'Step in z direction'
],
layout=Layout(
width='200px',
# flex='1 1 auto',
margin='7px 0 auto auto',
),
)
self.plane_toggle.style.button_width = 'auto'
self.plane_toggle.observe(self.on_plane_change, names='value')
self.toolbar = VBox(
children=[
self.save_button,
self.hide_button,
self.pan_zoom_toggle_button,
self.reset_pan_zoom_button,
self.reset_enhancements_button,
],
layout=Layout(
grid_area='toolbar',
margin='0',
),
)
# Image enhancements
self.min_max_slider = FloatRangeSlider(
value=[0, 255],
min=0,
max=255,
step=255 / self.enhancement_steps,
description='Min/Max:',
orientation='horizontal',
readout=True,
readout_format='.1f',
continuous_update=True,
layout=Layout(
grid_area='min-max-slider',
margin='10px 0 10px -10px',
width='100%',
),
)
self.min_max_slider.observe(self.on_min_max_change, names='value')
self.index_slider = IntSlider(
value=0,
min=0,
max=1,
step=1,
description='Index:',
orientation='horizontal',
readout=True,
readout_format='d',
continuous_update=True,
layout=Layout(
grid_area='index-slider',
margin='8px -20px 10px -36px',
width='100%',
),
)
self.index_slider.observe(self.on_image_index_change, names='value')
# Animation
self.play = Play(
value=self.index_slider.value,
min=self.index_slider.min,
max=self.index_slider.max,
step=self.index_slider.step,
)
jslink((self.play, 'value'), (self.index_slider, 'value'))
# Keep 'max' in sync as well
self.index_slider.observe(self.on_index_slider_max_change, names='max')
self.bottom_bar = HBox(
children=[
self.play,
self.index_slider,
self.plane_toggle,
],
layout=Layout(
grid_area='bottom-bar',
margin=f'10px -20px 0 0',
# overflow='hidden',
))
# Layout
self.gridbox = GridBox(children=[
self.figure,
self.toolbar,
self.mini_map,
self.min_max_slider,
self.bottom_bar,
], )
# Initially hidden without data
self.gridbox.layout.display = 'none'
self._dataset = None
if volume is not None:
self.dataset = VolumeDataset(volume)
# Hide pathchooser when using a volume
self.pathchooser.layout.display = 'none'
# Call VBox super class __init__
super().__init__(
children=[
self.pathchooser,
self.gridbox,
],
layout=Layout(width='auto'),
**kwargs,
)
