Talk: Compositional Blocks for Optimal Self-Healing Gradients

About this presentation

This presentation, which I held at SASO’17 Conference in Tucson (Arizona, USA), describes a novel nearly-MP-optimal gradient algorithm as well as an approach for defining a gradient algorithm with best performance trade-off to date by mixing different algorithms and techniques together.

A gradient is the adaptive field resulting from calculating, in any node of a networked system, the minimum distance to any source node. Such spatio-temporal structure is very useful and is at the basis of many self-organisation patterns.

Building performant gradient implementations is important to improve the dynamical characteristics of higher-level patterns.