BAYOEN

BAYOEN develops an organic neuromorphic hardware implementation of a Bayesian network for autonomous AI ECG sensors that will be inexpensive, green, consume very low power and be compatible with flexible technologies. Bayesian networks are built on the premise that uncertain knowledge can be represented by a probability. BAYOEN implements a stochastic computing paradigm to perform Bayesian inference by implementing low transistor count circuits using the novel properties of the organic electrochemical transistors (OECTs). The computational part of the project develops the methodology to transform ECG sensor data into stochastic bitstreams that represent the probabilities of different cardiovascular anomalies using coupled arrays of simulated spiking neurons. The experimental component uses these bitstreams to demonstrate a Bayesian classification of the data using a circuit of organic electrochemical transistors (OECTs). The main components of the project include: 1) the circuit design and optimization of the material properties of the OECTs, 2) the fabrication of the OECT circuits, 3) simulations of the coupled neurons that output stochastic bitstreams, 4) the implementation of the Bayesian inference in the OECT circuit. The project is a collaboration between the Center for Nanoscience and Nanotechnology (C2N) lab at the University Paris-Saclay, who focus on the computational and circuits design components and the IAPP at TU-Dresden, who bring their expertise in the fabrication of OECT devices and circuits. The main funding of the project is to support two PhD students and the costs of their research. BAYOEN addresses several of the societal issues in the National AI strategies of both Germany and France including: new energy efficient hardware for AI, Green AI, AI for improving healthcare and protection of personal healthcare data.