mlsanityData to 14 Oct 2026

Low-Precision Mixed-Computation Models for Inference on Edge

arXiv 2312.02210 ↗First posted 3 Dec 2023cs.LG cs.AIcited by 4

Not on the map. Why?

Abstract (arXiv)

This paper presents a mixed-computation neural network processing approach for edge applications that incorporates low-precision (low-width) Posit and low-precision fixed point (FixP) number systems. This mixed-computation approach employs 4-bit Posit (Posit4), which has higher precision around zero, for representing weights with high sensitivity, while it uses 4-bit FixP (FixP4) for representing other weights. A heuristic for analyzing the importance and the quantization error of the weights is presented to assign the proper number system to different weights. Additionally, a gradient approximation for Posit representation is introduced to improve the quality of weight updates in the backpropagation process. Due to the high energy consumption of the fully Posit-based computations, neural network operations are carried out in FixP or Posit/FixP. An efficient hardware implementation of a MAC operation with a first Posit operand and FixP for a second operand and accumulator is presented. The efficacy of the proposed low-precision mixed-computation approach is extensively assessed on vision and language models. The results show that, on average, the accuracy of the mixed-computation is about 1.5% higher than that of FixP with a cost of 0.19% energy overhead.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Builds on 29 of its 44 references that are in the map. First posted on arXiv 3 Dec 2023.

Builds on

29 of 44 references.

References · 29

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
  3. 2023QLoRA: Efficient Finetuning of Quantized LLMscited by 3,418
  4. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,079
  5. 2021MLP-Mixer: An all-MLP Architecture for Visioncited by 1,446
  6. 2016XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networkscited by 849
  7. 2021A Survey of Quantization Methods for Efficient Neural Network Inferencecited by 815
  8. 2018PACT: Parameterized Clipping Activation for Quantized Neural Networkscited by 467
  9. 2019HAWQ: Hessian AWare Quantization of Neural Networks with Mixed-Precisioncited by 402
  10. 2018LQ-Nets: Learned Quantization for Highly Accurate and Compact Deep Neural Networkscited by 274
  11. 2022FP8 Formats for Deep Learningcited by 196
  12. 2023LoftQ: LoRA-Fine-Tuning-Aware Quantization for Large Language Modelscited by 196
17 more
  1. 2018Training Deep Neural Networks with 8-bit Floating Point Numberscited by 164
  2. 2018Scalable Methods for 8-bit Training of Neural Networkscited by 142
  3. 2020Training Binary Neural Networks with Real-to-Binary Convolutionscited by 112
  4. 2021Nonuniform-to-Uniform Quantization: Towards Accurate Quantization via Generalized Straight-Through Estimationcited by 79
  5. 2017All You Need is Beyond a Good Init: Exploring Better Solution for Training Extremely Deep Convolutional Neural Networks with Orthonormality and Modulationcited by 58
  6. 2019Any-Precision Deep Neural Networkscited by 40
  7. 2018Discovering Low-Precision Networks Close to Full-Precision Networks for Efficient Embedded Inferencecited by 33
  8. 2020FracBNN: Accurate and FPGA-Efficient Binary Neural Networks with Fractional Activationscited by 32
  9. 20228-bit Numerical Formats for Deep Neural Networkscited by 25
  10. 2018NICE: Noise Injection and Clamping Estimation for Neural Network Quantizationcited by 24
  11. 2018Deep Positron: A Deep Neural Network Using the Posit Number Systemcited by 13
  12. 2019Cheetah: Mixed Low-Precision Hardware & Software Co-Design Framework for DNNs on the Edgecited by 8
  13. 2019Performance-Efficiency Trade-off of Low-Precision Numerical Formats in Deep Neural Networkscited by 7
  14. 2019Training Deep Neural Networks Using Posit Number Systemcited by 6
  15. 2020ExPAN(N)D: Exploring Posits for Efficient Artificial Neural Network Design in FPGA-based Systemscited by 3
  16. 2021PLAM: a Posit Logarithm-Approximate Multipliercited by 3
  17. 2023Sensitivity-Aware Mixed-Precision Quantization and Width Optimization of Deep Neural Networks Through Cluster-Based Tree-Structured Parzen Estimationcited by 3

Built on by

Cited by year

Citing papers per year: 2024: 2; 2025: 2.

4 papers.

  1. 2024Flex-PE: Flexible and SIMD Multi-Precision Processing Element for AI Workloadscited by 15
  2. 2025POLARON: Precision-aware On-device Learning and Adaptive Runtime-cONfigurable AI accelerationcited by 4
  3. 2024Fever Detection with Infrared Thermography: Enhancing Accuracy through Machine Learning Techniquescited by 3
  4. 2025XR-NPE: High-Throughput Mixed-precision SIMD Neural Processing Engine for Extended Reality Perception Workloadscited by 3

Similar, not cited

Close in wording, no citation either way

  1. 2022Mixed-Precision Neural Networks: A Survey
  2. 2022Edge Inference with Fully Differentiable Quantized Mixed Precision Neural Networks
  3. 2022Channel-wise Mixed-precision Assignment for DNN Inference on Constrained Edge Nodes
  4. 2023A Practical Mixed Precision Algorithm for Post-Training Quantization
  5. 2022AMED: Automatic Mixed-Precision Quantization for Edge Devices
  6. 2023Free Bits: Latency Optimization of Mixed-Precision Quantized Neural Networks on the Edge
  7. 2024MCU-MixQ: A HW/SW Co-optimized Mixed-precision Neural Network Design Framework for MCUs
  8. 2022FxP-QNet: A Post-Training Quantizer for the Design of Mixed Low-Precision DNNs with Dynamic Fixed-Point Representation