def __init__(self,
save_path: Path = None,
skip_images=True,
with_text_overlay=False,
advanced_eval=False,
max_epochs=-1,
data_loader=None):
super().__init__()
self.tp, self.fp, self.tn, self.fn = 0, 0, 0, 0
self.accuracy, self.balanced_accuracy, self.precision, self.recall, self.specificity = 0, 0, 0, 0, 0
self.confusion_matrix = np.zeros((2, 2), dtype=int)
self.evaluated_samples_epoch = 0
self.epoch = 0
self.max_epochs = max_epochs
self.class_names = ["OK", "Not OK"]
self.save_path = save_path
self.skip_images = skip_images
self.data_loader = data_loader
if not self.skip_images:
# when running TensorBoard, use '--samples_per_plugin images=100' to see all frames in slider
self.writer = SummaryWriter(log_dir=Path(get_artifact_uri()))
if save_path is not None:
self.im_save_path = save_path / "images"
self.im_save_path.mkdir(parents=True, exist_ok=True)
self.num = 0
self.with_text_overlay = with_text_overlay
plt.set_loglevel('warning')
if advanced_eval:
self.origin_tracker = {}
def distribution_of_features(df,
n_cols=3,
fig_size=(16, 26),
color_plot='gold',
kde_show=True,
label_rotation=45,
set_loglevel='warning'):
"""
Plots distribution of features in a pandas.DataFrame.
Does not work on feature encoded as category.
Parameters:
===========
df = pandas.DataFrame,
n_cols = int; default: 3,
controls number of columns per row of the figure.
fig_size = tuple (length, height); default: (16, 26),
controls the figure size of the output.
color_plot = str; default: 'gold',
controls color of the histplot and kde plot.
kde_show = `int` or `boolean`; default: `True`,
`True` shows kde plot.
`False` does not show kde plot.
label_rotation = int; default: 45,
sets x label rotation.
set_loglevel = str; default: 'warning',
The log level of matplotlib warning handler.
- options = {"notset", "debug", "info", "warning", "error", "critical"}
---version 1.2---
"""
plt.set_loglevel(set_loglevel)
fig, axes = plt.subplots(nrows=num_col_for_plotting(len(df.columns),
col=n_cols),
ncols=n_cols,
figsize=fig_size,
sharey=False)
for ax, column in zip(axes.flatten(), df):
sns.histplot(x=column, data=df, color=color_plot, ax=ax, kde=kde_show)
ax.set_title(f'Histplot of {column.title()}')
ax.tick_params('x', labelrotation=label_rotation)
sns.despine()
plt.tight_layout()
plt.suptitle('Histogram plots of the dataset',
fontsize=20,
fontweight=3,
va='bottom')
plt.show()
def show_image_plt(im: np.ndarray):
"""Show image in matplotlib window."""
if platform.system() != "Windows":
LOG.info("Cannot show plot on this OS")
return
plt.set_loglevel("info")
plt.figure()
cmap = None
if len(im.shape) == 2:
cmap = "gray"
else:
im = cv2.cvtColor(im, cv2.COLOR_BGR2RGB)
plt.imshow(im.astype(np.uint8), cmap=cmap, vmin=0, vmax=255)
plt.show()
def __init__(self, save_path: Path = None,
skip_images=True,
with_text_overlay=False,
summary_writer=None,
advanced_eval=False):
super().__init__()
self.tp, self.fp, self.tn, self.fn = 0, 0, 0, 0
self.accuracy, self.precision, self.recall, self.specificity = 0, 0, 0, 0
self.confusion_matrix = np.zeros((2, 2), dtype=int)
self.save_path = save_path
self.skip_images = skip_images
if save_path is not None:
self.im_save_path = save_path / "images"
if not self.skip_images:
self.im_save_path.mkdir(parents=True, exist_ok=True)
self.num = 0
self.with_text_overlay = with_text_overlay
self.summary_writer = summary_writer
plt.set_loglevel('warning')
if advanced_eval:
self.origin_tracker = {}
import copy
import warnings
from functools import reduce
from math import inf
try:
import graphviz
except ModuleNotFoundError as e:
graphviz = None
import joblib
import matplotlib.pyplot as plt
import numpy as np
plt.set_loglevel('warning')
def message_sort_key(message):
if type(message) == str:
return inf
number = reduce(lambda a, b: (a << 1) + int(b), message)
return number
def plot_stats(statistics, ylog=False, view=False, filename='avg_fitness.svg'):
""" Plots the population's average and best fitness. """
if plt is None:
warnings.warn(
"This display is not available due to a missing optional dependency (matplotlib)"
x = np.arange(10000)
y = np.random.randn(10000)
# disable matplotlib key shortcut
plt.rcParams['keymap.xscale'] = ''
plt.rcParams['keymap.yscale'] = ''
fig = plt.figure()
ax1 = plt.subplot(311)
ax2 = plt.subplot(312)
ax3 = plt.subplot(325)
ax4 = plt.subplot(326)
plt.set_loglevel("info")
#fig.subplots_adjust(hspace=0.5)
class Station_Cursor(object):
def __init__(self, ax):
global LOW_FREQ_POINT
global HIGH_FREQ_POINT
self.ax = ax
self.v1 = ax.axvline(LOW_FREQ_POINT, color='red', ls='--') #
self.v2 = ax.axvline(HIGH_FREQ_POINT, color='red', ls='--') #
# text location in axes coords
#self.txt = ax.text(0.7, 0.9, '', transform=ax.transAxes)
self.keyPressed = False
#print('init')
import numpy as np
import matplotlib as mpl
from matplotlib import pyplot as plt
from matplotlib.ticker import MaxNLocator
from seaborn import violinplot, histplot
from myUtilities import myParams
"""
myAna.py 所具有的功能
(1) 財務與風險指標評估
(2) 訓練結果視覺化
備註：財務與風險指標評估第(5)大項中，未平倉的交易不會被計入
"""
# Global variables and constants:
plt.set_loglevel("warning") # avoid log polluted by mpl
# --- --- #
def Training_Visualizer(metrics_df, isembedded=False):
"""
Training analytics:
Arg.: metrics_df with columns
- loss
- val_loss
- reward
- epsilon
isembedded: embedded to GUI (e.g. tk widget) or not
"""
plt.style.use('dark_background')
def set_plt_for_latex():
plt.rc("text", usetex=True)
plt.rc("font", **{"family": "serif", "sans-serif": ["Palatino"]})
plt.set_loglevel("info")
