mrdbourke/tensorflow-deep-learning
Tensorflow Deep Learning
This is a comprehensive deep learning course delivered entirely through Jupyter Notebooks, designed to teach neural network construction using TensorFlow 2.x. The curriculum follows a sequential-model-first pedagogy, introducing the Sequential API before moving to functional and subclassing approaches, and covers the full spectrum of model building from regression and classification through convolutional neural networks, natural language processing, and time series forecasting.
The course is structured around a checkpoint-based training workflow that saves the best model weights during training, enabling resumption and evaluation without retraining. It includes a transfer learning pipeline for reusing pre-trained model weights as feature extractors or fine-tuning later layers, and a mixed-precision training pipeline that accelerates training and reduces memory usage by combining 16-bit and 32-bit floating-point arithmetic. Models are exported in .h5 format for direct submission to the TensorFlow Developer Certificate exam platform.
The material covers building and training neural networks for image classification, text processing with embeddings and recurrent or transformer layers, and time series forecasting from sequential historical data. It also provides guidance on diagnosing common performance issues related to input shapes, datatypes, activation functions, and loss functions. The course includes study materials and practice exercises aligned with the official TensorFlow Developer Certificate exam objectives, along with instructions for configuring a local development environment to ensure hardware can train exam models within required time limits.
Features
- Deep Learning Courses - Delivers a comprehensive deep learning course teaching neural network construction using TensorFlow 2.x.
- Jupyter Notebook Curricula - Delivers a structured deep learning curriculum as Jupyter notebooks with embedded code and exercises.
- Convolutional Neural Network Training - Trains models to recognize and classify objects in images using convolutional and pooling layers.
- Image Classification - Builds convolutional neural networks to recognize and classify objects in images using TensorFlow 2.x.
- Neural Network Classification - Trains models to assign categorical labels to data points using activation functions and loss functions.
- Transfer Learning Pipelines - Reuses pre-trained model weights as feature extractors or fine-tunes later layers for new tasks.
- Natural Language Processing - Trains models to classify and understand text data using embeddings and recurrent or transformer layers.
- Neural Network Construction - Walks through constructing, training, and evaluating neural networks for regression, classification, and image recognition.
- Regression Neural Networks - Trains models to predict continuous numerical values from input features using dense layers and optimizers.
- Sequential-First Pedagogies - Teaches neural network construction by starting with the Sequential API before introducing other approaches.
- Neural Network Training - Provides walkthroughs for constructing, training, and evaluating neural network models.
- Pre-training Transfer Learning - Adapts pre-trained neural networks to new tasks through feature extraction and fine-tuning.
- Time Series Forecasting - Trains models to predict future values from sequential data using past observations and trends.
- Best Weight Checkpointers - Saves the best model weights during training for resumption and evaluation without retraining.
- Transfer Learning - Reuses features from a pre-trained model and adapts them to a new task with less training data.
- TensorFlow Developer Certificate Preparation - Provides study materials and practice exercises aligned with the official TensorFlow Developer Certificate exam.
- Certification Preparation Materials - Provides study materials and practice exercises aligned with the official TensorFlow Developer Certificate exam.
- Classification Training - Builds and trains neural networks to classify images using TensorFlow 2.x.
- Mixed Precision Training - Trains models faster and uses less memory by combining 16-bit and 32-bit floating-point arithmetic.
- Transfer Learning Guides - Provides guides on reusing pre-trained model features and fine-tuning them for new tasks.
- Model Checkpointing - Saves the best model weights during training and reloads them for evaluation or further training.
- Text Tokenization - Tokenizes text data to classify content or predict sequences using a neural network.
- Layer Unfreezing Fine-Tuners - Unfreezes and retrains later layers of a pre-trained model to adapt it to a new, similar task.
- Feature Extraction Pipelines - Uses a pre-trained model as a fixed feature extractor to train a new classifier on a smaller dataset.
- TensorFlow Time Series - Provides instructions for training models to predict future values from sequential data using TensorFlow.
- Model Export Submissions - Exports trained models in .h5 format for direct submission to the TensorFlow Developer Certificate exam.
- Educational Resources - Comprehensive tutorial series for learning deep learning with TensorFlow.
