Google can identify and transcribe all the views it has of street numbers in France in less than an hour, thanks to a neural network that’s just as good as human operators. Now its engineers reveal how they developed it.
Lots of sinergies here: ML/CS tools can be leveraged to obtain approximate solutions to hard game theoretical problems (like mechanism design), while at the same time game theoretical concepts inspire better, more realistic ML algorithms.
Stanford researchers, working with Google and NVIDIA, have created a new neural network system for machine learning that is six times the size of the unit built last year that taught itself how to recognize cats on the internet.
Facebook has detailed its extensive improvements to the open source Apache Giraph graph-processing platform. The project, which is built on top of Hadoop, can now process trillions of connections between people, places and things in minutes.
Many ML algorithms are most readily expressed in terms of local, vertex-centric computations. Think of label propagation, k-means, spectral clustering ... All in all, moving from vector spaces to graphs is the natural thing to do in many applications. Giraph provides an efficient way to run this kind of algorithms on top of an existing Hadoop infrastructure.