Valentin Radu

Lecturer (Assistant Professor), University of Sheffield

I lead the Ubiquitous AI Lab in the Department of Computer Science.
Before joining the University of Sheffield, I was a researcher at the University of Edinburgh, where I also received my PhD.

My research interests are in:
  • multimodal deep learning systems
  • distributed edge inference and training
  • deep neural network optimisation
  • context awareness by mobile sensing
For prospective PhD students (UK/EU): If you're interested in doing research in one of the topics mentioned above, please drop me an email.

News

[Jun. 2021] Two papers accepted at IPIN, 2021.

Joint work with Josh Barrows, Matthew Hill and Fabio Ciravegna, "Active Learning with Data Distribution Shift Detection for Updating Localization Systems".
Joint work with Xijia Wei, Zhiqiang Wei, "MM-Loc: Cross-sensor Indoor Smartphone Location Tracking using Multimodal Deep Neural Networks".

[Apr. 2021] Paper accepted at EdgeSys, collocated with EuroSys 2021.

Joint work with Hongrui Shi (University of Sheffield), "Towards Federated Learning with Attention Transfer to mitigate System and Data Heterogeneity of Clients".

[Mar. 2021] Paper accepted at EuroMLSys, collocated with EuroSys 2021.

Joint work with Rik Mulder (University of Edinburgh) and Christophe Dubach (McGill University), "Fast Optimisation of Convolutional Neural Network Inference using System Performance Models".

[Mar. 2021] I became a member of HiPEAC, the center of excellence in research for the future of computing systems in Europe.

[Mar. 2021] I became a member of the Insigneo Institute for in silico Medicine.

[Dec. 2020] We released the MM-Fit dataset that accompanies our IMWUT paper.

Let us know if you find this useful and we can spotlight your research on our project website.

[Oct. 2020] Paper accepted to appear in ACM IMWUT, vol. 4, issue 4 and to be presented at Ubicomp 2021.

Joint work with David Strömbäck (University of Edinburgh) and Sangxia Huang (Sony Sweden), "MM-Fit: Multimodal Deep Learning for Automatic Exercise Logging Across Sensing Devices"

[Jul. 2020] We're organising the 3rd Workshop on Accelerated Machine Learning (AccML) at HiPEAC 2021, Budapest

[Aug. 2019] Paper accepted to be presented at IISWC 2019.

Our team in Edinburgh, in collaboration with colleagues at Trinity College Dublin and University of Glasgow - "Performance Aware Convolutional Neural Network Channel Pruning for Embedded GPUs"

[Jul. 2019] We're organising the Accelerated Machine Learning (AccML) workshop at HiPEAC 2020, Bologna ("the largest computing systems research event in Europe").

Co-organising with colleagues at University of Glasgow, DeepMind and Google. Consider presenting your work with us!

[Jul. 2019] Paper accepted to appear in ACM IMWUT, vol. 3, issue 3 and to be presented at Ubicomp 2019.

Joint work with Maximilian Henne (University of Edinburgh) - "Vision2Sensor: Knowledge Transfer Across Sensing Modalities for Human Activity Recognition"

"The idea is quite interesting as it tackles one of the main challenges of the Ubicomp community when it comes to HAR: the annotation of data." (Anonymous Reviewer)

[Jul. 2019] 2 papers accepted to be presented at IPIN 2019. See you in Pisa, Italy

"Calibrating Recurrent Neural Networks on Smartphone Inertial Sensors for Location Tracking", with Xijia Wei (University of Edinburgh); and "CamLoc: Pedestrian Location Estimation through Body Pose Estimation on Smart Cameras", with Adrian Cosma (University Politehnica of Bucharest) and Ion Emilian Radoi (University Politehnica of Bucharest)

"One trillion new IoT devices will be produced between 2015 and 2035." (ARM)

Our research brings intelligence to many of these ubiquitous devices.

Research Team

I have had the pleasure of working with some amazing people in the Ubiquitous AI Lab:

Hongrui Shi (PhD, 2020 - )

Josh Barrows (UG)

Harry Woods (UG)

Alumni

Sudharshan Paindi Jayakumar

MSc thesis (2020): Anomaly Detection in Sensor Data through Multimodal Deep Neural Networks

David Strömbäck

MSc thesis (2019): Multimodal Learning for Automatic Exercise Logging (Best Thesis Awared)

David Ivorra

MSc thesis (2019): Efficient Motility Characterization at the Edge

Luca McArthur

MSc thesis (2019): Multi-task Neural Architecture Search for Edge Computing

Xingji Chen

MSc thesis (2018): "WiFi-based Indoor Localization using Deep Neural Networks on Smartphones"

Xijia Wei

MSc thesis (2018): "Smartphone-based Location Tracking using Recurrent Neural Networks"

Maximilian Henne

MSc thesis (2018): "Automatic Ground-truth Collection in Mobile Systems Through Multimodal Sensing"

Selected publications

The full list is on Google Scholar

  • [EuroMLSys'21] Fast Optimisation of Convolutional Neural Network Inference using System Performance Models
    Rik Mulder, Valentin Radu, Christophe Dubach
    In the ACM Workshop on Machine Learning and Systems, April 2021.
    [DOI] [PDF]
  • [EdgeSys'21] Towards Federated Learning with Attention Transfer to Mitigate System and Data Heterogeneity of Clients
    Hongrui Shi, Valentin Radu
    In ACM Edge Systems, Analytics and Networking, April 2021.
    [DOI] [PDF]
  • [IMWUT'20] MM-Fit: Multimodal Deep Learning for Automatic Exercise Logging Across Sensing Devices
    David Strömbäck, Sangxia Huang, Valentin Radu
    In ACM Interactive, Mobile, Wearable and Ubiquitous Technologies (IMWUT): Volume 4 Issue 4, December 2020.
    [DOI] [PDF] [bib] [dataset]
  • [IMWUT'19] Vision2Sensor: Knowledge Transfer Across Sensing Modalities for Human Activity Recognition
    Valentin Radu, Maximilian Henne
    In ACM Interactive, Mobile, Wearable and Ubiquitous Technologies: Volume 3 Issue 3, September 2019.
    [DOI] [PDF] [bib]
  • [IISWC'19] Performance Aware Convolutional Neural Network Channel Pruning for Embedded GPUs
    V. Radu, K. Kaszyk, Y. Wen, J. Turner, J. Cano, E. J. Crowley, B. Franke, A. Storkey, M. O’Boyle
    In the International Symposium on Workload Characterization (IISWC), IEEE, Orlando, USA, November 2019.
    [PDF]
  • [IPIN'19] CamLoc: Pedestrian Location Estimation through Body Pose Estimation on Smart Cameras
    Adrian Cosma, Ion Emilian Radoi, Valentin Radu
    In IEEE Indoor Positioning and Indoor Navigation (IPIN), Octomber 2019.
    [DOI] [PDF] [bib]
  • [IISWC'18] Characterising Across-Stack Optimisations for Deep Convolutional Neural Networks
    J. Turner, J. Cano, V. Radu, E. J. Crowley, M. O'Boyle, and A. Storkey
    In the International Symposium on Workload Characterization (IISWC), IEEE, North Carolina, USA, September 2018.
  • [IMWUT'17] Multimodal Deep Learning for Activity and Context Recognition
    Valentin Radu, Catherine Tong, Sourav Bhattacharya, Nicholas D. Lane, Cecilia Mascolo, Mahesh K. Marina, Fahim Kawsar
    In ACM Interactive, Mobile, Wearable and Ubiquitous Technologies: Volume 1 Issue 4, December 2017.
    [DOI] [PDF]
  • [UbiComp'16] Towards Multimodal Deep Learning for Activity Recognition on Mobile Devices
    Valentin Radu, Nicholas D. Lane, Sourav Bhattacharya, Cecilia Mascolo, Mahesh K. Marina, Fahim Kawsar
    In the Joint Conference on Pervasive and Ubiquitous Computing (UbiComp/ISWC'16) Adjunct, ACM, September 2016, Heidelberg, Germany.
    [DOI] [PDF] [bib]
  • [ATC'15] Impact of Indoor-Outdoor Context on Crowdsourcing based Mobile Coverage Analysis
    Mahesh K. Marina, Valentin Radu, Konstantinos Balampekos
    In All Things Cellular Workshop (ATC) collocated with SIGCOMM'15, ACM, August 2015, London, UK.
    [DOI] [PDF] [bib]
  • [SenSys'14] A Semi-Supervised Learning Approach for Robust Indoor-Outdoor Detection with Smartphones
    Valentin Radu, Panagiota Katsikouli, Rik Sarkar and Mahesh K. Marina
    In the Conference on Embedded Networked Sensor Systems (SenSys), ACM, November 2014, Memphis, USA.
    [DOI] [PDF] [bib]
  • [MobiCom'14] Poster: Am I Indoor or Outdoor?
    Valentin Radu, Panagiota Katsikouli, Rik Sarkar, Mahesh K. Marina
    In the 20th Annual International Conference on Mobile Computing and Networking (MobiCom), ACM, September 2014, Hawaii, USA.
    [DOI] [PDF] [bib] [poster]
  • [IPIN'13] HiMLoc: Indoor Smartphone Localization via Activity Aware Pedestrian Dead Reckoning with Selective Crowdsourced WiFi Fingerprinting - Nominated for the Best Paper Award
    Valentin Radu, Mahesh K. Marina
    In the 4th International Conference on Indoor Positioning and Indoor Navigation (IPIN), IEEE, October 2013, Belfort-Montbeliard, France.
    [DOI] [PDF] [bib]
  • [CNSM'13] Pazl: A Mobile Crowdsensing based Indoor WiFi Monitoring System
    Valentin Radu, Lito Kriara, Mahesh K. Marina
    In the 9th International Conference on Network and Service Management (CNSM), IEEE, October 2013, Zurich, Switzerland.
    [DOI] [PDF] [bib] [Presentation]

Contact

+44 114 222 1971

valentin.radu@sheffield.ac.uk

University of Sheffield, Department of Computer Science,
Regent Court (DCS), 211 Portobello, Sheffield, S1 4DP