My current work at the Fraunhofer-Chalmers Research Centre for Industrial Mathematics focuses on distributed data analytics in a large-scale industrial setting. We are closely collaborating with Volvo Cars and Volvo Trucks. As part of my work, I explored the suitability of Erlang for distributed machine learning tasks. I wrote up a draft paper on my implementation of Federated Learning in Erlang, which got accepted to IFL 2017, the 29th symposium on Implementation and Application of Functional Languages. It will take place in Bristol, UK, from 30 August to 1 September, 2017. The abstract is reproduced below.


Purely Functional Federated Learning in Erlang

Authors: Ulm, Gregor; Emil Gustavsson, Mats Jirstrand

Arguably the biggest strength of the functional programming language Erlang is how straightforward it is to implement concurrent and distributed programs with it. Numerical computing, on the other hand, is not necessarily seen as one of its strengths. The recent introduction of Federated Learning, a concept according to which edge devices are leveraged for decentralized machine learning tasks, while a central server only updates and distributes a global model, provided the motivation for exploring how well Erlang was suited to such a use case. We present a framework for Federated Learning in Erlang, written in a purely functional style, and compare two versions of it: one that has been exclusively written in Erlang, and one in which Erlang is relegated to coordinating client processes that rely on performing numerical computations in the programming language C. Initial results are promising, as we learnt that a real-world industrial use case of distributed data analytics can easily be tackled with a system purely written in Erlang.
The novelty of our work is that we present the first implementation of a Federated Learning framework in a functional programming language, with the added benefit of being purely functional. In addition, we demonstrate that Erlang can not only be leveraged for message passing but also performs adequately for practical machine learning tasks.