TensorFlow is the result of Google's tinkering with deep learning neural networks back in 2011, with the first generation machine learning proprietary system, dubbed DistBelief, and the second generation iteration is a core part of its commercial products like Google Voice Search, Google Photos, and YouTube.
Now, Google has extended the open source machine learning library to JavaScript (JS) with Tensorflow.js, a JS library for machine learning models deployment in the browser.
Tensorflow.js will work with Node.js server-side JavaScript runtime and as a WebGL-accelerated library remains part of the TensorFlow ecosystem; with the machine learning models running directly in the browser.
It eliminates the need for drivers, as developers can run codes right on the browser. And TensorFlow.js model converters can also run existing models under Node.js or in the browser, which models are retrained with sensor data connected to the browser.
TensorFlow.js APIs will enable developers to build models using low-level linear algebra library or the higher-level layers API, with the Keras-inspired API included for building high-level neural networks.
Albeit, TensorFlow.js isn't the only available JavaScript library for building neural networks, Synaptic offers free JavaScript neural network library for node.js and the browser, with architecture-free generalized algorithm.
Tensorflow.js API does not yet support all the functionality of the Python API, though the developers of the JavaScript API has pledged to work more in achieving an architecture-free JavaScript API.
No comments