import random
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from keras.models import Sequential
from keras.layers import Dense
from keras.utils import np_utils
from keras.layers import LocallyConnected1D
from keras import optimizers
from keras import initializers
from sklearn.model_selection import train_test_split

def IndexOfMax(arr):
    max = arr[0]
    indexOfMax = 0
    for i in range(0, len(arr)):
        if(max < arr[i]):
            max = arr[i]
            indexOfMax = i
    return indexOfMax

#  предикторы и отклик разделяем
# y - таблица ответов
X = np.array([[random.randint(0, 100) for g in range(3)] for i in range(400)])
y = np.array([IndexOfMax(arr) for arr in X])
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=12312, test_size=0.33)

#Keras работает с numpy массивами, так что из dataframe конвертируем в массив


y_train_bin = np_utils.to_categorical(y_train)
y_test_bin = np_utils.to_categorical(y_test)

init1 = initializers.TruncatedNormal(seed = 12332)
init2 = initializers.Constant(value = 1e-1)

#half deepleaning
model = Sequential()
model.add(Dense(4, input_dim=3, activation='relu', kernel_initializer = init1, bias_initializer = init2))
model.add(Dense(12, activation='relu', kernel_initializer = init1, bias_initializer = init2))
model.add(Dense(6, activation='relu', kernel_initializer = init1, bias_initializer = init2))
model.add(Dense(3, activation='softmax', kernel_initializer = init1, bias_initializer = init2))

# Compiling model
#lr - лямбда, decay - изменение лямбда
sgd = optimizers.SGD(lr = 0.1, decay = 1e-7, momentum = 0.1, nesterov = True)
na = optimizers.Nadam()
model.compile(loss='binary_crossentropy', optimizer=na, metrics=['accuracy'])


# Training a model
history = model.fit(X_train, y_train_bin, epochs=100, batch_size=10)

# evaluate the model
scores = model.evaluate(X_test, y_test_bin)
print("\nAccuracy: %.2f%%" % (scores[1]*100))
plt.plot(history.history['acc'])
plt.show()
# calculate predictions
predictions = model.predict(X_test)