Current students

CATTANEO LORENZO

Cycle: XXXVI

Section: Electronics
Advisor: IELMINI DANIELE
Tutor: GERACI ANGELO

Major Research topic:
Stochastic devices and arrays for hardware security and neuromorphic computing

Abstract:

The research project focuses on the study, design, and testing of array-based solutions using novel stochastic memory devices for neuromorphic and security applications. It is a well-established fact that stochasticity plays a key role in neuromorphic computing operations, and embedded non-volatile memory technologies appear promising alternatives to CMOS circuits for the emulation of the neuronal behaviour in the nanoscale, thanks to their inherently random transport mechanisms and properties. Furthermore, stochastic phenomena in memory devices can also provide ideal sources of entropy for true random number generators (TRNGs) and physical unclonable functions (PUFs), reviving the interest in applications based on the stochastic computing paradigm, such as the probabilistic inference from real-time data. The availability of TRNG in stochastic memory devices, together with their high-density 3-D integration with crossbar arrays capability, can provide the necessary bitstream within a small area and low power consumption. In this scenario, Bayesian networks are an example of stochastic networks that can be used in a probabilistic inference engine to estimate the probability of hidden causes, and that can be mapped in memory arrays. In these circuits, probabilities are encoded within spiking signals generated by the TRNG, showing compact design, power efficiency, and high resilience to variations. Other examples of hardware stochastic networks already implemented in memory arrays are Markov chain Monte Carlo sampling (MCMC), reservoir computing (RC), and echo state networks (ESN). On the other hand, the ability to generate high-quality random bitstream at low power and with low area occupation is gaining interest also in the security field, where designing hardware-intrinsic primitives (PUFs) capable of generating unrelated responses to different input is an essential feature.