BEGIN:VCALENDAR
VERSION:2.0
METHOD:PUBLISH
CALSCALE:GREGORIAN
PRODID:-//WordPress - MECv7.9.0//EN
X-ORIGINAL-URL:https://laboratoire-albert-fert.cnrs-thales.fr/
X-WR-CALNAME:Laboratoire Albert Fert
X-WR-CALDESC:CNRS - Thales
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-PUBLISHED-TTL:PT1H
X-MS-OLK-FORCEINSPECTOROPEN:TRUE
BEGIN:VEVENT
CLASS:PUBLIC
UID:MEC-f21e255f89e0f258accbe4e984eef486@laboratoire-albert-fert.cnrs-thales.fr
DTSTART:20261216T090000Z
DTEND:20261216T110000Z
DTSTAMP:20261103T125400Z
CREATED:20261103
LAST-MODIFIED:20261103
PRIORITY:5
SEQUENCE:16
TRANSP:OPAQUE
SUMMARY:PhD defense – Mohamed Menshawy
DESCRIPTION:On Wednesday, December 16, at 10:00 a.m. (C2N Amphitheater), Mohamed Menshawy will defend his PhD thesis titled “Programming of spintronic radio-frequency synapses through broadcast signals”, supervised by Julie Grollier and Frank Mizrahi.\nJury members:\nGregoire de Loubens, Research Director at CEA, Examiner and Chair\nTom Hayward, Professor at the University of Sheffield, Reviewer\nLiliana Buda, Professor at Grenoble INP, Reviewer\nAlex Jenkins, Researcher at INL, Examiner\nAbstract:\nSelectively programming a large number of non-volatile synaptic weights without compromising scalability is a key challenge for neuromorphic hardware. In this thesis, we experimentally demonstrate the programming of series-connected chains of 11 vortex-based magnetic tunnel junctions using broadcast radiofrequency signals applied through a shared strip line. Because each junction has a distinct resonance frequency, programming relies on frequency-selective reversal of the vortex-core polarity and does not require individual programming lines. Reconfiguring these binary magnetic states strongly modifies the continuous spectral responses used to process frequency-multiplexed RF inputs. Two chains are combined into a hardware network for handwritten-digit classification and drone RF-signature identification, reaching accuracies of 94.91 ± 0.26% and 97.33 ± 0.62%, respectively. Beyond inference, the binary vortex polarities are directly optimized using model-free training procedures, demonstrating that the physical magnetic parameters of the network can be trained for a target task. These results establish broadcast RF programming and hardware training as a route toward compact and reconfigurable spintronic neural networks.\n \n
URL:https://laboratoire-albert-fert.cnrs-thales.fr/en/events/phd-defense-maxime-ardisson-857/
CATEGORIES:PhD defence
LOCATION:C2N Amphitheater
END:VEVENT
END:VCALENDAR
