Kolloquium | Application-Oriented Memristor Co-Design | From Neuromorphic Synapses to Mixed-Mode In-Memory Logic
16. Juni 2026 | 10 Uhr c.t.
Dr. Nan Du von der Leibniz-IPHT hält am
16. Juni 2026 um
10 Uhr c.t.
im DFKI-Raum (MZH 4290)
ein Informatik-Kolloquium zum Thema "Application-Oriented Memristor Co-Design
From Neuromorphic Synapses to Mixed-Mode In-Memory Logic".
Wir freuen uns, Dich bei uns willkommen zu heißen.
Modern AI hardware is increasingly constrained not only by transistor scaling, but also by the energy and latency cost of moving data between memory and processing units. Memristors offer a different computing principle: the same nanoscale device can store information, process electrical signals, and adapt its internal state through rich, history-dependent dynamics. This makes them attractive building blocks for computing architectures that move beyond the classical separation of memory, logic, and learning. This talk presents memristors as adaptive computing elements in two complementary directions. The first part focuses on memristors for neuromorphic computing. Their analog, nonvolatile resistive switching enables artificial synapses that can emulate key biological learning functions, e.g. spike-timing- and spike-rate-dependent plasticity. These behaviors show how a single memristive device can replace complex CMOS synapse circuits and provide a compact route toward low-power brain-inspired hardware. The second part addresses memristive logic-in-memory computing. Here, memristive crossbar arrays are used not only for data storage, but also for logic processing directly inside the memory fabric. A mixed-mode computing framework is introduced, in which voltage-driven operations and resistance-state operations are co-designed to combine robustness, universality, and high parallelism. The approach reduces costly readout and data movement during logic cascading and is supported by crossbar-oriented synthesis and mapping. Together, these examples highlight a central message: memristors are not merely emerging memory devices, but application-dependent computing primitives. By co-designing device physics, circuit operation, architecture, and algorithms, memristive systems can enable energy-efficient hardware for brain-inspired learning, in-memory logic, and future adaptive AI systems.
Biografie
Dr. Nan Du is a Senior Scientist and Group Leader at the Leibniz Institute of Photonic Technology (Leibniz-IPHT), Germany, and holds a parallel group leadership position at the Institute of Solid-State Physics, Friedrich Schiller University Jena. She maintains close collaboration with industry, including TECHiFAB GmbH, enabling efficient translation of laboratory-scale nanodevices into scalable and manufacturing-ready technologies. Dr. Du received her Doctorate in Engineering (Dr.-Ing.) from Technische Universität Chemnitz and has more than 15 years of research experience in nanoscale memristor technology, spanning materials, devices, and system-level integration. Her research adopts a physics-aware, device–circuit co-design philosophy, leading to highly efficient and robust mixed-mode computing architectures and innovative in-memory computing architectures. She has demonstrated brain-inspired synaptic functionalities enabling advanced learning capabilities beyond conventional CMOS approaches. In parallel, she leads projects on security in emerging nanotechnologies, where her work has revealed fundamental physical mechanisms underlying side-channel leakage in nanomemory devices. Dr. Du has authored over 60+ publications, holds 10+ international patents. She is a principal investigator on projects funded by the German Research Foundation and serves as a coordination board member of the DFG Priority Program on Nano-Security.
Mehr Informationen gibt es bei Rolf Drechsler.
In X teilen
In LinkedIn teilen
