book

Applied Deep Learning with Keras

by Ritesh Bhagwat, Mahla Abdolahnejad, Matthew Moocarme

April 2019

Intermediate to advanced

412 pages

7h 46m

English

Packt Publishing

Read now

Unlock full access

About the BookAbout the AuthorsLearning ObjectivesAudienceApproachHardware RequirementsSoftware RequirementsConventionsInstallation and SetupInstalling the Code BundleAdditional Resources
Introduction to Machine Learning with KerasIntroductionData RepresentationTables of DataLoading DataExercise 1: Loading a Dataset from the UCI Machine Learning RepositoryData PreprocessingExercise 2: Cleaning the DataAppropriate Representation of the DataExercise 3: Appropriate Representation of the DataLife Cycle of Model CreationMachine Learning Librariesscikit-learnKerasAdvantages of KerasDisadvantages of KerasMore than Building ModelsModel TrainingClassifiers and Regression ModelsClassification TasksRegression TasksTraining and Test DatasetsModel Evaluation MetricsExercise 4: Creating a Simple ModelModel TuningBaseline ModelsExercise 5: Determining a Baseline ModelRegularizationCross-ValidationActivity 1: Adding Regularization to the ModelSummary
Machine Learning versus Deep LearningIntroductionAdvantages of ANNs over Traditional Machine Learning AlgorithmsAdvantages of Traditional Machine Learning Algorithms over ANNsHierarchical Data RepresentationLinear TransformationsScalars, Vectors, Matrices, and TensorsTensor AdditionExercise 6: Perform Various Operations with Vectors, Matrices, and TensorsReshapingMatrix TranspositionExercise 7: Matrix Reshaping and TranspositionMatrix MultiplicationExercise 8: Matrix MultiplicationExercise 9: Tensor MultiplicationIntroduction to KerasLayer TypesActivation FunctionsModel FittingActivity 2: Creating a Logistic Regression Model Using KerasSummary
Deep Learning with KerasIntroductionBuilding Your First Neural NetworkLogistic Regression to a Deep Neural NetworkActivation FunctionsForward Propagation for Making Predictions Loss FunctionBackpropagation for Computing Derivatives of Loss FunctionGradient Descent for Learning ParametersExercise 10: Neural Network Implementation with KerasActivity 3: Building a Single-Layer Neural Network for Performing Binary ClassificationModel EvaluationEvaluating a Trained Model with KerasSplitting Data into Training and Test SetsUnderfitting and OverfittingEarly StoppingActivity 4: Diabetes Diagnosis with Neural NetworksSummary
Evaluate Your Model with Cross-Validation using Keras WrappersIntroductionCross-ValidationDrawbacks of Splitting a Dataset Only OnceK-Fold Cross-ValidationLeave-One-Out Cross-ValidationComparing the K-Fold and LOO MethodsCross-Validation for Deep Learning ModelsKeras Wrapper with scikit-learnExercise 11: Building the Keras Wrapper with scikit-learn for a Regression ProblemCross-Validation with scikit-learnCross-Validation Iterators in scikit-learnExercise 12: Evaluate Deep Neural Networks with Cross-ValidationActivity 5: Model Evaluation Using Cross-Validation for a Diabetes Diagnosis ClassifierModel Selection with Cross-validationCross-Validation for Model Evaluation versus Model SelectionExercise 13: Write User-Defined Functions to Implement Deep Learning Models with Cross-ValidationActivity 6: Model Selection Using Cross-Validation for the Diabetes Diagnosis ClassifierActivity 7: Model Selection Using Cross-validation on the Boston House Prices DatasetSummary
Improving Model AccuracyIntroductionRegularizationThe Need for RegularizationReducing Overfitting with RegularizationL1 and L2 RegularizationL1 and L2 Regularization FormulationL1 and L2 Regularization Implementation in KerasActivity 8: Weight Regularization on a Diabetes Diagnosis ClassifierDropout RegularizationPrinciples of Dropout RegularizationReducing Overfitting with DropoutExercise 14: Dropout Implementation in KerasActivity 9: Dropout Regularization on Boston Housing DatasetOther Regularization MethodsEarly StoppingExercise 15: Implementing Early Stopping in KerasData AugmentationAdding NoiseHyperparameter Tuning with scikit-learnGrid Search with scikit-learnRandomized Search with scikit-learnActivity 10: Hyperparameter Tuning on the Diabetes Diagnosis ClassifierSummary
Model EvaluationIntroductionAccuracyExercise 16: Calculating Null Accuracy on a Dummy Healthcare DatasetAdvantages and Limitations of AccuracyImbalanced DatasetsWorking with Imbalanced DatasetsConfusion MatrixMetrics Computed from a Confusion MatrixExercise 17: Computing Accuracy and Null Accuracy with Healthcare DataActivity 11: Computing the Accuracy and Null Accuracy of a Neural Network When We Change the Train/Test SplitActivity 12: Derive and Compute Metrics Based on a Confusion MatrixExercise 18: Calculate the ROC and AUC CurvesSummary
Computer Vision with Convolutional Neural NetworksIntroductionComputer VisionConvolutional Neural NetworksArchitecture of a CNNInput ImageConvolution LayerPooling LayerFlatteningImage AugmentationAdvantages of Image AugmentationExercise 19: Build a CNN and Identify Images of Cats and DogsActivity 13: Amending our Model with Multiple Layers and the Use of SoftMaxExercise 20: Amending our model by reverting to the Sigmoid activation functionExercise 21: Changing the Optimizer from Adam to SGDExercise 22: Classifying a New ImageActivity 14: Classify a New ImageSummary
Transfer Learning and Pre-Trained ModelsIntroductionPre-Trained Sets and Transfer LearningFeature ExtractionFine-Tuning a Pre-Trained Network The ImageNet DatasetSome Pre-Trained Networks in KerasExercise 23: Identify an Image Using the VGG16 NetworkActivity 15: Use the VGG16 Network to Train a Deep Learning Network to Identify ImagesExercise 24: Classification of Images That Are Not Present in the ImageNet Database.Exercise 25: Fine-Tune the VGG16 ModelExercise 26: Image Classification with ResNetActivity 16: Image Classification with ResNet Summary
Sequential Modeling with Recurrent Neural NetworksIntroductionSequential Memory and Sequential Modeling Recurrent Neural Networks (RNNs)The Vanishing Gradient ProblemLong Short-Term Memory (LSTM)Exercise 27: Predict the Trend of Apple's Stock Price Using an LSTM with 50 Units (Neurons)Activity 17: Predict the Trend of Microsoft's Stock Price Using an LSTM with 50 Units (Neurons)Exercise 28: Predicting the Trend of Apple's Stock Price Using an LSTM with 100 unitsActivity 18: Predicting Microsoft's Stock Price with Added RegularizationActivity 19: Predicting the Trend of Microsoft's Stock Price Using an LSTM with an Increasing Number of LSTM Neurons (100 Units) Summary

Chapter 1: Introduction to Machine Learning with KerasActivity 1: Adding Regularization to the ModelChapter 2: Machine Learning versus Deep LearningActivity 2: Creating a Logistic Regression Model Using KerasChapter 3: Deep Learning with KerasActivity 3: Building a Single-Layer Neural Network for Performing Binary ClassificationActivity 4: Diabetes Diagnosis with Neural NetworksChapter 4: Evaluate Your Model with Cross-Validation with Keras WrappersActivity 5: Model Evaluation Using Cross-Validation for a Diabetes Diagnosis ClassifierActivity 6: Model Selection Using Cross-Validation for the Diabetes Diagnosis ClassifierActivity 7: Model Selection Using Cross-validation on the Boston House Prices DatasetChapter 5: Improving Model AccuracyActivity 8: Weight Regularization on a Diabetes Diagnosis ClassifierActivity 9: Dropout Regularization on Boston House Prices DatasetActivity 10: Hyperparameter Tuning on the Diabetes Diagnosis ClassifierChapter 6: Model EvaluationActivity 11: Computing Accuracy and Null Accuracy of Neural Network When We Change the Train/Test SplitActivity 12: Derive and Compute Metrics Based on the Confusion MatrixChapter 7: Computer Vision with Convolutional Neural NetworksActivity 13: Amending our Model with Multiple Layers and Use of SoftMaxActivity 14: Classify a New ImageChapter 8: Transfer Learning and Pre-trained ModelsActivity 15: Use the VGG16 Network to Train a Deep Learning Network to Identify ImagesActivity 16: Image Classification with ResNet Chapter 9: Sequential Modeling with Recurrent Neural NetworksActivity 17: Predict the Trend of Microsoft’s Stock Price Using an LSTM with 50 Units (Neurons)Activity 18: Predicting Microsoft’s stock price with added regularizationActivity 19: Predicting the Trend of Microsoft’s Stock Price Using LSTM with 100 Units

Overview

Delve into the world of deep learning with 'Applied Deep Learning with Keras.' This book offers a comprehensive guide, starting with the basics of machine learning and Python, and progressing to advanced applications using Keras to build efficient deep neural networks. Perfect for those looking to solve real-world problems with practical examples and cutting-edge techniques.

What this Book will help me do

Understand the differences between machine learning and deep learning architectures, and the scenarios in which each is applied.
Become proficient in using Keras to build and train various types of neural networks including CNNs, RNNs, and DNNs.
Learn techniques like L1, L2, and dropout regularization to enhance the performance of your neural network models.
Master model evaluation methods such as cross-validation and performance metrics to ensure the reliability of your models.
Apply your skills to practical problems such as customer churn prediction, disease diagnosis, and more through step-by-step projects.

Author(s)

Ritesh Bhagwat, Mahla Abdolahnejad, and Matthew Moocarme bring a rich technical background as practitioners and educators in the field of AI and deep learning. They combine their vast experience in neural networks and data science to create engaging and hands-on learning content. Their detailed and approachable methods aim to help readers build confidence and expertise in deep learning applications.

Who is it for?

This book is designed for individuals familiar with data science and machine learning who wish to deepen their knowledge of artificial neural networks and their practical applications. If you've worked with Python and have insights into statistics and logistic regression, this book will enhance your abilities. Familiarity with libraries like scikit-learn will serve as a bonus, but not a necessity. Ultimately, it caters to data science enthusiasts eager to explore Keras for building cutting-edge machine learning solutions.