Manon Dampfhoffer
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Manon Dampfhoffer has a total of 47 co-authors. Here are the top co-authors by number of shared publications:
- Thomas Mesquida
CEA-List, Université Grenoble Alpes
7 shared publications - Hiroaki Nishi
Graduate School of Science and Technology, Keio University, Yokohama, Japan
4 shared publications - Hiroshi Nakano
Graduate School of Science and Technology, Keio University, Yokohama, Japan
4 shared publications - Kamil Jeziorek
Embedded Vision Systems Group AGH University of Krakow Poland
4 shared publications - Krzysztof Blachut
AGH University of Krakow
4 shared publications - Piotr Wzorek
AGH University of Krakow
4 shared publications
Latest Publications
- Open Access
- 8 September 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Reconfigurable Technology and Systems
- Vol. 19 (3) , 1-34
Hardware-Accelerated Graph Neural Networks: An Alternative Approach for Event-Based Audio Classification and Keyword Spotting on SoC FPGA
As the volume of data recorded by embedded edge sensors increases, particularly from neuromorphic devices producing discrete event streams, there is a need for hardware-aware neural architectures that enable efficient, low-latency and energy-conscious local processing. To address this research gap, we present an FPGA implementation of event-graph neural networks for audio processing. We utilise an artificial cochlea that converts time-series signals into sparse event data, reducing memory and computation costs. Our architecture was implemented on an SoC FPGA and evaluated on two open source datasets. For the classification task, our baseline floating-point model achieves 92.7% accuracy on the SHD dataset—only 2.4% and 2% below the state-of-the-art—while requiring 10 \(\times\) and 67 \(\times\) fewer parameters, respectively. Compared to FPGA-based spiking neural networks, our quantised model reaches 92.3% accuracy, outperforming them by 19.3 pps while reducing resource usage and latency. For the SSC dataset, we report the first hardware-accelerated evaluation, and our models achieve 66.9–71.0% accuracy. Extending our conference work, we further demonstrate the first end-to-end FPGA implementation of the event-audio keyword spotting task, combining graph convolutional layers with recurrent sequence modelling for continuous, streaming operation. The system achieves 95% word-end detection accuracy, with only 10.53 µs latency and 1.18 W power consumption, establishing a strong benchmark for energy-efficient event-driven KWS.- 8 September 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Reconfigurable Technology and Systems
- Vol. 19 (3) , 1-34
Open AccessHardware-Accelerated Graph Neural Networks: An Alternative Approach for Event-Based Audio Classification and Keyword Spotting on SoC FPGA
As the volume of data recorded by embedded edge sensors increases, particularly from neuromorphic devices producing discrete event streams, there is a need for hardware-aware neural architectures that enable efficient, low-latency and energy-conscious local processing. To address this research gap, we present an FPGA implementation of event-graph neural networks for audio processing. We utilise an artificial cochlea that converts time-series signals into sparse event data, reducing memory and computation costs. Our architecture was implemented on an SoC FPGA and evaluated on two open source datasets. For the classification task, our baseline floating-point model achieves 92.7% accuracy on the SHD dataset—only 2.4% and 2% below the state-of-the-art—while requiring 10 \(\times\) and 67 \(\times\) fewer parameters, respectively. Compared to FPGA-based spiking neural networks, our quantised model reaches 92.3% accuracy, outperforming them by 19.3 pps while reducing resource usage and latency. For the SSC dataset, we report the first hardware-accelerated evaluation, and our models achieve 66.9–71.0% accuracy. Extending our conference work, we further demonstrate the first end-to-end FPGA implementation of the event-audio keyword spotting task, combining graph convolutional layers with recurrent sequence modelling for continuous, streaming operation. The system achieves 95% word-end detection accuracy, with only 10.53 µs latency and 1.18 W power consumption, establishing a strong benchmark for energy-efficient event-driven KWS. - 23 May 2026
- Published by Springer Nature
- p. 3-35
Understanding Neural Network Fault Tolerance: From Analog Hardware to the GPU
- 23 May 2026
- Published by Springer Nature
- p. 3-35
Understanding Neural Network Fault Tolerance: From Analog Hardware to the GPU
- Open Access
- 16 May 2026
- Published by Elsevier in SSRN Electronic Journal
Laboratory and season-long field evaluation of AI-based predictive control for residential heat pumps in DHW and space-heating operation
Residential heat pumps are commonly controlled using static rule-based controllers such as fixed heating curves and hysteresis-based DHW control, which can lead to unnecessary temperature lifts and avoidable storage losses as building and occupancy conditions change. This paper presents two AI-enabled predictive control strategies for (i) DHW production and (ii) space-heating supply-temperature control. In both cases, forecasting/modeling and optimization are separated to facilitate implementation. For DHW, a learned demand forecast is combined with Modelica-based simulation-in-the-loop scheduling that minimizes electrical energy while enforcing service constraints. Laboratory experiments show reductions in electrical energy use of 5-8% for an individual-housing configuration and 38-43% for a collective-housing configuration, together with a substantial increase in the heat-pump coefficient of performance (COP). For space heating, an attention-augmented encoder-decoder LSTM temperature predictor is embedded in a fast surrogate-assisted receding-horizon controller. A field trial in one residential building during the 2024-2025 heating season showed an 8% reduction in daily electricity use, an increase in mean COP from 3.95 to 4.15, and a reduction in mean absolute comfort deviation from 0.3 to 0.1 K. The results demonstrate that forecast-informed optimization can improve heat-pump efficiency while maintaining comfort, although broader field validation across building types and climates remains necessary.
- 16 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessLaboratory and season-long field evaluation of AI-based predictive control for residential heat pumps in DHW and space-heating operation
Residential heat pumps are commonly controlled using static rule-based controllers such as fixed heating curves and hysteresis-based DHW control, which can lead to unnecessary temperature lifts and avoidable storage losses as building and occupancy conditions change. This paper presents two AI-enabled predictive control strategies for (i) DHW production and (ii) space-heating supply-temperature control. In both cases, forecasting/modeling and optimization are separated to facilitate implementation. For DHW, a learned demand forecast is combined with Modelica-based simulation-in-the-loop scheduling that minimizes electrical energy while enforcing service constraints. Laboratory experiments show reductions in electrical energy use of 5-8% for an individual-housing configuration and 38-43% for a collective-housing configuration, together with a substantial increase in the heat-pump coefficient of performance (COP). For space heating, an attention-augmented encoder-decoder LSTM temperature predictor is embedded in a fast surrogate-assisted receding-horizon controller. A field trial in one residential building during the 2024-2025 heating season showed an 8% reduction in daily electricity use, an increase in mean COP from 3.95 to 4.15, and a reduction in mean absolute comfort deviation from 0.3 to 0.1 K. The results demonstrate that forecast-informed optimization can improve heat-pump efficiency while maintaining comfort, although broader field validation across building types and climates remains necessary.
- 8 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Models and algorithms for implementing energy-efficient spiking neural networks on neuromorphic hardware at the edge
Modèles et algorithmes pour l'implémentation de réseaux de neurones impulsionnels à faible consommation énergétique sur du matériel neuromorphique: L'apprentissage profond dans les réseaux de neurones artificiels (ANNs), une branche de l'intelligence artificielle (IA), est considéré comme une révolution dans l'informatique et a un impact sur tous les secteurs de l'économie. Cependant, les ANNs sont très gourmands en ressources de calcul et en mémoire, ce qui limite leur intégration à la périphérie du réseau pour des applications embarquées. Les réseaux de neurones impulsionnels (SNNs) sont des alternatives prometteuses aux ANNs en termes d'efficacité énergétique et sont donc de bons candidats pour les implémentations IA embarquées utilisant du matériel neuromorphique. En effet, les SNNs encodent les informations en utilisant des événements temporels épars (appelés « spikes ») au lieu d'activations denses et précises. Cependant, l'écart entre le développement algorithmique des SNNs d'une part, et leur implémentation matérielle d'autre part, rend difficile l'obtention de solutions réellement efficaces. Dans ce contexte, cette thèse suit une approche tenant compte du matériel pour conduire les développements algorithmiques des SNNs. En particulier, des modèles et des algorithmes sont proposés pour améliorer la précision et l'efficacité énergétique des SNNs, en considérant des implémentations matérielles numériques et analogiques.Afin de comparer les implémentations des SNNs et des ANNs sur des accélérateurs de réseaux de neurones dédiés, un modèle de leur efficacité énergétique est fourni. En particulier, on constate que la parcimonie des activations joue un rôle clé dans l'efficacité des SNNs. Par conséquent, un nouveau modèle de SNN, SpikGRU, est proposé. Remarquablement, il combine la précision des ANNs récurrents à porte avec une parcimonie des activations, et s'avère très efficace pour les applications de faible puissance telles que le repérage de mots clés. En outre, l'implémentation des poids synaptiques utilisant des mémoires analogiques non volatiles est envisagée pour augmenter encore l'efficacité énergétique. Avec une méthodologie d'apprentissage adaptée, les SNNs se révèlent très robustes à ces poids hautement quantifiés et avec un haut niveau de bruit. Une étude de cas utilisant des mémoires résistives valide l'approche.En encourageant le co-développement algorithme-matériel, ce travail vise à ouvrir la voie à des implémentations efficaces de réseaux de neurones embarqués.
- 8 April 2026
- Published by Agence Bibliographique de l'Enseignement Supérieur
Models and algorithms for implementing energy-efficient spiking neural networks on neuromorphic hardware at the edge
Modèles et algorithmes pour l'implémentation de réseaux de neurones impulsionnels à faible consommation énergétique sur du matériel neuromorphique: L'apprentissage profond dans les réseaux de neurones artificiels (ANNs), une branche de l'intelligence artificielle (IA), est considéré comme une révolution dans l'informatique et a un impact sur tous les secteurs de l'économie. Cependant, les ANNs sont très gourmands en ressources de calcul et en mémoire, ce qui limite leur intégration à la périphérie du réseau pour des applications embarquées. Les réseaux de neurones impulsionnels (SNNs) sont des alternatives prometteuses aux ANNs en termes d'efficacité énergétique et sont donc de bons candidats pour les implémentations IA embarquées utilisant du matériel neuromorphique. En effet, les SNNs encodent les informations en utilisant des événements temporels épars (appelés « spikes ») au lieu d'activations denses et précises. Cependant, l'écart entre le développement algorithmique des SNNs d'une part, et leur implémentation matérielle d'autre part, rend difficile l'obtention de solutions réellement efficaces. Dans ce contexte, cette thèse suit une approche tenant compte du matériel pour conduire les développements algorithmiques des SNNs. En particulier, des modèles et des algorithmes sont proposés pour améliorer la précision et l'efficacité énergétique des SNNs, en considérant des implémentations matérielles numériques et analogiques.Afin de comparer les implémentations des SNNs et des ANNs sur des accélérateurs de réseaux de neurones dédiés, un modèle de leur efficacité énergétique est fourni. En particulier, on constate que la parcimonie des activations joue un rôle clé dans l'efficacité des SNNs. Par conséquent, un nouveau modèle de SNN, SpikGRU, est proposé. Remarquablement, il combine la précision des ANNs récurrents à porte avec une parcimonie des activations, et s'avère très efficace pour les applications de faible puissance telles que le repérage de mots clés. En outre, l'implémentation des poids synaptiques utilisant des mémoires analogiques non volatiles est envisagée pour augmenter encore l'efficacité énergétique. Avec une méthodologie d'apprentissage adaptée, les SNNs se révèlent très robustes à ces poids hautement quantifiés et avec un haut niveau de bruit. Une étude de cas utilisant des mémoires résistives valide l'approche.En encourageant le co-développement algorithme-matériel, ce travail vise à ouvrir la voie à des implémentations efficaces de réseaux de neurones embarqués.
- Open Access
Hardware-accelerated graph neural networks: an alternative approach for neuromorphic event-based audio classification and keyword spotting on SoC FPGA
As the volume of data recorded by embedded edge sensors increases, particularly from neuromorphic devices producing discrete event streams, there is a growing need for hardware-aware neural architectures that enable efficient, low-latency, and energy-conscious local processing. We present an FPGA implementation of event-graph neural networks for audio processing. We utilise an artificial cochlea that converts time-series signals into sparse event data, reducing memory and computation costs. Our architecture was implemented on a SoC FPGA and evaluated on two open-source datasets. For classification task, our baseline floating-point model achieves 92.7% accuracy on SHD dataset - only 2.4% below the state of the art - while requiring over 10x and 67x fewer parameters. On SSC, our models achieve 66.9-71.0% accuracy. Compared to FPGA-based spiking neural networks, our quantised model reaches 92.3% accuracy, outperforming them by up to 19.3% while reducing resource usage and latency. For SSC, we report the first hardware-accelerated evaluation. We further demonstrate the first end-to-end FPGA implementation of event-audio keyword spotting, combining graph convolutional layers with recurrent sequence modelling. The system achieves up to 95% word-end detection accuracy, with only 10.53 microsecond latency and 1.18 W power consumption, establishing a strong benchmark for energy-efficient event-driven KWS.
Open AccessHardware-accelerated graph neural networks: an alternative approach for neuromorphic event-based audio classification and keyword spotting on SoC FPGA
As the volume of data recorded by embedded edge sensors increases, particularly from neuromorphic devices producing discrete event streams, there is a growing need for hardware-aware neural architectures that enable efficient, low-latency, and energy-conscious local processing. We present an FPGA implementation of event-graph neural networks for audio processing. We utilise an artificial cochlea that converts time-series signals into sparse event data, reducing memory and computation costs. Our architecture was implemented on a SoC FPGA and evaluated on two open-source datasets. For classification task, our baseline floating-point model achieves 92.7% accuracy on SHD dataset - only 2.4% below the state of the art - while requiring over 10x and 67x fewer parameters. On SSC, our models achieve 66.9-71.0% accuracy. Compared to FPGA-based spiking neural networks, our quantised model reaches 92.3% accuracy, outperforming them by up to 19.3% while reducing resource usage and latency. For SSC, we report the first hardware-accelerated evaluation. We further demonstrate the first end-to-end FPGA implementation of event-audio keyword spotting, combining graph convolutional layers with recurrent sequence modelling. The system achieves up to 95% word-end detection accuracy, with only 10.53 microsecond latency and 1.18 W power consumption, establishing a strong benchmark for energy-efficient event-driven KWS.
- 4 April 2025
- Published by Springer Nature
- p. 51-68
Hardware-Accelerated Event-Graph Neural Networks for Low-Latency Time-Series Classification on SoC FPGA
- 4 April 2025
- Published by Springer Nature
- p. 51-68
Hardware-Accelerated Event-Graph Neural Networks for Low-Latency Time-Series Classification on SoC FPGA
- Open Access
Hardware-Accelerated Event-Graph Neural Networks for Low-Latency Time-Series Classification on SoC FPGA
As the quantities of data recorded by embedded edge sensors grow, so too does the need for intelligent local processing. Such data often comes in the form of time-series signals, based on which real-time predictions can be made locally using an AI model. However, a hardware-software approach capable of making low-latency predictions with low power consumption is required. In this paper, we present a hardware implementation of an event-graph neural network for time-series classification. We leverage an artificial cochlea model to convert the input time-series signals into a sparse event-data format that allows the event-graph to drastically reduce the number of calculations relative to other AI methods. We implemented the design on a SoC FPGA and applied it to the real-time processing of the Spiking Heidelberg Digits (SHD) dataset to benchmark our approach against competitive solutions. Our method achieves a floating-point accuracy of 92.7% on the SHD dataset for the base model, which is only 2.4% and 2% less than the state-of-the-art models with over 10% and 67% fewer model parameters, respectively. It also outperforms FPGA-based spiking neural network implementations by 19.3% and 4.5%, achieving 92.3% accuracy for the quantised model while using fewer computational resources and reducing latency.
Open AccessHardware-Accelerated Event-Graph Neural Networks for Low-Latency Time-Series Classification on SoC FPGA
As the quantities of data recorded by embedded edge sensors grow, so too does the need for intelligent local processing. Such data often comes in the form of time-series signals, based on which real-time predictions can be made locally using an AI model. However, a hardware-software approach capable of making low-latency predictions with low power consumption is required. In this paper, we present a hardware implementation of an event-graph neural network for time-series classification. We leverage an artificial cochlea model to convert the input time-series signals into a sparse event-data format that allows the event-graph to drastically reduce the number of calculations relative to other AI methods. We implemented the design on a SoC FPGA and applied it to the real-time processing of the Spiking Heidelberg Digits (SHD) dataset to benchmark our approach against competitive solutions. Our method achieves a floating-point accuracy of 92.7% on the SHD dataset for the base model, which is only 2.4% and 2% less than the state-of-the-art models with over 10% and 67% fewer model parameters, respectively. It also outperforms FPGA-based spiking neural network implementations by 19.3% and 4.5%, achieving 92.3% accuracy for the quantised model while using fewer computational resources and reducing latency.
- Top Cited Papers
- 7 April 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 35 (9) , 11906-11921
Backpropagation-Based Learning Techniques for Deep Spiking Neural Networks: A Survey
With the adoption of smart systems, artificial neural networks (ANNs) have become ubiquitous. Conventional ANN implementations have high energy consumption, limiting their use in embedded and mobile applications. Spiking neural networks (SNNs) mimic the dynamics of biological neural networks by distributing information over time through binary spikes. Neuromorphic hardware has emerged to leverage the characteristics of SNNs, such as asynchronous processing and high activation sparsity. Therefore, SNNs have recently gained interest in the machine learning community as a brain-inspired alternative to ANNs for low-power applications. However, the discrete representation of the information makes the training of SNNs by backpropagation-based techniques challenging. In this survey, we review training strategies for deep SNNs targeting deep learning applications such as image processing. We start with methods based on the conversion from an ANN to an SNN and compare these with backpropagation-based techniques. We propose a new taxonomy of spiking backpropagation algorithms into three categories, namely, spatial, spatiotemporal, and single-spike approaches. In addition, we analyze different strategies to improve accuracy, latency, and sparsity, such as regularization methods, training hybridization, and tuning of the parameters specific to the SNN neuron model. We highlight the impact of input encoding, network architecture, and training strategy on the accuracy–latency tradeoff. Finally, in light of the remaining challenges for accurate and efficient SNN solutions, we emphasize the importance of joint hardware–software codevelopment.
- 7 April 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Neural Networks and Learning Systems
- Vol. 35 (9) , 11906-11921
Top Cited PapersBackpropagation-Based Learning Techniques for Deep Spiking Neural Networks: A Survey
With the adoption of smart systems, artificial neural networks (ANNs) have become ubiquitous. Conventional ANN implementations have high energy consumption, limiting their use in embedded and mobile applications. Spiking neural networks (SNNs) mimic the dynamics of biological neural networks by distributing information over time through binary spikes. Neuromorphic hardware has emerged to leverage the characteristics of SNNs, such as asynchronous processing and high activation sparsity. Therefore, SNNs have recently gained interest in the machine learning community as a brain-inspired alternative to ANNs for low-power applications. However, the discrete representation of the information makes the training of SNNs by backpropagation-based techniques challenging. In this survey, we review training strategies for deep SNNs targeting deep learning applications such as image processing. We start with methods based on the conversion from an ANN to an SNN and compare these with backpropagation-based techniques. We propose a new taxonomy of spiking backpropagation algorithms into three categories, namely, spatial, spatiotemporal, and single-spike approaches. In addition, we analyze different strategies to improve accuracy, latency, and sparsity, such as regularization methods, training hybridization, and tuning of the parameters specific to the SNN neuron model. We highlight the impact of input encoding, network architecture, and training strategy on the accuracy–latency tradeoff. Finally, in light of the remaining challenges for accurate and efficient SNN solutions, we emphasize the importance of joint hardware–software codevelopment.
- 25 October 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 7 (3) , 731-741
Are SNNs Really More Energy-Efficient Than ANNs? an In-Depth Hardware-Aware Study
Spiking Neural Networks (SNNs) hold the promise of lower energy consumption in embedded hardware due to their spike-based computations compared to traditional Artificial Neural Networks (ANNs). The relative energy efficiency of this emerging technology compared to traditional digital hardware has not been fully explored. Many studies do not consider memory accesses, which account for an important fraction of the energy consumption, use naive ANN hardware implementations, or lack generality. In this paper, we compare the relative energy efficiency of classical digital implementations of ANNs with novel event-based SNN implementations based on variants of the Integrate and Fire (IF) model. We provide a theoretical upper bound on the relative energy efficiency of ANNs, by computing the maximum possible benefit from ANN data reuse and sparsity. We also use the Eyeriss ANN accelerator as a case study. We show that the simpler IF model is more energy-efficient than the Leaky IF and temporal continuous synapse models. Moreover, SNNs with the IF model can compete with efficient ANN implementations when there is a very high spike sparsity, i.e. between 0.15 and 1.38 spikes per synapse per inference, depending on the ANN implementation. Our analysis shows that hybrid ANN-SNN architectures, leveraging a SNN event-based approach in layers with high sparsity and ANN parallel processing for the others, are a promising new path for further energy savings.
- 25 October 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Emerging Topics in Computational Intelligence
- Vol. 7 (3) , 731-741
Are SNNs Really More Energy-Efficient Than ANNs? an In-Depth Hardware-Aware Study
Spiking Neural Networks (SNNs) hold the promise of lower energy consumption in embedded hardware due to their spike-based computations compared to traditional Artificial Neural Networks (ANNs). The relative energy efficiency of this emerging technology compared to traditional digital hardware has not been fully explored. Many studies do not consider memory accesses, which account for an important fraction of the energy consumption, use naive ANN hardware implementations, or lack generality. In this paper, we compare the relative energy efficiency of classical digital implementations of ANNs with novel event-based SNN implementations based on variants of the Integrate and Fire (IF) model. We provide a theoretical upper bound on the relative energy efficiency of ANNs, by computing the maximum possible benefit from ANN data reuse and sparsity. We also use the Eyeriss ANN accelerator as a case study. We show that the simpler IF model is more energy-efficient than the Leaky IF and temporal continuous synapse models. Moreover, SNNs with the IF model can compete with efficient ANN implementations when there is a very high spike sparsity, i.e. between 0.15 and 1.38 spikes per synapse per inference, depending on the ANN implementation. Our analysis shows that hybrid ANN-SNN architectures, leveraging a SNN event-based approach in layers with high sparsity and ANN parallel processing for the others, are a promising new path for further energy savings.
- 15 September 2022
- Published by Springer Nature
- p. 359-370
Investigating Current-Based and Gating Approaches for Accurate and Energy-Efficient Spiking Recurrent Neural Networks
- 15 September 2022
- Published by Springer Nature
- p. 359-370
Investigating Current-Based and Gating Approaches for Accurate and Energy-Efficient Spiking Recurrent Neural Networks