Neural Network Projects with Python: The ultimate guide to using Python to explore the true power of neural networks through six projects

Cover
Title Page
Copyright and Credits
Dedication
About Packt
Contributors
Table of Contents
Preface
Chapter 1: Machine Learning and Neural Networks 101
	What is machine learning?
		Machine learning algorithms
		The machine learning workflow
	Setting up your computer for machine learning
	Neural networks
		Why neural networks?
		The basic architecture of neural networks
		Training a neural network from scratch in Python
			Feedforward
			The loss function
			Backpropagation
			Putting it all together
		Deep learning and neural networks
	pandas – a powerful data analysis toolkit in Python
		pandas DataFrames
		Data visualization in pandas
		Data preprocessing in pandas
			Encoding categorical variables
			Imputing missing values
		Using pandas in neural network projects
	TensorFlow and Keras – open source deep learning libraries
		The fundamental building blocks in Keras
			Layers – the atom of neural networks in Keras
			Models – a collection of layers
			Loss function – error metric for neural network training
			Optimizers – training algorithm for neural networks
		Creating neural networks in Keras
	Other Python libraries
	Summary
Chapter 2: Predicting Diabetes with Multilayer Perceptrons
	Technical requirements
	Diabetes – understanding the problem
	AI in healthcare
		Automated diagnosis
	The diabetes mellitus dataset
	Exploratory data analysis
	Data preprocessing
		Handling missing values
		Data standardization
		Splitting the data into training, testing, and validation sets
	MLPs
		Model architecture
			Input layer
			Hidden layers
			Activation functions
				ReLU
				Sigmoid activation function
	Model building in Python using Keras
		Model building
		Model compilation
		Model training
	Results analysis
		Testing accuracy
		Confusion matrix
		ROC curve
		Further improvements
	Summary
	Questions
Chapter 3: Predicting Taxi Fares with Deep Feedforward Networks
	Technical requirements
	Predicting taxi fares in New York City
	The NYC taxi fares dataset
	Exploratory data analysis
		Visualizing geolocation data
		Ridership by day and hour
	Data preprocessing
		Handling missing values and data anomalies
	Feature engineering
		Temporal features
		Geolocation features
	Feature scaling
	Deep feedforward networks
		Model architecture
		Loss functions for regression problems
	Model building in Python using Keras
	Results analysis
	Putting it all together
	Summary
	Questions
Chapter 4: Cats Versus Dogs - Image Classification Using CNNs
	Technical requirements
	Computer vision and object recognition
	Types of object recognition tasks
	Digital images as neural network input
	Building blocks of CNNs
		Filtering and convolution
		Max pooling
	Basic architecture of CNNs
	A review of modern CNNs
		LeNet (1998)
		AlexNet (2012)
		VGG16 (2014)
		Inception (2014)
		ResNet (2015)
		Where we stand today
	The cats and dogs dataset
	Managing image data for Keras
	Image augmentation
	Model building
		Building a simple CNN
		Leveraging on pre-trained models using transfer learning
	Results analysis
	Summary
	Questions
Chapter 5: Removing Noise from Images Using Autoencoders
	Technical requirements
	What are autoencoders?
	Latent representation
	Autoencoders for data compression
	The MNIST handwritten digits dataset
	Building a simple autoencoder
		Building autoencoders in Keras
		Effect of hidden layer size on autoencoder performance
	Denoising autoencoders
		Deep convolutional denoising autoencoder
	Denoising documents with autoencoders
		Basic convolutional autoencoder
		Deep convolutional autoencoder
	Summary
	Questions
Chapter 6: Sentiment Analysis of Movie Reviews Using LSTM
	Technical requirements
	Sequential problems in machine learning
	NLP and sentiment analysis
		Why sentiment analysis is difficult
	RNN
		What\'s inside an RNN?
		Long- and short-term dependencies in RNNs
		The vanishing gradient problem
	The LSTM network
		LSTMs – the intuition
		What\'s inside an LSTM network?
			Forget gate
			Input gate
			Output gate
			Making sense of this
	The IMDb movie reviews dataset
	Representing words as vectors
		One-hot encoding
		Word embeddings
	Model architecture
		Input
		Word embedding layer
		LSTM layer
		Dense layer
		Output
	Model building in Keras
		Importing data
		Zero padding
		Word embedding and LSTM layers
		Compiling and training models
	Analyzing the results
		Confusion matrix
	Putting it all together
	Summary
	Questions
Chapter 7: Implementing a Facial Recognition System with Neural Networks
	Technical requirements
	Facial recognition systems
	Breaking down the face recognition problem
		Face detection
			Face detection in Python
		Face recognition
	Requirements of face recognition systems
		Speed
		Scalability
		High accuracy with small data
	One-shot learning
		Naive one-shot prediction – Euclidean distance between two vectors
	Siamese neural networks
	Contrastive loss
	The faces dataset
	Creating a Siamese neural network in Keras
	Model training in Keras
	Analyzing the results
	Consolidating our code
	Creating a real-time face recognition program
		The onboarding process
		Face recognition process
		Future work
	Summary
	Questions
Chapter 8: What\'s Next?
	Putting it all together
		Machine Learning and Neural Networks 101
		Predicting Diabetes with Multilayer Perceptrons
		Predicting Taxi Fares with Deep Feedforward Nets
		Cats Versus Dogs – Image Classification Using CNNs
		Removing Noise from Images Using Autoencoders
		Sentiment Analysis of Movie Reviews Using LSTM
		Implementing a Facial Recognition System with Neural Networks
	Cutting edge advancements in neural networks
		Generative adversarial networks
		Deep reinforcement learning
	Limitations of neural networks
	The future of artificial intelligence and machine learning
		Artificial general intelligence
		Automated machine learning
	Keeping up with machine learning
		Books
		Scientific journals
		Practicing on real-world datasets
	Favorite machine learning tools
	Summary
Other Books You May Enjoy
Index