book

AI and Machine Learning for Coders

by Laurence Moroney

October 2020

Intermediate to advanced

392 pages

9h 36m

English

O'Reilly Media, Inc.

Audiobook available

Read now

Unlock full access

Who Should Read This BookWhy I Wrote This BookNavigating This BookTechnology You Need to UnderstandOnline ResourcesConventions Used in This BookUsing Code ExamplesO’Reilly Online LearningHow to Contact UsAcknowledgments
What Is Machine Learning?Limitations of Traditional ProgrammingFrom Programming to LearningWhat Is TensorFlow?Using TensorFlowInstalling TensorFlow in PythonUsing TensorFlow in PyCharmUsing TensorFlow in Google ColabGetting Started with Machine LearningSeeing What the Network LearnedSummary
Recognizing Clothing ItemsThe Data: Fashion MNISTNeurons for VisionDesigning the Neural NetworkThe Complete CodeTraining the Neural NetworkExploring the Model OutputTraining for Longer—Discovering OverfittingStopping TrainingSummary
ConvolutionsPoolingImplementing Convolutional Neural NetworksExploring the Convolutional NetworkBuilding a CNN to Distinguish Between Horses and HumansThe Horses or Humans DatasetThe Keras ImageDataGeneratorCNN Architecture for Horses or HumansAdding Validation to the Horses or Humans DatasetTesting Horse or Human ImagesImage AugmentationTransfer LearningMulticlass ClassificationDropout RegularizationSummary
Getting Started with TFDSUsing TFDS with Keras ModelsLoading Specific VersionsUsing Mapping Functions for AugmentationUsing TensorFlow AddonsUsing Custom SplitsUnderstanding TFRecordThe ETL Process for Managing Data in TensorFlowOptimizing the Load PhaseParallelizing ETL to Improve Training PerformanceSummary
Encoding Language into NumbersGetting Started with TokenizationTurning Sentences into SequencesRemoving Stopwords and Cleaning TextWorking with Real Data SourcesGetting Text from TensorFlow DatasetsGetting Text from CSV FilesGetting Text from JSON FilesSummary
Establishing Meaning from WordsA Simple Example: Positives and NegativesGoing a Little Deeper: VectorsEmbeddings in TensorFlowBuilding a Sarcasm Detector Using EmbeddingsReducing Overfitting in Language ModelsUsing the Model to Classify a SentenceVisualizing the EmbeddingsUsing Pretrained Embeddings from TensorFlow HubSummary
The Basis of RecurrenceExtending Recurrence for LanguageCreating a Text Classifier with RNNsStacking LSTMsUsing Pretrained Embeddings with RNNsSummary

Turning Sequences into Input SequencesCreating the ModelGenerating TextPredicting the Next WordCompounding Predictions to Generate TextExtending the DatasetChanging the Model ArchitectureImproving the DataCharacter-Based EncodingSummary
Common Attributes of Time SeriesTrendSeasonalityAutocorrelationNoiseTechniques for Predicting Time SeriesNaive Prediction to Create a BaselineMeasuring Prediction AccuracyLess Naive: Using Moving Average for PredictionImproving the Moving Average AnalysisSummary
Creating a Windowed DatasetCreating a Windowed Version of the Time Series DatasetCreating and Training a DNN to Fit the Sequence DataEvaluating the Results of the DNNExploring the Overall PredictionTuning the Learning RateExploring Hyperparameter Tuning with Keras TunerSummary
Convolutions for Sequence DataCoding ConvolutionsExperimenting with the Conv1D HyperparametersUsing NASA Weather DataReading GISS Data in PythonUsing RNNs for Sequence ModelingExploring a Larger DatasetUsing Other Recurrent MethodsUsing DropoutUsing Bidirectional RNNsSummary
What Is TensorFlow Lite?Walkthrough: Creating and Converting a Model to TensorFlow LiteStep 1. Save the ModelStep 2. Convert and Save the ModelStep 3. Load the TFLite Model and Allocate TensorsStep 4. Perform the PredictionWalkthrough: Transfer Learning an Image Classifier and Converting to TensorFlow LiteStep 1. Build and Save the ModelStep 2. Convert the Model to TensorFlow LiteStep 3. Optimize the ModelSummary
What Is Android Studio?Creating Your First TensorFlow Lite Android AppStep 1. Create a New Android ProjectStep 2. Edit Your Layout FileStep 3. Add the TensorFlow Lite DependenciesStep 4. Add Your TensorFlow Lite ModelStep 5. Write the Activity Code to Use TensorFlow Lite for InferenceMoving Beyond “Hello World”—Processing ImagesTensorFlow Lite Sample AppsSummary
Creating Your First TensorFlow Lite App with XcodeStep 1. Create a Basic iOS AppStep 2. Add TensorFlow Lite to Your ProjectStep 3. Create the User InterfaceStep 4. Add and Initialize the Model Inference ClassStep 5. Perform the InferenceStep 6. Add the Model to Your AppStep 7. Add the UI LogicMoving Beyond “Hello World”—Processing ImagesTensorFlow Lite Sample AppsSummary
What Is TensorFlow.js?Installing and Using the Brackets IDEBuilding Your First TensorFlow.js ModelCreating an Iris ClassifierSummary
JavaScript Considerations for TensorFlow DevelopersBuilding a CNN in JavaScriptUsing Callbacks for VisualizationTraining with the MNIST DatasetRunning Inference on Images in TensorFlow.jsSummary
Converting Python-Based Models to JavaScriptUsing the Converted ModelsUsing Preconverted JavaScript ModelsUsing the Toxicity Text ClassifierUsing MobileNet for Image Classification in the BrowserUsing PoseNetSummary
Transfer Learning from MobileNetStep 1. Download MobileNet and Identify the Layers to UseStep 2. Create Your Own Model Architecture with the Outputs from MobileNet as Its InputStep 3. Gather and Format the DataStep 4. Train the ModelStep 5. Run Inference with the ModelTransfer Learning from TensorFlow HubUsing Models from TensorFlow.orgSummary
What Is TensorFlow Serving?Installing TensorFlow ServingInstalling Using DockerInstalling Directly on LinuxBuilding and Serving a ModelExploring Server ConfigurationSummary
Fairness in ProgrammingFairness in Machine LearningTools for FairnessThe What-If ToolFacetsFederated LearningStep 1. Identify Available Devices for TrainingStep 2. Identify Suitable Available Devices for TrainingStep 3. Deploy a Trainable Model to Your Training SetStep 4. Return the Results of the Training to the ServerStep 5. Deploy the New Master Model to the ClientsSecure Aggregation with Federated LearningFederated Learning with TensorFlow FederatedGoogle’s AI PrinciplesSummary

Content preview from AI and Machine Learning for Coders

Chapter 3. Going Beyond the Basics: Detecting Features in Images

In Chapter 2 you learned how to get started with computer vision by creating a simple neural network that matched the input pixels of the Fashion MNIST dataset to 10 labels, each representing a type (or class) of clothing. And while you created a network that was pretty good at detecting clothing types, there was a clear drawback. Your neural network was trained on small monochrome images that each contained only a single item of clothing, and that item was centered within the image.

To take the model to the next level, you need to be able to detect features in images. So, for example, instead of looking merely at the raw pixels in the image, what if we could have a way to filter the images down to constituent elements? Matching those elements, instead of raw pixels, would help us to detect the contents of images more effectively. Consider the Fashion MNIST dataset that we used in the last chapter—when detecting a shoe, the neural network may have been activated by lots of dark pixels clustered at the bottom of the image, which it would see as the sole of the shoe. But when the shoe is no longer centered and filling the frame, this logic doesn’t hold.

One method to detect features comes from photography and the image processing methodologies that you might be familiar with. If you’ve ever used a tool like Photoshop or GIMP to sharpen an image, you’re using a mathematical filter that works on the pixels of the image. ...