Édney M. V. Freitas, Nicolas Guimarães, Rafael Maria
et al.
This work compares open-source electronic design automation tools with a commercial environment using three representative integrated circuit blocks in the IHP 130 nm open PDK: a common-mode noise filter, a finite-state machine, and a voltage-controlled oscillator. The study reports design effort and quality of results for digital logic, including area, power, and timing closure, and examines analog layout feasibility. For the finite-state machine at 50 MHz, the open-source flow reached 0.029 mm$^2$ (post-layout) and 4.37 mW (estimated) with 828 standard cells, whereas the commercial flow achieved 0.019 mm$^2$ and 2.00 mW with 497 cells, corresponding to increases of 53\% in area and 118\% in power. The common-mode noise filter totals 1.879 mm$^2$ with 1703 flip-flops at 50 MHz. The voltage-controlled oscillator occupies 0.0025 mm$^2$ and achieves a simulated maximum oscillation frequency of 2.65 GHz. The contribution is a side-by-side quantification of quality of results across digital and analog blocks in the IHP open PDK. The results indicate that open-source tools are viable for early prototyping, training, and collaboration, while commercial flows retain advantages in automation and quality of results when strict targets on power and area or precision analog layout are required.
James Andrew Gooding, Leon Bozianu, Carlos Cocha Toapaxi
et al.
Synergies between MAchine learning, Real-Time analysis and Hybrid architectures for efficient Event Processing and decision-making (SMARTHEP) is a European Training Network, training a new generation of Early Stage Researchers (ESRs) to advance real-time decision-making, driving data-collection and analysis towards synonymity. SMARTHEP brings together scientists from major LHC collaborations at the frontiers of real-time analysis (RTA) and key specialists from computer science and industry. By solving concrete problems as a community, SMARTHEP will further the adoption of RTA techniques, enabling future High Energy Physics (HEP) discoveries and generating impact in industry. ESRs will contribute to European growth, leveraging their hands-on experience in machine learning and accelerators towards commercial deliverables in fields that can profit most from RTA, e.g., transport, manufacturing, and finance. This contribution presents the training and outreach plan for the network, and is intended as an opportunity for further collaboration and feedback from the CHEP community.
The application of machine learning (ML) in high energy physics (HEP), specifically in heavy-flavor jet tagging at Large Hadron Collider (LHC) experiments, has experienced remarkable growth and innovation in the past decade. This review provides a detailed examination of current and past ML techniques in this domain. It starts by exploring various data representation methods and ML architectures, encompassing traditional ML algorithms and advanced deep learning techniques. Subsequent sections discuss specific instances of successful ML applications in jet flavor tagging in the ATLAS and CMS experiments at the LHC, ranging from basic fully-connected layers to graph neural networks employing attention mechanisms. To systematically categorize the advancements over the LHC's three runs, the paper classifies jet tagging algorithms into three generations, each characterized by specific data representation techniques and ML architectures. This classification aims to provide an overview of the chronological evolution in this field. Finally, a brief discussion about anticipated future developments and potential research directions in the field is presented.
Jennet Dickinson, Rachel Kovach-Fuentes, Lindsey Gray
et al.
The combinatorics of track seeding has long been a computational bottleneck for triggering and offline computing in High Energy Physics (HEP), and remains so for the HL-LHC. Next-generation pixel sensors will be sufficiently fine-grained to determine angular information of the charged particle passing through from pixel-cluster properties. This detector technology immediately improves the situation for offline tracking, but any major improvements in physics reach are unrealized since they are dominated by lowest-level hardware trigger acceptance. We will demonstrate track angle and hit position prediction, including errors, using a mixture density network within a single layer of silicon as well as the progress towards and status of implementing the neural network in hardware on both FPGAs and ASICs.