ICE Seminar: Computing with responsible electronics
To improve capabilities of future AI systems, researchers are exploring new computational paradigms and materials, which can exploit the specific needs of a set of applications and the interactions between hardware, software and algorithms in a computing system. An exciting example is to integrate the use of approximate computing, where the computational result is ‘good enough’. The computations are not as precise as in a conventional von Neumann architecture but there is significant savings in the number of circuit elements and power and typically the task to be solved involves some inherent error resilience. One such computational paradigm is stochastic computing (SC), which uses fewer gates and is highly fault-tolerant. Organic devices realized with low energy manufacturing processes are an exciting technology for realizing stochastic computing because they entail smaller circuits compared to conventional methodologies and variability can be used as an advantage for decorrelating the bitstreams in addition to having a very promising sustainability profile. In this talk I will present our latest results on how innovations in organic materials can be used in circuits to perform classification tasks using stochastic computing.
Bio
Laurie Calvet is a Franco-American scientist. Originally from the US, she earned her BS at Columbia University in NY in 1995, USA and her PhD at Yale University in Connecticut, USA in 2001. She arrived in France for a post-doc at CNRS-Thales and was then recruited into CNRS in 2007. She is currently chargée de recherche at LPICM, Ecole Polytechnique. Her current research focuses on taking ideas from biology and implementing them into new device concepts and circuits. She currently coordinates that EIC pathfinder grant BAYFLEX.