Google Open-Sources TensorFlow

Google open-sourced TensorFlow in November 2015 under the Apache 2.0 license. The project gave researchers and developers access to a production-oriented machine-learning framework that could run on CPUs and GPUs.

TensorFlow’s graph-based model described computations as operations connected by tensors. The system handled automatic differentiation and could place operations on available devices. This was especially useful for neural networks, where training repeatedly computes gradients across millions of parameters.

The release arrived during rapid growth in deep learning and became one of the field’s dominant frameworks. It also helped standardize the idea that machine-learning research should be expressed in reusable software systems rather than as one-off numerical code.