This week’s featured GitHub project is Keras, a modular neural network and deep-learning library based on the Theano Python library.
Keras allows for fast prototyping of deep-learning applications through a minimalist library structure of modular, extensible components. Developed by François Chollet, the neural network library is designed for quick experimentation, and supports both convolutional and recurrent deep-learning networks on the CPU and GPU.
The library is made up of models: sequences of standalone configurable modules that can be plugged into neural layers and functions. Each module consists of less than 100 lines of code, and the library is written in Python for simpler debugging and syntax highlighting.
Running on the Theano library, Keras holds a greater ability for developers to define, optimize and evaluate mathematical expressions involving multi-dimensional arrays. The popular Python library is commonly used in large-scale scientific computations and experiments, and includes dynamic C code generation and unit testing for application self-verification.
According to Chollet, Keras is named after a reference in ancient Greek and Latin literature, first found in the Homer’s ”Odyssey,” of dream spirits divided between those who deceive men with false visions, who arrive on Earth through a gate of ivory, and those who announce a future that will come to pass, who arrive through a gate of horn.
More details about Keras are available here.
Top 5 projects trending on GitHub this week
#1: LeakCanary, a memory leak detection library for Android and Java.
#2: JSON Server was featured in last week’s Top 5 trending projects.
#3: PathPicker is a Facebook creation: a simple command-line tool that solves the perpetual problem of selecting files out of bash output.
#4: Awesome, a curated list of all the Awesome lists pervading GitHub. There are oodles and oodles of ‘em.
#5: Zoom.js is a simple jQuery plug-in for image zooming, used in production by the content publishing and social sharing platform Medium.