Manuele Rusci

MSCA Post-doctoral fellow at KU Leuven, PSI division

bio_conf.jpg

I am specialized in Embedded Machine Learning and Deep Learning (aka TinyML) for ultra-low power IoT sensor devices, and multi-core RISC-V Microcontrollers in particular. My contributions concern low-bitwith quantization methods and optimized software frameworks for inference on MCUs. I have also worked on full-system integration of smart audio or video sensor solutions with a power consumpions as low as few mWs (with GreenWaves Technologies).

Recently, my research focus has moved towards On-Device Learning to improve at runtime the predicion capacitiy of IoT smart sensors. In this space, I am investigating the integration of training algorithms on MCUs and efficient (continual) learning strategies for resource-constrained platforms. Thanks to my EU-funded Marie-Curie grant (on-going at KU Leuven with Prof. Tinne Tuytelaars), I am now studying novel adaptation mechanisms for ultra-low power speech recognition sensors (SEA2Learn project).

Previously, I was at Universita’ di Bologna (Prof. Benini’s group), where I was involved in the PULP project - targeting new RISC-V based IoT systems - and sevaral projects in the digital design domain, e.g., FPGA-based equipments for railway applications or, during my PhD, the design of smart camera sensors with event-based cameras.

News

Nov 30, 2023 The new tutorial On-Device Continual Learning Meets Ultra-Low Power Processing, organized with C. Cioflan (ETHZ) and D. Nadalini (UNIBO) has been accepted at DATE24 in Valencia!
Sep 1, 2023 Our paper On-device Customization of Tiny Deep Learning Models for Keyword Spotting with Few Examples, has been accepted for publication at IEEE micro!
Aug 22, 2023 Presenting our latest paper at Interspeech 2023 in the poster session.
Jul 15, 2023 New paper accepted for publication in the Special Issue of Future Generation Computer Systems journal: Reduced Precision Floating-Point Optimization for Deep Neural Network On-Device Learning on MicroControllers led by Davide Nadalini. The code is also available.
Jun 1, 2023 Landed in Grenoble :fr: at GreenWaves Technologies for the secondment (3 months) of my MSCA SEA2Learn project.

Selected Publications

  1. On-device Customization of Tiny Deep Learning Models for Keyword Spotting with Few Examples
    Manuele Rusci, and Tinne Tuytelaars
    IEEE Micro, 2023
  2. Reduced precision floating-point optimization for Deep Neural Network On-Device Learning on microcontrollers
    Davide Nadalini, Manuele Rusci, Luca Benini, and 1 more author
    Future Generation Computer Systems, 2023
  3. Accelerating RNN-Based Speech Enhancement on a Multi-core MCU with Mixed FP16-INT8 Post-training Quantization
    Manuele Rusci, Marco Fariselli, Martin Croome, and 2 more authors
    In Joint European Conference on Machine Learning and Knowledge Discovery in Databases, 2022
  4. A tinyml platform for on-device continual learning with quantized latent replays
    Leonardo Ravaglia, Manuele Rusci, Davide Nadalini, and 3 more authors
    IEEE Journal on Emerging and Selected Topics in Circuits and Systems, 2021
  5. Memory-driven mixed low precision quantization for enabling deep network inference on microcontrollers
    Manuele Rusci, Alessandro Capotondi, and Luca Benini
    Proceedings of Machine Learning and Systems, 2020
    Code
  6. PULP-NN: accelerating quantized neural networks on parallel ultra-low-power RISC-V processors
    Angelo Garofalo, Manuele Rusci, Francesco Conti, and 2 more authors
    Philosophical Transactions of the Royal Society A, 2020