Fragmentation of Distributed Exchanges

Abstract

Distributed securities exchanges may become de facto fragmented if they span geographical regions with asymmetric computer infrastructure. First, we build an economic model of a decentralized exchange with two miner clusters, standing in for compact areas of economic activity (e.g., cities). "Local" miners in the area with relatively higher trading activity only join a decentralized exchange if they enjoy a large speed advantage over "long-distance" competitors. This is due to a transfer of economic value across miners, specifically from high-to lowactivity clusters. Second, we estimate the speed advantage of "local" over "long-distance" miners in a series of Monte Carlo experiments over a two-cluster, unstructured peer-to-peer network simulated in C. We find that the speed advantage increases in the level of infrastructure asymmetry between clusters. Crossregion DEX blockchains are feasible as long as the asymmetry levels in trading activity and infrastructure availability across regions are positively correlated.

Graph-based Neural Networks Framework using Microcontrollers for Energy-Efficient Traffic Forecasting

“…A vehicle on a road (an edge) should forecast the traffic to choose a route that is less busy to increase the transportation energy efficiency. The neural network was training using MATLAB) [13] and it was tested using a C language implementation on microcontrollers on each vehicle [8,14]. The number of LSTM hidden layers is about 80 (our evaluation shows that the increasing of number of layers will not improve the neural network performance).…”