from keras.datasets import mnist from keras.models import Sequential from keras.layers import Dense, Dropout, Flatten from keras.layers import Conv2D, MaxPooling2D from keras import backend as K import matplotlib.pyplot as plt from general import generateData, train_model import numpy as np num_classes = 10 epochs = 10 #generate data (x_train, y_train, x_test, y_test, input_shape) = generateData(num_classes) #list to collect the accuracies for different learning rates accuracies = [] #what learning rates to iterate through learningrates = [0.001, 0.005, 0.025, 0.05, 0.1, 0.25, 0.5, 1] for learningRate in learningrates: accuracy = 0 #3 times for each learning rate for j in range(3): #Train model model, fit_info = train_model(x_train, y_train, x_test,
from keras import backend as K import matplotlib.pyplot as plt from general import generateData, train_model import numpy as np import math as m from seaborn.matrix import heatmap import pandas as pd from matplotlib import cm num_classes = 10 epochs = 10 # Generate the datasets (x_train, y_train, x_test, y_test, _) = generateData(num_classes) neuronsToTry = [10, 25, 50, 100, 150, 250, 500, 750, 1000] learningRatesToTry = [0.001, 0.005, 0.01, 0.05, 0.1] #Colors for plotting colors = ["blue","red","green","orange","purple","cyan","pink","brown"] #List to collect the performances of the different models modelPerformances = [] for neurons in neuronsToTry: for lr in learningRatesToTry: # Train model model, fit_info = train_model(
import keras from keras.datasets import mnist from keras.models import Sequential from keras.layers import Dense, Dropout, Flatten, GaussianNoise from keras.layers import Conv2D, MaxPooling2D from keras import backend as K from keras.regularizers import l2 import numpy as np from general import generateData, train_model num_classes = 10 #generate data (x_train, y_train, x_test, y_test, input_shape) = generateData(num_classes, 1, 3) #Train convolutional model convModel, _ = train_model( x_train, y_train, x_test, y_test, [ GaussianNoise(0.1), Conv2D(45, (5, 5), activation="relu", input_shape=input_shape, kernel_regularizer=l2(0.001)), MaxPooling2D(2, 2), Conv2D(60, (5, 5), activation="relu", input_shape=input_shape),