Audrey Corbeil Therrien
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Audrey Corbeil Therrien has a total of 175 co-authors. Here are the top co-authors by number of shared publications:
- David Espes
University of Western Brittany - LabSTICC
4 shared publications - Hamza Ezzaoui Rahali
Université de Sherbrooke, Sherbrooke, Quebec, Canada
4 shared publications - Mohammad Mehdi Rahimifar
Université de Sherbrooke, Sherbrooke, Quebec, Canada
4 shared publications Berthié Gouin-Ferland
Interdisciplinary Inst Technol Innovat 3IT, Sherbrooke, PQ, Canada
3 shared publicationsCatherine Dezan
3 shared publicationsEnzo Zamaï
3 shared publications
Latest Publications
- Open Access
- 13 October 2026
- Published by American Physical Society (APS) in PRX Intelligence
- Vol. 1 (2) , 021201
Machine Learning on Heterogeneous, Edge, and Quantum Hardware for Particle Physics
The next generation of particle physics experiments will face an era of challenges in data acquisition, due to unprecedented data rates and volumes along with extreme environments and operational constraints. Harnessing these data for scientific discovery demands real-time inference and decision-making, intelligent data reduction, and efficient processing architectures beyond current capabilities. Crucial to the success of this experimental paradigm are several emerging technologies, such as artificial intelligence and machine learning, silicon microelectronics, and the advent of quantum algorithms and processing. Their intersection includes areas of research such as low-power and low-latency devices for edge computing, heterogeneous accelerator systems, reconfigurable hardware, codesign and synthesis strategies, readout for cryogenic or high-radiation environments, and analog computing. This white paper presents a community-driven vision to identify and prioritize research and development opportunities in hardware-based ML systems and corresponding physics applications, contributing toward a successful transition to the data frontier of fundamental science.
- 13 October 2026
- Published by American Physical Society (APS) in PRX Intelligence
- Vol. 1 (2) , 021201
Open AccessMachine Learning on Heterogeneous, Edge, and Quantum Hardware for Particle Physics
The next generation of particle physics experiments will face an era of challenges in data acquisition, due to unprecedented data rates and volumes along with extreme environments and operational constraints. Harnessing these data for scientific discovery demands real-time inference and decision-making, intelligent data reduction, and efficient processing architectures beyond current capabilities. Crucial to the success of this experimental paradigm are several emerging technologies, such as artificial intelligence and machine learning, silicon microelectronics, and the advent of quantum algorithms and processing. Their intersection includes areas of research such as low-power and low-latency devices for edge computing, heterogeneous accelerator systems, reconfigurable hardware, codesign and synthesis strategies, readout for cryogenic or high-radiation environments, and analog computing. This white paper presents a community-driven vision to identify and prioritize research and development opportunities in hardware-based ML systems and corresponding physics applications, contributing toward a successful transition to the data frontier of fundamental science.
- Open Access
- 29 August 2026
- Published by Elsevier in Computers & Security
- Vol. 171
Temporal-symbolic data representation for intrusion detection in Industrial Control Systems
- 29 August 2026
- Published by Elsevier in Computers & Security
- Vol. 171
Open AccessTemporal-symbolic data representation for intrusion detection in Industrial Control Systems
- Open Access
- 27 May 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Reconfigurable Technology and Systems
- Vol. 19 (2) , 1-29
wa-hls4ml: A Benchmark and Surrogate Models for hls4ml Resource and Latency Estimation
As machine learning (ML) is increasingly implemented in hardware to address real-time challenges in scientific applications, the development of advanced toolchains has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as hardware synthesis, are becoming limiting factors in the rapid iteration of designs. To mitigate these emerging constraints, multiple efforts have been undertaken to develop an ML-based surrogate model that estimates resource usage of ML accelerator architectures. We introduce wa-hls4ml, a benchmark for ML accelerator resource and latency estimation, and its corresponding initial dataset of over 680,000 fully connected and convolutional neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, and the average performance across a subset of the dataset. Additionally, we introduce GNN- and transformer-based surrogate models that predict latency and resources for ML accelerators. We present the architecture and performance of the models and find that the models generally predict latency and resources for the 75th percentile within several percent of the synthesized resources on the synthetic test dataset.
- 27 May 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Reconfigurable Technology and Systems
- Vol. 19 (2) , 1-29
Open Accesswa-hls4ml: A Benchmark and Surrogate Models for hls4ml Resource and Latency Estimation
As machine learning (ML) is increasingly implemented in hardware to address real-time challenges in scientific applications, the development of advanced toolchains has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as hardware synthesis, are becoming limiting factors in the rapid iteration of designs. To mitigate these emerging constraints, multiple efforts have been undertaken to develop an ML-based surrogate model that estimates resource usage of ML accelerator architectures. We introduce wa-hls4ml, a benchmark for ML accelerator resource and latency estimation, and its corresponding initial dataset of over 680,000 fully connected and convolutional neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, and the average performance across a subset of the dataset. Additionally, we introduce GNN- and transformer-based surrogate models that predict latency and resources for ML accelerators. We present the architecture and performance of the models and find that the models generally predict latency and resources for the 75th percentile within several percent of the synthesized resources on the synthetic test dataset.
- Open Access
- 20 March 2026
- Published by Elsevier in SSRN Electronic Journal
Temporal-Symbolic Data Representation for Intrusion Detection in Industrial Control Systems
Industrial Control Systems (ICS) increasingly integrate both information technology (IT) networks and operational technology (OT) processes, creating complex cyber-physical environments exposed to a wide range of cyber threats. In such contexts, effective intrusion detection requires models capable of capturing both communication patterns and the temporal evolution of system behavior. This paper introduces a temporal-symbolic data representation for intrusion detection in ICS environments. The proposed approach transforms sequential observations into structured 2D matrices that preserve temporal ordering while maintaining the semantic structure of system features, allowing the recent history of communications or process states to be explicitly modeled. To effectively exploit this temporal structure, we propose a mask-guided variable-window pooling mechanism that dynamically adapts the temporal aggregation according to the effective temporal horizon required for each sample. This adaptive pooling strategy preserves fine temporal resolution for recent observations while progressively summarizing older parts of the history. The resulting matrices are processed by a lightweight convolutional neural network designed to efficiently extract temporal patterns while maintaining a compact architecture. Experimental evaluations conducted on CICIDS2017, SWaT, and WADI demonstrate that the proposed method achieves strong detection performance with low computational complexity. These results show that combining structured temporal representations with adaptive temporal pooling provides an effective and scalable solution for intrusion detection in modern industrial cyber-physical systems.
- 20 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessTemporal-Symbolic Data Representation for Intrusion Detection in Industrial Control Systems
Industrial Control Systems (ICS) increasingly integrate both information technology (IT) networks and operational technology (OT) processes, creating complex cyber-physical environments exposed to a wide range of cyber threats. In such contexts, effective intrusion detection requires models capable of capturing both communication patterns and the temporal evolution of system behavior. This paper introduces a temporal-symbolic data representation for intrusion detection in ICS environments. The proposed approach transforms sequential observations into structured 2D matrices that preserve temporal ordering while maintaining the semantic structure of system features, allowing the recent history of communications or process states to be explicitly modeled. To effectively exploit this temporal structure, we propose a mask-guided variable-window pooling mechanism that dynamically adapts the temporal aggregation according to the effective temporal horizon required for each sample. This adaptive pooling strategy preserves fine temporal resolution for recent observations while progressively summarizing older parts of the history. The resulting matrices are processed by a lightweight convolutional neural network designed to efficiently extract temporal patterns while maintaining a compact architecture. Experimental evaluations conducted on CICIDS2017, SWaT, and WADI demonstrate that the proposed method achieves strong detection performance with low computational complexity. These results show that combining structured temporal representations with adaptive temporal pooling provides an effective and scalable solution for intrusion detection in modern industrial cyber-physical systems.
- Open Access
- 17 March 2026
- Published by Elsevier in SSRN Electronic Journal
Temporal-Symbolic Data Representation for Intrusion Detection in Industrial Control Systems
Industrial Control Systems (ICS) increasingly integrate both information technology (IT) networks and operational technology (OT) processes, creating complex cyber-physical environments exposed to a wide range of cyber threats. In such contexts, effective intrusion detection requires models capable of capturing both communication patterns and the temporal evolution of system behavior. This paper introduces a temporal-symbolic data representation for intrusion detection in ICS environments. The proposed approach transforms sequential observations into structured 2D matrices that preserve temporal ordering while maintaining the semantic structure of system features, allowing the recent history of communications or process states to be explicitly modeled. To effectively exploit this temporal structure, we propose a mask-guided variable-window pooling mechanism that dynamically adapts the temporal aggregation according to the effective temporal horizon required for each sample. This adaptive pooling strategy preserves fine temporal resolution for recent observations while progressively summarizing older parts of the history. The resulting matrices are processed by a lightweight convolutional neural network designed to efficiently extract temporal patterns while maintaining a compact architecture. Experimental evaluations conducted on CICIDS2017, SWaT, and WADI demonstrate that the proposed method achieves strong detection performance with low computational complexity. These results show that combining structured temporal representations with adaptive temporal pooling provides an effective and scalable solution for intrusion detection in modern industrial cyber-physical systems.
- 17 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessTemporal-Symbolic Data Representation for Intrusion Detection in Industrial Control Systems
Industrial Control Systems (ICS) increasingly integrate both information technology (IT) networks and operational technology (OT) processes, creating complex cyber-physical environments exposed to a wide range of cyber threats. In such contexts, effective intrusion detection requires models capable of capturing both communication patterns and the temporal evolution of system behavior. This paper introduces a temporal-symbolic data representation for intrusion detection in ICS environments. The proposed approach transforms sequential observations into structured 2D matrices that preserve temporal ordering while maintaining the semantic structure of system features, allowing the recent history of communications or process states to be explicitly modeled. To effectively exploit this temporal structure, we propose a mask-guided variable-window pooling mechanism that dynamically adapts the temporal aggregation according to the effective temporal horizon required for each sample. This adaptive pooling strategy preserves fine temporal resolution for recent observations while progressively summarizing older parts of the history. The resulting matrices are processed by a lightweight convolutional neural network designed to efficiently extract temporal patterns while maintaining a compact architecture. Experimental evaluations conducted on CICIDS2017, SWaT, and WADI demonstrate that the proposed method achieves strong detection performance with low computational complexity. These results show that combining structured temporal representations with adaptive temporal pooling provides an effective and scalable solution for intrusion detection in modern industrial cyber-physical systems.
- 6 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Nuclear Science
- Vol. 73 (5) , 1770-1777
Neural Network Acceleration on MPSoC Board: Integrating SLAC’s SNL and Auto-SNL
The LCLS-II Free Electron Laser (FEL) will generate X-ray pulses for beamline experiments at rates of up to 1 MHz, with detectors producing data throughputs exceeding 1 TB/s. Managing such massive data streams presents significant challenges, as transmission and storage infrastructures become prohibitively expensive. Machine learning (ML) offers a promising solution for real-time data reduction, but conventional implementations introduce excessive latency, making them unsuitable for high-speed experimental environments. To address these challenges, SLAC developed the SLAC Neural Network Library (SNL), a specialized framework designed to deploy real-time ML inference models on Field-Programmable Gate Arrays (FPGA). SNL’s key feature is the ability to dynamically update model weights without requiring FPGA resynthesis, enhancing flexibility for adaptive learning applications. To further enhance usability and accessibility, we introduce Auto-SNL, a Python extension that streamlines the process of converting Python-based neural network models into SNL-compatible high-level synthesis code. This paper presents a benchmark comparison against hls4ml, the current state-of-the-art tool, across multiple neural network architectures, fixed-point precisions, and synthesis configurations targeting a Xilinx ZCU102 FPGA. The results showed that SNL achieves competitive or superior latency in most tested architectures, while in some cases also offering FPGA resource savings. This adaptation demonstrates SNL’s versatility, opening new opportunities for researchers and academics in fields such as high-energy physics, medical imaging, robotics, and many more.
- 6 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Nuclear Science
- Vol. 73 (5) , 1770-1777
Neural Network Acceleration on MPSoC Board: Integrating SLAC’s SNL and Auto-SNL
The LCLS-II Free Electron Laser (FEL) will generate X-ray pulses for beamline experiments at rates of up to 1 MHz, with detectors producing data throughputs exceeding 1 TB/s. Managing such massive data streams presents significant challenges, as transmission and storage infrastructures become prohibitively expensive. Machine learning (ML) offers a promising solution for real-time data reduction, but conventional implementations introduce excessive latency, making them unsuitable for high-speed experimental environments. To address these challenges, SLAC developed the SLAC Neural Network Library (SNL), a specialized framework designed to deploy real-time ML inference models on Field-Programmable Gate Arrays (FPGA). SNL’s key feature is the ability to dynamically update model weights without requiring FPGA resynthesis, enhancing flexibility for adaptive learning applications. To further enhance usability and accessibility, we introduce Auto-SNL, a Python extension that streamlines the process of converting Python-based neural network models into SNL-compatible high-level synthesis code. This paper presents a benchmark comparison against hls4ml, the current state-of-the-art tool, across multiple neural network architectures, fixed-point precisions, and synthesis configurations targeting a Xilinx ZCU102 FPGA. The results showed that SNL achieves competitive or superior latency in most tested architectures, while in some cases also offering FPGA resource savings. This adaptation demonstrates SNL’s versatility, opening new opportunities for researchers and academics in fields such as high-energy physics, medical imaging, robotics, and many more.
- 16 May 2025
- Published by Springer Nature
- p. 195-209
Data Transformation for IDS: Leveraging Symbolic and Temporal Aspects
- 16 May 2025
- Published by Springer Nature
- p. 195-209
Data Transformation for IDS: Leveraging Symbolic and Temporal Aspects
- Open Access
- 27 February 2025
- Published by Association for Computing Machinery (ACM)
- p. 43
wa-hls4ml and lui-gnn: A Benchmark and GNN based Surrogate Model for hls4ml Resource and Latency Estimation
As machine learning (ML) increasingly serves as a tool for addressing real-time challenges in scientific applications, the development of advanced tooling has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as model synthesis, are now becoming limiting factors in the rapid iteration of designs. To reduce these emerging constraints, multiple efforts are being launched toward designing an ML-based surrogate model that estimates resource usage of synthesized accelerator architectures. This model would reduce the design iteration time, especially when designing within a set of given hardware constraints. This approach shows considerable potential, but as it stands, the effort is early and would benefit from coordination and standardization to assist future work as it emerges. We introduce wa-hls4ml, a benchmark for ML accelerator resource and latency estimation, and its corresponding initial dataset of more than 100,000 fully connected neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, as well as the average performance across a subset of the dataset. We measure the performance of a given predictor model through multiple metrics, including R2 score and SMAPE on regression tasks, as well as inference time to further characterize the estimator under test. Additionally, we introduce the latency/utilization inference graph neural network (lui-gnn), a surrogate model that uses a graph neural network to represent input architectures in the form of a directed graph. We present the architecture and performance of the model and find that lui-gnn generally predicts latency and utilization for the 75% quantile within several percent of the synthesized resources on the synthetic test dataset, indicating that this approach of estimating resource and latency via a surrogate models has promise and warrants further research.
- 27 February 2025
- Published by Association for Computing Machinery (ACM)
- p. 43
Open Accesswa-hls4ml and lui-gnn: A Benchmark and GNN based Surrogate Model for hls4ml Resource and Latency Estimation
As machine learning (ML) increasingly serves as a tool for addressing real-time challenges in scientific applications, the development of advanced tooling has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as model synthesis, are now becoming limiting factors in the rapid iteration of designs. To reduce these emerging constraints, multiple efforts are being launched toward designing an ML-based surrogate model that estimates resource usage of synthesized accelerator architectures. This model would reduce the design iteration time, especially when designing within a set of given hardware constraints. This approach shows considerable potential, but as it stands, the effort is early and would benefit from coordination and standardization to assist future work as it emerges. We introduce wa-hls4ml, a benchmark for ML accelerator resource and latency estimation, and its corresponding initial dataset of more than 100,000 fully connected neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, as well as the average performance across a subset of the dataset. We measure the performance of a given predictor model through multiple metrics, including R2 score and SMAPE on regression tasks, as well as inference time to further characterize the estimator under test. Additionally, we introduce the latency/utilization inference graph neural network (lui-gnn), a surrogate model that uses a graph neural network to represent input architectures in the form of a directed graph. We present the architecture and performance of the model and find that lui-gnn generally predicts latency and utilization for the 75% quantile within several percent of the synthesized resources on the synthetic test dataset, indicating that this approach of estimating resource and latency via a surrogate models has promise and warrants further research.
- Open Access
- 20 January 2025
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 6 (1) , 015009
rule4ml: an open-source tool for resource utilization and latency estimation for ML models on FPGA
Implementing machine learning (ML) models on field-programmable gate arrays (FPGAs) is becoming increasingly popular across various domains as a low-latency and low-power solution that helps manage large data rates generated by continuously improving detectors. However, developing ML models for FPGAs is time-consuming, as optimization requires synthesis to evaluate FPGA area and latency, making the process slow and repetitive. This paper introduces a novel method to predict the resource utilization and inference latency of neural networks (NNs) before their synthesis and implementation on FPGA. We leverage HLS4ML, a tool-flow that helps translate NNs into high-level synthesis (HLS) code, to synthesize a diverse dataset of NN architectures and train resource utilization and inference latency predictors. While HLS4ML requires full synthesis to obtain resource and latency insights, our method uses trained regression models for immediate pre-synthesis predictions. The prediction models estimate the usage of block RAM, digital signal processors, flip-flops, and look-Up tables, as well as the inference clock cycles. The predictors were evaluated on both synthetic and existing benchmark architectures and demonstrated high accuracy with R 2 scores ranging between 0.8 and 0.98 on the validation set and sMAPE values between 10% and 30%. Overall, our approach provides valuable preliminary insights, enabling users to quickly assess the feasibility and efficiency of NNs on FPGAs, accelerating the development and deployment processes. The open-source repository can be found at https://github.com/IMPETUS-UdeS/rule4ml, while the datasets are publicly available at https://borealisdata.ca/dataverse/rule4ml.
- 20 January 2025
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 6 (1) , 015009
Open Accessrule4ml: an open-source tool for resource utilization and latency estimation for ML models on FPGA
Implementing machine learning (ML) models on field-programmable gate arrays (FPGAs) is becoming increasingly popular across various domains as a low-latency and low-power solution that helps manage large data rates generated by continuously improving detectors. However, developing ML models for FPGAs is time-consuming, as optimization requires synthesis to evaluate FPGA area and latency, making the process slow and repetitive. This paper introduces a novel method to predict the resource utilization and inference latency of neural networks (NNs) before their synthesis and implementation on FPGA. We leverage HLS4ML, a tool-flow that helps translate NNs into high-level synthesis (HLS) code, to synthesize a diverse dataset of NN architectures and train resource utilization and inference latency predictors. While HLS4ML requires full synthesis to obtain resource and latency insights, our method uses trained regression models for immediate pre-synthesis predictions. The prediction models estimate the usage of block RAM, digital signal processors, flip-flops, and look-Up tables, as well as the inference clock cycles. The predictors were evaluated on both synthetic and existing benchmark architectures and demonstrated high accuracy with R 2 scores ranging between 0.8 and 0.98 on the validation set and sMAPE values between 10% and 30%. Overall, our approach provides valuable preliminary insights, enabling users to quickly assess the feasibility and efficiency of NNs on FPGAs, accelerating the development and deployment processes. The open-source repository can be found at https://github.com/IMPETUS-UdeS/rule4ml, while the datasets are publicly available at https://borealisdata.ca/dataverse/rule4ml.
- Open Access
- 14 November 2024
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 5 (4) , 045041
Accelerating data acquisition with FPGA-based edge machine learning: a case study with LCLS-II
New scientific experiments and instruments generate vast amounts of data that need to be transferred for storage or further processing, often overwhelming traditional systems. Edge machine learning (EdgeML) addresses this challenge by integrating machine learning (ML) algorithms with edge computing, enabling real-time data processing directly at the point of data generation. EdgeML is particularly beneficial for environments where immediate decisions are required, or where bandwidth and storage are limited. In this paper, we demonstrate a high-speed configurable ML model in a fully customizable EdgeML system using a field programmable gate array (FPGA). Our demonstration focuses on an angular array of electron spectrometers, referred to as the ‘CookieBox,’ developed for the Linac Coherent Light Source II project. The EdgeML system captures 51.2 Gbps from a 6.4 GS s−1 analog to digital converter and is designed to integrate data pre-processing and ML inside an FPGA. Our implementation achieves an inference latency of 0.2 µs for the ML model, and a total latency of 0.4 µs for the complete EdgeML system, which includes pre-processing, data transmission, digitization, and ML inference. The modular design of the system allows it to be adapted for other instrumentation applications requiring low-latency data processing.
- 14 November 2024
- Published by IOP Publishing in Machine Learning: Science and Technology
- Vol. 5 (4) , 045041
Open AccessAccelerating data acquisition with FPGA-based edge machine learning: a case study with LCLS-II
New scientific experiments and instruments generate vast amounts of data that need to be transferred for storage or further processing, often overwhelming traditional systems. Edge machine learning (EdgeML) addresses this challenge by integrating machine learning (ML) algorithms with edge computing, enabling real-time data processing directly at the point of data generation. EdgeML is particularly beneficial for environments where immediate decisions are required, or where bandwidth and storage are limited. In this paper, we demonstrate a high-speed configurable ML model in a fully customizable EdgeML system using a field programmable gate array (FPGA). Our demonstration focuses on an angular array of electron spectrometers, referred to as the ‘CookieBox,’ developed for the Linac Coherent Light Source II project. The EdgeML system captures 51.2 Gbps from a 6.4 GS s−1 analog to digital converter and is designed to integrate data pre-processing and ML inside an FPGA. Our implementation achieves an inference latency of 0.2 µs for the ML model, and a total latency of 0.4 µs for the complete EdgeML system, which includes pre-processing, data transmission, digitization, and ML inference. The modular design of the system allows it to be adapted for other instrumentation applications requiring low-latency data processing.