The Kolmogorov complexity measures the length of the shortest program required to reproduce a pattern. We approximate the Kolmogorov complexity using a lossless compression technique. In this way, we found that model 1 and 2 have both an approximate byte count of 416 (despite the delta in evolutionary model complexity), and model 15 has a byte count 768. That is, in (metric) information theoretical terms the first two simplest models present invariance with respect to affine transformations in the trust space. This is a disambiguation aid that the AI uses for decision making: a description of the world with a lower error (model 2) can be encoded at the same level of computational complexity as an inferior alternative.
Configuring blockchain protocols' parameters based on the networks' topology analysis
Authors have explored graph-based techniques to automatically detect realistic decentralized network growth models from empirical data \cite{Menezes_2014} and to study systemic risk in cryptocurrency markets \cite{Venegasa}. The causal inference network in Figure 4 uses a parametric approach based on symbolic regression to derive the relationship between nodes, starting from a sample of 2000 price and consumption of services in the cryptocurrency markets during the period of August 2016 to January 2018.