Research

Design space exploration for efficient AI accelerator / algorithm co-design

AI algorithms are increasingly coming to the fore in current research and also in industry. Even today, algorithms are limited by hardware constraints such as energy consumption or memory bandwidth. This is why it is becoming increasingly important to develop dedicated new AI accelerators and take algorithms into account. In particular, we at ITIV are working on architecture exploration with the aim of achieving an optimal trade-off between different optimization strategies, algorithm accuracy and performance.

Energy optimization of AI accelerator architectures

Energy consumption plays a very important role for many AI applications. This applies not only to battery-powered mobile devices, but increasingly also to data centers. Today, the training of algorithms in particular is responsible for high power consumption. ITIV is therefore researching new types of accelerator architectures that are specifically optimized for low energy consumption. Various strategies such as approximate computing, quantization or pruning are used for this purpose

Facilitated mapping of AI algorithms dedicated accelerators

Specialized accelerators for neural networks and AI make it possible to execute these algorithms efficiently on embedded systems. For this reason, many such accelerators have been introduced recently. Nevertheless, the important connection between the algorithms and the accelerators is becoming increasingly clear. Easy-to-use development tools are essential for the rapid roll-out of algorithms to suitable accelerators and for their co-design. ITIV is working on new mapping strategies to simplify the use of AI accelerators.

Supervised completed student projects (selection)

MA: "Evaluation of Concepts for Hardware Accelerated Neural Network Training;"
BA: "Implementation and Evaluation of Mixed Precision Systolic Arrays in the Field of Convolutional Neural Networks; Implementation and Evaluation of Mixed Precision Systolic Arrays for Convolutional Neural Networks"
BA: "Concept and Evaluation of a High Throughput Neural Network Accelerator Hardware Architecture"
MA: "Machine Learning for Material Analysis"

Publications

2024

PhD Theses

Efficient Optimization of Convolutional Neural Networks for Modern Embedded High-Performance Applications. PhD dissertation
Hotfilter, T.
2024, August 14. Karlsruher Institut für Technologie (KIT). doi:10.5445/IR/1000173259

Conference Papers

LOTTA: An FPGA-based Low-Power Temporal Convolutional Network Hardware Accelerator
Kreß, F.; Serdyuk, A.; Kobsar, D.; Hotfilter, T.; Höfer, J.; Harbaum, T.; Becker, J.
2024. 2024 IEEE 37th International System-on-Chip Conference (SOCC), Dresden, Germany, 16-19 September 2024, 126–131, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/SOCC62300.2024.10737863

Automated Deep Neural Network Inference Partitioning for Distributed Embedded Systems
Kreß, F.; El Annabi, E. M.; Hotfilter, T.; Hoefer, J.; Harbaum, T.; Becker, J.
2024. 2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI), 1st-3rd July 2024, Knoxville, 39–44, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/ISVLSI61997.2024.00019

2023

Journal Articles

EFFECT: An End-to-End Framework for Evaluating Strategies for Parallel AI Anomaly Detection
Stammler, M.; Höfer, J.; Kraus, D.; Schmidt, P.; Hotfilter, T.; Harbaum, T.; Becker, J.
2023. Procedia Computer Science, 222, 499 – 508. doi:10.1016/j.procs.2023.08.188

CNNParted: An open source framework for efficient Convolutional Neural Network inference partitioning in embedded systems
Kreß, F.; Sidorenko, V.; Schmidt, P.; Hoefer, J.; Hotfilter, T.; Walter, I.; Harbaum, T.; Becker, J.
2023. Computer Networks, 229, Article no: 109759. doi:10.1016/j.comnet.2023.109759

Conference Papers

ATLAS: An Approximate Time-Series LSTM Accelerator for Low-Power IoT Applications
Kreß, F.; Serdyuk, A.; Hiegle, M.; Waldmann, D.; Hotfilter, T.; Hoefer, J.; Hamann, T.; Barth, J.; Kämpf, P.; Harbaum, T.; Becker, J.
2023. 26th Euromicro Conference on Digital System Design (DSD 2023), 569–576, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/DSD60849.2023.00084

A Low-Stall Methodology for an Interleaved Processor State Replication
Kempf, F.; Höfer, J.; Hotfilter, T.; Becker, J.
2023. 2023 IEEE 16th International Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC), 276 – 283, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/MCSoC60832.2023.00048

Leveraging Mixed-Precision CNN Inference for Increased Robustness and Energy Efficiency
Hotfilter, T.; Hoefer, J.; Merz, P.; Kreß, F.; Kempf, F.; Harbaum, T.; Becker, J.
2023. 2023 IEEE 36th International System-on-Chip Conference (SOCC), Santa Clara, USA, 05-08 September 2023, 1–6, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/SOCC58585.2023.10256738

European Processor Initiative Demonstration of Integrated Semi-Autonomous Driving System
Hofman, D.; Brcic, M.; Kovac, M.; Hotfilter, T.; Becker, J.; Reinhardt, D.; Grigorescu, S. M.; Stevens, R.; Vo, T. T.
2023. 2023 IEEE 36th International System-on-Chip Conference (SOCC), Santa Clara, USA, 05-08 September 2023, 1–6, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/SOCC58585.2023.10257105

A Hardware-Aware Sampling Parameter Search for Efficient Probabilistic Object Detection
Hoefer, J.; Hotfilter, T.; Kreß, F.; Qiu, C.; Harbaum, T.; Becker, J.
2023. Computer Vision Systems – 14th International Conference, ICVS 2023, Vienna, Austria, September 27–29, 2023. Ed.: H. Christensen, 299–309, Springer Nature Switzerland. doi:10.1007/978-3-031-44137-0_25

A Hardware-Centric Approach to Increase and Prune Regular Activation Sparsity in CNNs
Hotfilter, T.; Höfer, J.; Kreß, F.; Kempf, F.; Kraft, L.; Harbaum, T.; Becker, J.
2023. 2023 IEEE 5th International Conference on Artificial Intelligence Circuits and Systems (AICAS), 1–5, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/AICAS57966.2023.10168566

LETSCOPE: Lifecycle Extensions Through Software-Defined Predictive Control of Power Electronics
Chu, A.; Hermann, C. M.; Silz, J.; Pfau, J.; Barón, K. M.; Anantharajaiah, N.; Schmidt, P.; Hotfilter, T.; Xie, X.; Becker, J.; Kallfass, I.; Roth-Stielow, J.; Stork, W.
2023. IEEE EUROCON 2023 - 20th International Conference on Smart Technologies, 665–670, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/EUROCON56442.2023.10199076

SiFI-AI: A Fast and Flexible RTL Fault Simulation Framework Tailored for AI Models and Accelerators
Hoefer, J.; Kempf, F.; Hotfilter, T.; Kreß, F.; Harbaum, T.; Becker, J.
2023. Proceedings of the Great Lakes Symposium on VLSI 2023, 287–292, Association for Computing Machinery (ACM). doi:10.1145/3583781.3590226

An Analytical Model of Configurable Systolic Arrays to find the Best-Fitting Accelerator for a given DNN Workload
Hotfilter, T.; Schmidt, P.; Höfer, J.; Kreß, F.; Harbaum, T.; Becker, J.
2023. DroneSE and RAPIDO: System Engineering for constrained embedded systems, 73–78, Association for Computing Machinery (ACM). doi:10.1145/3579170.3579258

Automated Search for Deep Neural Network Inference Partitioning on Embedded FPGA
Kreß, F.; Hoefer, J.; Hotfilter, T.; Walter, I.; El Annabi, E. M.; Harbaum, T.; Becker, J.
2023. Machine Learning and Principles and Practice of Knowledge Discovery in Databases. Hrsg.: I. Koprinska. Pt. 1, 557–568, Springer International Publishing. doi:10.1007/978-3-031-23618-1_37

2022

Conference Papers

Runtime Adaptive Cache Checkpointing for RISC Multi-Core Processors
Kempf, F.; Höfer, J.; Kreß, F.; Hotfilter, T.; Harbaum, T.; Becker, J.
2022. Conference Proceedings: 2022 IEEE 35th International System-on-Chip Conference (SOCC) Ed.: S. Sezer, 1–6, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/SOCC56010.2022.9908110

Data Movement Reduction for DNN Accelerators: Enabling Dynamic Quantization Through an eFPGA
Hotfilter, T.; Kreß, F.; Kempf, F.; Becker, J.; Baili, I.
2022. 2022 IEEE Computer Society Annual Symposium on VLSI (ISVLSI), Nicosia, Cyprus, 04-06 July 2022, 371–372. doi:10.1109/ISVLSI54635.2022.00082

Hardware-aware Partitioning of Convolutional Neural Network Inference for Embedded AI Applications
Kreß, F.; Hoefer, J.; Hotfilter, T.; Walter, I.; Sidorenko, V.; Harbaum, T.; Becker, J.
2022. 18th International Conference on Distributed Computing in Sensor Systems (DCOSS), 133–140, IEEEXplore. doi:10.1109/DCOSS54816.2022.00034

Hardware-aware Workload Distribution for AI-based Online Handwriting Recognition in a Sensor Pen
Kreß, F.; Serdyuk, A.; Hotfilter, T.; Höfer, J.; Harbaum, T.; Becker, J.; Hamann, T.
2022. 2022 11th Mediterranean Conference on Embedded Computing (MECO). Ed.: IEEE, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/MECO55406.2022.9797131

Towards Reconfigurable Accelerators in HPC: Designing a Multipurpose eFPGA Tile for Heterogeneous SoCs
Hotfilter, T.; Kreß, F.; Kempf, F.; Becker, J.; Haro, J. M. De; Jiménez-González, D.; Moretó, M.; Álvarez, C.; Labarta, J.; Baili, I.
2022. 2022 Design, Automation & Test in Europe Conference & Exhibition (DATE), Antwerp, Belgium, 14-23 March 2022, 628–631, Institute of Electrical and Electronics Engineers (IEEE). doi:10.23919/DATE54114.2022.9774716

Embedded Face Recognition for Personalized Services in the Assistive Robotics
Walter, I.; Ney, J.; Hotfilter, T.; Rybalkin, V.; Hoefer, J.; Wehn, N.; Becker, J.
2022. Machine Learning and Principles and Practice of Knowledge Discovery in Databases – International Workshops of ECML PKDD 2021, Virtual Event, September 13-17, 2021, Proceedings, Part I. Ed.: M. Kamp, 339–350, Springer International Publishing. doi:10.1007/978-3-030-93736-2_26

2021

Conference Papers

FLECSim-SoC: A Flexible End-to-End Co-Design Simulation Framework for System on Chips
Hotfilter, T.; Hoefer, J.; Kreß, F.; Kempf, F.; Becker, J.
2021. IEEE 34th International System-on-Chip Conference (SOCC), 14th-17th September 2021, Las Vegas, Nevada, USA, 83–88, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/SOCC52499.2021.9739212

2020

Conference Papers

QUA³CK - A Machine Learning Development Process
Stock, S. C.; Becker, J.; Grimm, D.; Hotfilter, T.; Molinar, G.; Stang, M.; Stork, W.
2020. Proceedings of Artificial Intelligence for Science, Industry and Society — PoS(AISIS2019), 026, Scuola Internazionale Superiore di Studi Avanzati (SISSA). doi:10.22323/1.372.0026

Embedded Image Processing the European Way: A new platform for the future automotive market
Hotfilter, T.; Kempf, F.; Becker, J.; Reinhardt, D.; Baili, I.
2020. 6th IEEE World Forum on Internet of Things, WF-IoT 2020, New Orleans, United States, 2 - 16 June 2020, Art.Nr. 9221396, Institute of Electrical and Electronics Engineers (IEEE). doi:10.1109/WF-IoT48130.2020.9221396

2019

Journal Articles

Evaluation of a high-throughput communication link for future automotive ADAS controllers
Yigui, L.; Youteng, S.; Schade, F.; Hotfilter, T.; Becker, J.; Yuan, Z.; Zizhou, O.; Weiming, L.
2019. Proceedings of the Institution of Mechanical Engineers / D, 233 (9), 2371–2378. doi:10.1177/0954407019851334

Presentations

The QUA³CK Machine Learning Development Process and the Laboratory for Applied Machine Learning Approaches (LAMA)
Becker, J.; Grimm, D.; Hotfilter, T.; Meier, C.; Molinar, G.; Stang, M.; Stock, S.; Stork, W.
2019, October 22. Symposium Artificial Intelligence for Science, Industry and Society (AISIS 2019), Mexico City, Mexico, October 20–December 25, 2019