High Quality Down-Sampling for Deterministic Approaches to Stochastic Computing

Najafi, M. Hassan; Lilja, David J.

doi:10.1109/tetc.2017.2789243

Cited by 28 publications

(5 citation statements)

References 21 publications

(59 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Deterministic sequences that enable error-free multiplication have been proposed (Fig. 3) [25], [26] and investigated from various aspects [27]- [30]. Error-free multiplication is possible because each bit in a sequence is "seen" exactly once by every bit of the other sequence during stochastic multiplication (e.g.…”

Section: Related Workmentioning

confidence: 99%

A Deterministic Approach to Approximate Computing on Stochastic Computing Hardware, with Reduced Sequence Lengths for Convolutional Neural Networks

Hekmatshoar¹

2023

Preprint

View full text Add to dashboard Cite

<p>In this work, a deterministic sequence suitable for approximate computing on stochastic computing hardware is proposed and its effectiveness in achieving high accuracies with relatively short sequence lengths is studied for convolutional neural networks. It is shown that in the range of interest for neural network computations, multiplication errors can be lower than quantization errors with this approach. The sequence lengths required for achieving accuracies within ~0.5% of the floating-point baseline are of the order of 16 and 32 for CIFAR10 classification with VGG16 and ResNet20 networks, respectively, when all convolutions and matrix multiplications are performed using the proposed sequence. For ImageNet classification, the sequence lengths required for accuracies within ~1% of the floating-point baseline are of the order of 32 for MobileNetV1 and ResNet50 networks. This work suggests that stochastic computing hardware and approaches may be feasible for approximate neural network computations with higher accuracies, lower latencies and/or larger networks than previously reported. </p>

show abstract

Section: Related Workmentioning

confidence: 99%

A Deterministic Approach to Approximate Computing on Stochastic Computing Hardware, with Reduced Sequence Lengths for Convolutional Neural Networks

Hekmatshoar¹

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…To perform normalization and increase the throughput, an efficient converging normalization unit had been used [21]. A deterministic stochastic system with down sampling has been introduced to overcome fluctuation and correlation to develop an accurate system [22]. An isolation-based method has been developed for handling the accuracy issue in an SC system.…”

Section: Related Work On Stochastic Computing Elementsmentioning

confidence: 99%

Accuracy Analysis on Design of Stochastic Computing in Arithmetic Components and Combinational Circuit

Ashok,

Bala Tripura Sundari

2023

Computation

View full text Add to dashboard Cite

Stochastic circuits are used in applications that require low area and power consumption. The computing performed using these circuits is referred to as Stochastic computing (SC). The arithmetic operations in this computing can be realized using minimum logic circuits. The SC system allows a tradeoff of computational accuracy and area; thereby, the challenge in SC is improving the accuracy. The accuracy depends on the SC system’s stochastic number generator (SNG) part. SNGs provide the appropriate stochastic input required for stochastic computation. Hence we explore the accuracy in SC for various arithmetic operations performed using stochastic computing with the help of logic circuits. The contributions in this paper are; first, we have performed stochastic computing for arithmetic components using two different SNGs. The SNGs considered are Linear Feed-back Shift Register (LFSR) -based traditional stochastic number generators and S-box-based stochastic number generators. Second, the arithmetic components are implemented in a combinational circuit for algebraic expression in the stochastic domain using two different SNGs. Third, computational analysis for stochastic arithmetic components and the stochastic algebraic equation has been conducted. Finally, accuracy analysis and measurement are performed between LFSR-based computation and S-box-based computation. The novel aspect of this work is the use of S-box-based SNG in the development of stochastic computing in arithmetic components. Also, the implementation of stochastic computing in the combinational circuit using the developed basic arithmetic components, and exploration of accuracy with respect to stochastic number generators used is presented.

show abstract

“…The particular digital implementations require only a few logic gates. Furthermore, stochastic computations can be performed with no accuracy loss when employing converters based on linear feedback shift registers [17]. Computational latency can also be efficiently reduced for certain digital signal processing systems when employing noise-shaping converters [18].…”

Section: Stochastic Computing Basicsmentioning

confidence: 99%

A Novel Stochastic Polar Architecture for All-Digital Transmission

Andriakopoulos

Papachatzopoulos

Paliouras

2021

2021 IEEE International Symposium on Circuits and Systems (ISCAS)

View full text Add to dashboard Cite

A novel architecture of an all-digital transmitter is proposed in this paper, introducing stochastic computation to a single-bit polar topology. The proposed architecture exploits the benefits of stochastic computing to simplify the design of an alldigital transmitter and minimize hardware complexity. Complexity reduction is achieved by implementing a digital oscillator as a cosine function in the stochastic domain instead of resorting to LUT-based or CORDIC-based implementations. An evaluation of the introduced all-digital polar architecture shows that the power spectral density of the transmitted passband signal is similar to that of a conventional transmitter. Synthesis of the proposed stochastic-enabled transmitter at a 28-nm FDSOI technology node reveals its minimal complexity requirements and 87.13% area reduction compared to conventional implementations.

show abstract

High Quality Down-Sampling for Deterministic Approaches to Stochastic Computing

Cited by 28 publications

References 21 publications

A Deterministic Approach to Approximate Computing on Stochastic Computing Hardware, with Reduced Sequence Lengths for Convolutional Neural Networks

A Deterministic Approach to Approximate Computing on Stochastic Computing Hardware, with Reduced Sequence Lengths for Convolutional Neural Networks

Accuracy Analysis on Design of Stochastic Computing in Arithmetic Components and Combinational Circuit

A Novel Stochastic Polar Architecture for All-Digital Transmission

Contact Info

Product

Resources

About