devforfu

"""
Combination of Sutton's approach dividing space into boxes with simple 
TD-learning algorithm (see basic_rl.py somewhere on gym portal).

Some simulation paramteres are hardcoded and learner is not universal.
"""

from collections import defaultdict
import math

devforfu

"""
Cart pole environment learner that uses something like Cross Entropy Method.

"Something like" b/c not sure if implemented it correctly. And it is EXTREMELY slow. Anyway, 
looks like some kind of randomized search. 
"""

from operator import itemgetter
import heapq

devforfu

"""
A source code converted from:

  https://github.com/tensorflow/tensorflow/blob/master/tensorflow/examples/udacity/5_word2vec.ipynb
  
Into single script. But somehow, this implementation doesnt' show any improvement after 100000 iterations.
"""

import os
import math

devforfu

class TemperatureConverter(metaclass=abc.ABCMeta):    
    """
    Base class of temperature converters which supports dynamic substitution
    of implementations.
    """
    symbol = 'K'

    def __new__(cls, convert_to='celsius'):
        if issubclass(cls, TemperatureConverter):
            if convert_to == 'celsius':

devforfu

class _CelsiusConverter(TemperatureConverter):
    """
    Concrete implementation of temperature converted which converts from Kelvin
    into Celsius degrees.
    """

    symbol = '°C'

    def _convert(self, value):
        return value - 273.15

devforfu

"""
A bit more involved demonstration of __new__ method usage.

For more information see: https://iliazaitsev.me/blog/2018/02/14/python-new
"""
import io
import abc
import json
from os.path import exists
from textwrap import dedent

devforfu

"""
Simple example shows usage of generators as coroutines accepting values from 
outside and yield processed values to caller.
"""
import re


def get_loss(regex='^[\d\w]+(\d.\d+).hdf5$'):
    """
    Checks if name of file with model weights matches regular expression and

devforfu

class FeaturesExtractor:
    """Runs pretrained model without top layers on dataset and saves generated
    bottleneck features onto disk.
    """
    def __init__(self, build_fn, preprocess_fn, source,
                 target_size=(299, 299, 3), batch_size=128):        
        
        self.build_fn = build_fn
        self.preprocess_fn = preprocess_fn
        self.source = source

devforfu

def sgd(x_train, y_train, x_valid, y_valid, variance_threshold=0.1):    
    threshold = VarianceThreshold(variance_threshold)
        
    sgd_classifier = SGDClassifier(
        alpha=1./len(x_train),
        class_weight='balanced',
        loss='log', penalty='elasticnet', 
        fit_intercept=False, tol=0.001, n_jobs=-1)
    
    bagging = BaggingClassifier(

devforfu

class DNNClassifier:
    def __init__(self, hidden_layers, n_classes, activation=tf.nn.elu):
        self.hidden_layers = hidden_layers
        self.n_classes = n_classes
        self.activation = activation
        self._graph = None
        self._session = None
        self._saver = None
        self.stop_training = False