• Search by Module
• Search by Words
• Search Projects

• Java
• Python
• JavaScript
• TypeScript
• C++
• Scala
• Blog

• MKLpy-master
  • examples
    • test_preprocessing.py
    • test_training.py
    • test_multiclass.py
    • test_string_kernls.py
    • test_GRAM.py
    • test_modelselection.py
    • test_metrics.py
  • mkdocs.yml
  • LICENSE
  • .readthedocs.yml
  • MKLpy
    • utils
      • misc.py
      • exceptions.py
      • validation.py
      • __init__.py
    • arrange.py
    • generators.py
    • algorithms
      • AverageMKL.py
      • AlternateMKL.py
      • EasyMKL.py
      • MEMO.py
      • komd.py
      • __init__.py
      • GRAM.py
      • base.py
    • multiclass.py
    • scheduler.py
    • model_selection.py
    • __init__.py
    • callbacks.py
    • preprocessing
      • data_preprocessing.py
      • kernel_preprocessing.py
      • __init__.py
      • binarization.py
    • metrics
      • alignment.py
      • evaluate.py
      • __init__.py
      • pairwise
        • vector.py
        • boolean.py
        • string.py
        • __init__.py
  • setup.py
  • setup.cfg
  • .travis.yml
  • README.md
  • tests
    • unit_tests.py
  • requirements.txt
  • .gitignore
  • docs
    • resources
      • mklpy_logo.svg
    • faq.md
    • 404.md
    • examples
      • examples_generators.md
      • proj_kernet.md
      • examples_modelselection.md
      • examples_training.md
      • examples_kernels.md
      • examples_loading.md
      • examples_evaluation.md
    • index.md
    • about.md
    • developers
      • algorithms.md
      • generators.md
      • scheduler.md
      • callbacks.md
    • stylesheets
      • extra.css
    • requirements.txt

import numpy as np
from sklearn.metrics import roc_auc_score, accuracy_score
import warnings
from .algorithms.base import Solution
from .metrics import ratio


class Callback():

	def __init__(self):
		self.model = None

	def register(self, model):
		self.model = model

	def on_train_begin(self, KL, Y):
		pass

	def on_train_end(self):
		pass

	def on_step_begin(self, step):
		pass

	def on_step_end(self, step):
		pass


class EarlyStopping(Callback):

	def __init__(self,
				KLva,
				Yva,
				patience=5,
				cooldown=1,
				metric='val_obj',
				restore_best_solution=True,
			):

		super().__init__()
		self.KLva = KLva
		self.Yva  = Yva
		self.patience = patience
		self.cooldown = cooldown
		self.metric = metric
		self.restore_best_solution = restore_best_solution
		assert metric in ['obj', 'val_obj', 'auc', 'accuracy']

	def register(self, model):
		self.model = model
		if self.metric in ['auc', 'accuracy'] or model.direction=='max':
			self.monitor_op = np.greater
			self.best = -np.Inf
		else:
			self.monitor_op = np.less
			self.best = np.Inf


	def on_train_begin(self):
		self.wait = 0
		self.stopped_epoch = 0
		self.baseline = None
		self.best_solution = self.model.solution #initial sol
		#todo: set self.best
		#self.on_step_end(0)
		self.vals = []#ratio(self.model.solution.ker_matrix, self.model.Y)]
		#print ('[0][earlystopping] initial obj', self.model.solution.objective, self.vals[0])


	def on_step_end(self, step):
		if step % self.cooldown:
			return

		if self.metric == 'obj':
			current = self.model.solution.objective
		elif self.metric == 'val_obj':
			pass
		elif self.metric == 'auc':
			Kva = self.model.func_form(self.KLva, self.model.solution.weights)
			ys = self.model.learner.fit(self.model.solution.ker_matrix, self.model.Y) \
				.decision_function(Kva)
			current = roc_auc_score(self.Yva, ys)
		elif self.metric == 'accuracy':
			Kva = self.model.func_form(self.KLva, self.model.solution.weights)
			ys = self.model.learner.fit(self.model.solution.ker_matrix, self.model.Y) \
				.predict(Kva)
			current = accuracy_score(self.Yva, ys)
		else:
			raise Error('Metric error')

		self.vals.append(current)#ratio(self.model.solution.ker_matrix, self.model.Y))
		#print ('[%d][earlystopping] current' % step, \
		#	self.model.solution.objective, self.vals[-1], current)


		if self.monitor_op(current, self.best):
			self.best_solution = Solution(
				weights = self.model.solution.weights,
				objective = self.model.solution.objective,
				ker_matrix = self.model.solution.ker_matrix,
				)
			self.best = current
			self.wait = 0
		else:
			self.wait += 1
			if self.wait >= self.patience:
				self.stoppend_epoch = step
				self.model.convergence = True


	def on_train_end(self):
		if self.restore_best_solution:
			self.model.solution = self.best_solution