A deep learning framework is a software system that provides reusable abstractions and computational capabilities for building and training neural networks.
A deep learning framework replaces low-level, hand-written gradient and hardware code with reusable software for building and training neural networks. section 3.1
Deep learning frameworks combine three capabilities:
Compute gradients automatically for arbitrary differentiable functions. section 3.1
Run tensor computations on CPUs, GPUs, and potentially specialized deep learning hardware. section 3.1
Distribute computation across multiple devices or machines. section 3.1
How Deep Learning with R, Third Edition covers it
How deep it goes
Defines three core capabilities and sketches their historical emergence, but does not compare framework internals.
Examples
Names Keras, JAX, TensorFlow, and PyTorch as major frameworks, contrasting them with hand-written C++ gradient logic.
What you can do after
Evaluate Keras, JAX, TensorFlow, and PyTorch as reusable alternatives to low-level neural-network implementations.
How it connects
Connects automatic differentiation, tensor computation, GPU computing, and distributed computing.
PyTorchincludes · PyTorch is an open-source machine-learning framework for Python that provides tensor computation, automatic differentiation, and tools for building and training neural networks.
Kerasincludes · Keras is an open-source, high-level deep-learning API for building, training, evaluating, and deploying neural-network models.
JAXincludes · JAX is a Python library for high-performance array-based numerical computing with composable transformations including automatic differentiation, compilation, vectorization, and parallelization.
TensorFlowincludes · TensorFlow is an open-source software platform for numerical computation and machine learning, with tools for developing, training, and deploying models.
automatic differentiationuses · Automatic differentiation is a computational technique that evaluates derivatives of a program by systematically applying the chain rule to its elementary operations.
Model Servingaddresses · A model-serving platform is software infrastructure that hosts trained models and exposes inference by accepting requests and returning model outputs.
This explainer was generated by Manning's knowledge graph from the text of Deep Learning with R, Third Edition. The section links point to the parts of the book it draws on.
