Paper Title: Machine Learning and Non-Volatile Memories

Paper Abstract: The Talk will be based on the content of a new book published by Springer, titled "Machine Learning and Non-volatile Memories". At a first sight, ML and NVMs seem very far away from each other. ML implies mathematics, algorithms and a lot of computation; and NVMs are solid-state devices used to store information. However, these two worlds can work together. In particular, we can identify two main fields of application: analog Neural Networks (NNs) and SSDs. 1. NNs perform a specific computation called Vector-by-Matrix (VbM) multiplication. NVMs enable the analog implementation of the VbM, which can beat the equivalent digital implementation in terms of speed and power. 2. Inside SSDs, NANDs are managed by a microcontroller which executes the Flash management algorithms. For a storage designer it is difficult to directly access the internal architecture of the device, and therefore going to a higher abstraction level (i.e., algorithms handling data from/to the drive) is the only viable solution. These algorithms are usually referred to as prognostics tools, and ML has emerged as a very promising candidate in this context. All the above topics will be covered in the Talk.

Paper Author: Rino Micheloni, Research Fellow, University of Ferrara
Cristian Zambelli, Assistant Professsor, University of Ferrara (Italy)

Author Bio: Dr. Rino Micheloni is currently a Research Fellow at the University of Ferrara, Italy. Before that, he was Vice-President and Fellow at Microsemi/Microchip Corporation, where he established the Flash Signal Processing Labs in Milan, Italy, with special focus on NAND Flash technology characterization, Machine Learning techniques for improving memory reliability, and Error Correction Codes. Prior to joining Microsemi, he was Fellow at PMC-Sierra, working on NAND Flash technology characterization, LDPC, and NAND Signal Processing as part of the team developing Flash controllers for PCIe SSDs. Before that, he was with IDT (Integrated Device Technology) as Lead Flash Technologist, driving the architecture and design of the BCH engine in the world’s first PCIe NVMe SSD controller. Early in his career, he led NAND design teams at STMicroelectronics, Hynix, and Infineon; during this time, he developed the industry’s first MLC NOR device with embedded ECC technology and the industry’s first MLC NAND with embedded BCH. Dr. Micheloni is IEEE Senior Member, he has co-authored more than100 publications in peer-reviewed journals and international conferences, and he holds 291 patents worldwide (including 139 US patents). He received the STMicroelectronics Exceptional Patent Award in 2003 and 2004, the Infineon IP Award in 2007, and he was elected to the PMC-Sierra Inventor Wall of Fame in 2013. In 2020 Dr. Micheloni was selected for the European Inventor Award. Dr. Micheloni has published the following books: 3D Flash Memories (Chinese edition by Tsinghua University Press, 2020), Inside Solid State Drives – 2nd edition – (Springer, 2018), Solid-State-Drives (SSDs) Modeling (Springer, 2017), 3D Flash Memories (Springer, 2016), Inside Solid State Drives (Springer, 2013), Inside NAND Flash Memories (Springer, 2010), Error Correction Codes for Non-Volatile Memories (Springer, 2008), Memories in Wireless Systems (Springer, 2008), VLSI-Design of Non-Volatile Memories (Springer, 2005), and Memorie in Sistemi Wireless (Franco Angeli, 2005).

Author 2 Bio: