TensorFlow is an open-source machine learning framework developed by Google's Brain Team, designed to build, train, and deploy machine learning models. It represents data as multidimensional arrays called tensors, which flow through a computational graph of mathematical operations. This architecture allows the same model to run on CPUs, GPUs, and TPUs without rewriting code, making it flexible enough for both research and production environments.
The framework supports the full machine learning workflow, from data preprocessing and model training to deployment. It includes Keras, a high-level API built on top of TensorFlow that simplifies model building for common use cases, and TensorBoard, a visualization tool for monitoring training progress and debugging model performance. TensorFlow also handles automatic gradient computation, which is central to backpropagation and gradient descent during training.
TensorFlow is one of several widely used machine learning frameworks, alongside PyTorch and JAX.