book

The Regularization Cookbook

by Vincent Vandenbussche

July 2023

Intermediate to advanced

424 pages

9h 20m

English

Packt Publishing

Read now

Unlock full access

ForewordContributorsAbout the authorAbout the reviewer
Who this book is forWhat this book coversTo get the most out of this bookDownload the example code filesConventions usedSectionsGetting readyHow to do it…How it works…There’s more…See alsoGet in touchShare Your ThoughtsDownload a free PDF copy of this book
Technical requirementsIntroducing regularizationExamples of models that did not pass the deployment testIntuition about regularizationKey concepts of regularizationBias and varianceUnderfitting and overfittingRegularization – from overfitting to underfittingUnavoidable biasDiagnosing bias and varianceRegularization – a multi-dimensional problemSummary
Technical requirementsLoading dataGetting readyHow to do it…There’s more…See alsoSplitting dataGetting readyHow to do it…See alsoPreparing quantitative dataGetting readyHow to do it…There’s more…See alsoPreparing qualitative dataGetting readyHow to do it…There’s more…See alsoTraining a modelGetting readyHow to do it…See alsoEvaluating a modelGetting readyHow to do it…See alsoPerforming hyperparameter optimizationGetting readyHow to do it…
Technical requirementsTraining a linear regression model with scikit-learnGetting readyHow to do it…There’s more…See alsoRegularizing with ridge regressionGetting readyHow to do it…There’s more…See alsoRegularizing with lasso regressionGetting readyHow to do it…There’s more…See alsoRegularizing with elastic net regressionGetting readyHow to do it…See alsoTraining a logistic regression modelGetting readyHow to do it…Regularizing a logistic regression modelGetting readyHow to do it…There’s more…Choosing the right regularizationGetting readyHow to do it…See also
Technical requirementsBuilding a classification treeDisorder measurementLoss functionGetting readyHow to do it…There’s more…See alsoBuilding regression treesGetting readyHow to do it…See alsoRegularizing a decision treeGetting readyHow to do it…How it works…There’s more…See alsoTraining the Random Forest algorithmGetting readyHow to do it…See alsoRegularization of Random ForestGetting startedHow to do it…Training a boosting model with XGBoostGetting readyHow to do it…See alsoRegularization with XGBoostGetting readyHow to do it…There’s more…
Technical requirementsHashing high cardinality featuresGetting startedHow to do it...See alsoAggregating featuresGetting readyHow to do it...There’s more...Undersampling an imbalanced datasetGetting readyHow to do it...There’s more...See alsoOversampling an imbalanced datasetGetting readyHow to do it...There’s more...See alsoResampling imbalanced data with SMOTEGetting readyHow to do it...There’s more...See also
Technical requirementsTraining a perceptronGetting startedHow to do it…There’s more…See alsoTraining a neural network for regressionGetting startedHow to do it…There’s more…See alsoTraining a neural network for binary classificationGetting readyHow to do it…There’s more…See alsoTraining a multiclass classification neural networkGetting readyHow to do it…There’s more…See also
Technical requirementsRegularizing a neural network with L2 regularizationGetting readyHow to do it...There’s more...See alsoRegularizing a neural network with early stoppingGetting readyHow to do it...There’s more...Regularization with network architectureGetting readyHow to do it...There’s more...Regularizing with dropoutGetting readyHow to do it...There’s more...See also
Technical requirementsTraining an RNNGetting startedHow to do it…There’s more…See alsoTraining a GRUGetting startedHow to do it…There’s more…See alsoRegularizing with dropoutGetting readyHow to do it…There’s more…Regularizing with the maximum sequence lengthGetting readyHow to do it…There’s more…

Technical requirementsRegularization using a word2vec embeddingGetting readyHow to do it…There’s more…See alsoData augmentation using word2vecGetting readyHow to do it…There’s more…See alsoZero-shot inference with pre-trained modelsGetting readyHow to do it…There’s more…See alsoRegularization with BERT embeddingsGetting readyHow to do it…There’s more…See alsoData augmentation using GPT-3Getting readyHow to do it…There’s more…See also
Technical requirementsTraining a CNNGetting startedHow to do it…There’s more…See alsoRegularizing a CNN with vanilla NN methodsGetting startedHow to do it…There’s more…See alsoRegularizing a CNN with transfer learning for object detectionObject detectionMean average precisionCOCO datasetGetting startedHow to do it…There’s more…See alsoSemantic segmentation using transfer learningGetting startedHow to do it…There’s more…See also
Technical requirementsApplying image augmentation with AlbumentationsSpatial-level augmentationPixel-level augmentationAlbumentationsGetting startedHow to do it…There’s more…See alsoCreating synthetic images for object detectionGetting startedHow to do it…There’s more…See alsoImplementing real-time style transferStable DiffusionPerceptual lossGetting startedHow to do it…There’s more…See also
Other Books You May EnjoyPackt is searching for authors like youShare Your ThoughtsDownload a free PDF copy of this book

Content preview from The Regularization Cookbook

10 Regularization in Computer Vision

In this chapter, we will explore another popular field of deep learning – computer vision. Computer vision is a large field with many tasks, from classification through generative models to object detection. Even though we can’t cover all of them, we will supply methods that can apply to all tasks.

In this chapter, we’ll cover the following recipes:

Training a convolutional neural network (CNN)
Regularizing a CNN with vanilla neural network (NN) methods
Regularizing a CNN with transfer learning for object detection
Semantic segmentation using transfer learning

At the end of this chapter, you will be able to handle several computer vision tasks such as image classification, object detection, instance segmentation, ...