LogicWiSARD: Memoryless Synthesis of Weightless Neural Networks

Miranda, Igor; Arora, Aman; Susskind, Zachary; Villon, Luis A. Q.; Katopodis, Rafael F.; Dutra, Diego Leonel Cadette; Araújo, Leandro Santiago de; Lima, Priscila M. V.; França, Felipe M. G.; John, Lizy K.; Breternitz, Maurício

doi:10.1109/asap54787.2022.00014

Cited by 6 publications

(2 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For each class in the problem domain, WiSARD stores a set of RAM nodes in the form of a discriminator that is responsible for recognizing patterns of that class. Besides lower training time, models like WiSARD present a structure that can be directly implemented at a hardware level, which makes them suitable for highperformance and low-power embedded applications 37, 38 .…”

Section: Methodsmentioning

confidence: 99%

A Data-Driven Approach for Estimating Temperature Variations Based on B-mode Ultrasound Images and Changes in Backscattered Energy

Oliveira,

França,

Pereira

2023

Ultrason Imaging

View full text Add to dashboard Cite

Thermal treatments that use ultrasound devices as a tool have as a key point the temperature control to be applied in a specific region of the patient’s body. This kind of procedure requires caution because the wrong regulation can either limit the treatment or aggravate an existing injury. Therefore, determining the temperature in a region of interest in real-time is a subject of high interest. Although this is still an open problem, in the field of ultrasound analysis, the use of machine learning as a tool for both imaging and automated diagnostics are application trends. In this work, a data-driven approach is proposed to address the problem of estimating the temperature in regions of a B-mode ultrasound image as a supervised learning problem. The proposal consists in presenting a novel data modeling for the problem that includes information retrieved from conventional B-mode ultrasound images and a parametric image built based on changes in backscattered energy (CBE). Then, we compare the performance of classic models in the literature. The computational results presented that, in a simulated scenario, the proposed approach that a Gradient Boosting model would be able to estimate the temperature with a mean absolute error of around 0.5°C, which is acceptable in practical environments both in physiotherapic treatments and high intensity focused ultrasound (HIFU).

show abstract

Section: Methodsmentioning

confidence: 99%

A Data-Driven Approach for Estimating Temperature Variations Based on B-mode Ultrasound Images and Changes in Backscattered Energy

Oliveira,

França,

Pereira

2023

Ultrason Imaging

View full text Add to dashboard Cite

show abstract

“…In order to make the classification of MNIST images on a real-world FPGA feasible, Ferreira et al [12] used hash tables which alleviated the exploding memory consumption. LogicWiSARD [23] converts RAM contents into Boolean logic functions to reduce the use of arithmetic circuits.…”

Section: Table Lookup-based Computingmentioning

confidence: 99%

Table-Lookup MAC: Scalable Processing of Quantised Neural Networks in FPGA Soft Logic

Gerlinghoff,

Choong,

Goh

et al. 2024

Proceedings of the 2024 ACM/SIGDA International Symposium on Field Programmable Gate Arrays

View full text Add to dashboard Cite

Recent advancements in neural network quantisation have yielded remarkable outcomes, with three-bit networks reaching state-ofthe-art full-precision accuracy in complex tasks. These achievements present valuable opportunities for accelerating neural networks by computing in reduced precision. Implementing it on FPGAs can take advantage of bit-level reconfigurability, which is not available on conventional CPUs and GPUs. Simultaneously, the high data intensity of neural network processing has inspired computing-in-memory paradigms, including on FPGA platforms. By programming the effects of trained model weights as lookup operations in soft logic, the transfer of weight data from memory units can be avoided, alleviating the memory bottleneck. However, previous methods face poor scalability -the high logic utilisation limiting them to small networks/sub-networks of binary models with low accuracy. In this paper, we introduce Table Lookup Multiply-Accumulate (TLMAC) as a framework to compile and optimise quantised neural networks for scalable lookup-based processing. TLMAC clusters and maps unique groups of weights to lookupbased processing elements, enabling highly parallel computation while taking advantage of parameter redundancy. Further place and route algorithms are proposed to reduce LUT utilisation and routing congestion. We demonstrate that TLMAC significantly improves the scalability of previous related works. Our efficient logic mapping and high degree of reuse enables entire ImageNet-scale quantised models with full-precision accuracy to be implemented using lookup-based computing on one commercially available FPGA.

show abstract

Object modeling through weightless tracking

do Nascimento,

França

2024

Neural Comput & Applic

View full text Add to dashboard Cite

LogicWiSARD: Memoryless Synthesis of Weightless Neural Networks

Cited by 6 publications

References 16 publications

A Data-Driven Approach for Estimating Temperature Variations Based on B-mode Ultrasound Images and Changes in Backscattered Energy

A Data-Driven Approach for Estimating Temperature Variations Based on B-mode Ultrasound Images and Changes in Backscattered Energy

Table-Lookup MAC: Scalable Processing of Quantised Neural Networks in FPGA Soft Logic

Object modeling through weightless tracking

Contact Info

Product

Resources

About