Titles, Abstracts, and Bios for the talks, posters and demos in the Program
In-Memory Computing-based Deep Learning Accelerators: An Overview and Future Prospects
Abu Sebastian, IBM Research - Zurich
A status update on the most promising approach towards deep learning acceleration.
Bio: Distinguished scientist at IBM Research Zurich
Device integration of graphene nanoribbons
Mikael Perrin, ETHZ/Empa
In this talk, I'll cover our recent efforts in integrating bottom-up synthesized graphene nanoribbons into electronic quantum devices.
Bio: Mickael Perrin is a group leader at Empa and assistant professor at ETH Zurich. He recently was awarded the ERC starting grant and SNSF Eccellenza professorial fellowship to investigate quantum heat engines for efficient energy conversion at the nanoscale.
Introducing SwissChips
Frank K. Gürkaynak, ETH Zürich
A short introduction to SwissChips which is a three year program to support academic IC Design in Switzerland
Bio: Frank is a Senior Scientist at the Integrated Systems Laboratory of ETH Zürich and also leads the Microelectronics Design Center.
Placeholder, C. Studer may give the talk
20 years of Bio/CMOS interfaces in Biomedical
Sandro Carrara, EPFL
The emerging era of wearables leads to distributed diagnostics. To that, new and definitely more reliable and specific devices are required. Therefore, we will discuss new Bio/CMOS to this aim.
Bio: IEEE Fellow, IEEE Sensors Council Technical Achievement Award in 2016, and Scientist Medal by the International Association of Advanced Materials in 2024. 7 books, 400 papers, 19 patents.
From Powering low power Wireless to Empowering 6G Communication Systems
Mathieu Coustans, FHNW
This talk highlights the evolution from research on autonomous low-power wireless networks to cutting-edge developments in 6G optical networks.
Bio: Ph.D. from EPFL. With extensive industry experience, he co-founded Riot-Micro, worked for renowned semiconductor companies (EM, GF, STM, Sony), now a Professor at FHNW. Holds four patents and received the IEEE ISSCC Jan van Vessem Award in 2021.
Daniele Palossi, IDSIA USI-SUPSI and ETH Zurich
This talk will explore miniaturized autonomous robots characterized by a mW-scale computational power envelope. The talk will dive into their onboard ultra-low-power brains, their TinyML-based perception pipelines, and some real-world applications.
Bio: Daniele Palossi is a Senior Researcher at the IDSIA, USI-SUPSI, where he leads the nanorobotics research group, and at the IIS, ETH Zürich. His research stands at the intersection of AI, ultra-low-power embedded systems, and miniaturized robotics.
1. Moody - ULP emotion detection
Pierre-François Rüedi & Petar Jokic, CSEM
Based on CSEM's ULP imager and ULP ML accelerator and powered by a coin cell, Moody embeds a small game controlled with your face emotions.
2. Spoken Language Understanding with Tiny RNN
Longbiao Cheng, Sensors Group, INII, UZH-ETHZ
This demo shows that even a 42k-parameter RNN with 16-dimensional input can achieve high accuracy in a 32-class spoken language understanding task while being robust to real-world acoustic environments.
3. Towards Novel Human-Machine Interfaces: An ultrasound centric approach
Giusy Spacone, ETH Zurich - IIS
We demonstrate real-time gesture recognition using an armband powered by WULPUS, an ultra-low-power and truly wearable sensing platform for ultrasound data acquisition and real-time computation.
4. Dextra - the rock-scissors-paper robot
Xiang Deng and Tobi Delbruck, Sensors Group, INII, UZH-ETHZ
Dextra beats humans at the classic guessing game of rock-scissors-paper by turning it into a physical sport of raw speed. It demonstrates activity-driven-perception by combining a quick event camera, a small CNN, and a quick tendon-driven hand.
1. SwissChips
Tracy Ewen and Frank K. Gürkaynak, IIS ETHZ
Introduces SwissChips, which is a three year program to support academic IC Design in Switzerland
2. Experimental Design of Ferroelectric Devices for Neuromorphic Hardware
Anwesha Panda, Jiajie Gao, Laura Bégon-Lours. Neuromorphic Electronics with Oxides Lab, ITET ETHZ
We outline a concise workflow, from concept and design to fabrication, for HfZrO₂-based ferroelectric devices. By enabling multi-functional testing on a single chip and exploring diverse channel materials, we aim to advance neuromorphic hardware.
Anwesha Panda holds a M.Sc in Materials Science and is currently pursuing a PhD at the Neuromorphic Electronics with Oxides group (IIS, D-ITET, ETHZ).
3. Memristor-based Perceptron trained with a Bio-Inspired Learning Rule
Long Ching Ip. Ashira Long Ching Ip & Pau Vilimelis Aceituno, Institute of Neuroinformatics, University of Zurich and ETH Zurich
Nadia Jimenez Olalla, Institute of Electromagnetic Fields, ETH Zurich
We present memristive circuits that can be trained in the loop with a bio-inspired learning rule to act as perceptrons.
4. Three Terminal Memristive Artificial Neuron with Tunable Firing Probability
Kshipra Srikrishnaprabhu, Institute of Electromagnetic Fields, ETH Zurich
Lewerenz, Mila ; Passerini, Elias ; Weber, Luca; Fischer, Markus; Olalla, Nadia Jimenez; Gisler, Raphael; Emboras, Alexandros; Luisier, Mathieu ; Csontos, Miklos ; Koch, Ueli ; Leuthold, Juerg
Compact and simplistic spiking neural network hardware is necessary to efficiently implement neuromorphic functions on the device level. We present a versatile, ultra-small footprint three terminal memristive artificial neuron where the set voltage of the memristor, and in turn the spiking probability of the neuron can be tuned by the gate voltage.
5. Neuronal and Synaptic functionalities integrated in a single memristive device
Nadia Jimenez Olalla. ETH Zürich
Elias Passerini, Mila Lewerenz, Arnaud Schneuwly, Nadia Jimenez Olalla, Markus Fischer, Raphael Gisler, Alexandros Emboras, Yuriy Fedoryshyn, Mathieu Luisier, Thomas Schimmel, Miklós Csontos, Ueli Koch and Juerg Leuthold
Resistive Random Access Memories (RRAM) hold great potential to be used for Neuromorphic Computing applications. In this work, we present a simple, reliable resistive system which can emulate both synaptic and neuronal functionalities in the same device.
Nadia Jimenez Olalla is pursuing a PhD at ETH Zürich to develop memristive devices as a platform to develop neuromorphic applications.
6. Picosecond Femtojoule Resistive Switching in Nanoscale VO2 Memristors
Miklos Csontos. ETH IEF
S. W. Schmid, L. Pósa, T. N. Török, B. Sánta, Z. Pollner, G. Molnár, Y. Horst, J. Volk, J. Leuthold, A. Halbritter and M. Csontos
We demonstrate tunable low-resistance states in nanoscale VO2 memristors set within 15 ps and switching energies starting from 1 fJ. The high-resistance state can be recovered in 600 ps.
Dr. Miklos Csontos received his diploma in engineering-physics (2002) and PhD in physics (2007) at the Budapest University of Technology and Economics. Currently he is a researcher at ETH Zurich.
7. Continuous-time RC-Chain ADC
Hampus Malmberg, ETHZ
An amplifier-less continuous-time analog-to-digital converter consisting of only passives, comparators, and inverters is presented
Hampus Malmberg received the B.S. degree in electrical engineering from Chalmers University in 2012, the M.Sc. and Ph.D. degrees in electrical engineering from ETH Zürich, Zürich, Switzerland, in 2014 and 2020, respectively. He is currently a Postdoctoral Researcher with ETH Zürich.
8. From Picoseconds to Biological Timescales: Conductance Changes in Hafnia Synapses
Alexandre Baigol Sisó, Laura Bégon-Lours, Neuromorphic Electronics with Oxides Lab, ITET ETHZ
The switching speed and ultrafast dynamics of HfZrO4-based ferroelectric devices are investigated to establish a baseline for future neuromorphic applications. We identify 3 critical parameters: minimum pulse time, voltage ranges, and device size.
Alexandre Baigol Sisó is a PhD student at the Neuromorphic Electronics with Oxides group (IIS, DITET, ETHZ) focusing on ferroelectric materials for artificial synapses. He has MSc in Materials Science and a BS in Materials Engineering.
9. Scaling Effects of Transistor Leakage Current and IR Drop on 1T1R Memory Arrays
Junren Chen, INI, UZH-ETHZ
Junren Chen, Giacomo Indiveri
We found that an optimal resistance range of memory cells exists for good 1T1R array scaling capability, bounded by the joint effects of IR drop and transistor leakage.
10. A 28nm 576k RRAM-Based Computing-in-Memory Macro with Direct-Current ADC Architecture for Optimizing On-Chip Programming
Siqi Liu, INI, UZH-ETHZ
A 28nm RRAM-based CIM macro with a novel direct-current ADC achieves 4.67× programming speed, 0.15× power savings, and 4.31× compact weight distribution.
11. SRAM in-memory computing
Anqgi Liu, INI, UZH-ETHZ
This poster presents an in-depth analysis of SRAM-based in-memory computing for Spiking Neural Networks and conventional Artificial Neural Networks, with comparative study of Analog In-Memory Computing and Digital In-Memory Computing. It also introduces “Lokum,” a custom research chip developed to investigate innovative approaches in charge-domain and current-domain computing.
12. Event-based perception for nano-UAVs
Lorenzo Lamberti, University of Bologna
Lorenzo Lamberti, Alfio di Mauro, Francesco Conti, Daniele Palossi, Luca Benini
This poster explores event-based perception on Nano-UAVs. We introduce the Kraken shield, a 7g multi-sensor board for Nano-UAVs, featuring event- and frame-based imagers and a Parallel Ultra-Low-Power SoC featuring spiking and ternary neural networks accelerators.
13. Training on the Fly: On-Device Self-Supervised Learning Aboard Nano-Drones Within 20 mW
Elia Cereda, IDSIA, USI-SUPSI
Elia Cereda*, Alessandro Giusti*, and Daniele Palossi*^. *IDSIA, USI-SUPSI. ^IIS, ETH Zürich
TinyML models made autonomous sub-50g nano-drones possible despite tight hardware constraints. However, these models still suffer in unseen environments from domain shift. We propose on-device self-supervised learning as a solution.