# Distribution-Aware Hadamard Quantization for Hardware-Efficient Implicit Neural Representations

arXiv: https://arxiv.org/abs/2508.13478 · HTML: https://mlsanity.com/paper/2508.13478/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 23 references that are in the map.
- First posted on arXiv 19 Aug 2025.

## Builds on (most cited first, 17 of 17)

- [DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation](https://mlsanity.com/paper/1901.05103/) (2019)
- [Implicit Neural Representations with Periodic Activation Functions](https://mlsanity.com/paper/2006.09661/) (2020)
- [Occupancy Networks: Learning 3D Reconstruction in Function Space](https://mlsanity.com/paper/1812.03828/) (2018)
- [Implicit Geometric Regularization for Learning Shapes](https://mlsanity.com/paper/2002.10099/) (2020)
- [Vector Neurons: A General Framework for SO(3)-Equivariant Networks](https://mlsanity.com/paper/2104.12229/) (2021)
- [WIRE: Wavelet Implicit Neural Representations](https://mlsanity.com/paper/2301.05187/) (2023)
- [Texture Fields: Learning Texture Representations in Function Space](https://mlsanity.com/paper/1905.07259/) (2019)
- [FINER: Flexible spectral-bias tuning in Implicit NEural Representation by Variable-periodic Activation Functions](https://mlsanity.com/paper/2312.02434/) (2023)
- [ACORN: Adaptive Coordinate Networks for Neural Scene Representation](https://mlsanity.com/paper/2105.02788/) (2021)
- [Compressive Neural Representations of Volumetric Scalar Fields](https://mlsanity.com/paper/2104.04523/) (2021)
- [FAST: DNN Training Under Variable Precision Block Floating Point with Stochastic Rounding](https://mlsanity.com/paper/2110.15456/) (2021)
- [DINER: Disorder-Invariant Implicit Neural Representation](https://mlsanity.com/paper/2211.07871/) (2022)
- [On Quantizing Implicit Neural Representations](https://mlsanity.com/paper/2209.01019/) (2022)
- [Nonparametric Teaching of Implicit Neural Representations](https://mlsanity.com/paper/2405.10531/) (2024)
- [ASMR: Activation-sharing Multi-resolution Coordinate Networks For Efficient Inference](https://mlsanity.com/paper/2405.12398/) (2024)
- [Exploring the Potential of Low-bit Training of Convolutional Neural Networks](https://mlsanity.com/paper/2006.02804/) (2020)
- [Toward INT4 Fixed-Point Training via Exploring Quantization Error for Gradients](https://mlsanity.com/paper/2407.12637/) (2024)

## Built on by (most cited first, 1 of 1)

- [PatchINR: Patch-Based Implicit Neural Representations for Efficient and Scalable Inference](https://mlsanity.com/paper/2606.25534/) (2026)

## Abstract (arXiv)

Implicit Neural Representations (INRs) encode discrete signals using Multi-Layer Perceptrons (MLPs) with complex activation functions. While INRs achieve superior performance, they depend on full-precision number representation for accurate computation, resulting in significant hardware overhead. Previous INR quantization approaches have primarily focused on weight quantization, offering only limited hardware savings due to the lack of activation quantization. To fully exploit the hardware benefits of quantization, we propose DHQ, a novel distribution-aware Hadamard quantization scheme that targets both weights and activations in INRs. Our analysis shows that the weights in the first and last layers have distributions distinct from those in the intermediate layers, while the activations in the last layer differ significantly from those in the preceding layers. Instead of customizing quantizers individually, we utilize the Hadamard transformation to standardize these diverse distributions into a unified bell-shaped form, supported by both empirical evidence and theoretical analysis, before applying a standard quantizer. To demonstrate the practical advantages of our approach, we present an FPGA implementation of DHQ that highlights its hardware efficiency. Experiments on diverse image reconstruction tasks show that DHQ outperforms previous quantization methods, reducing latency by 32.7%, energy consumption by 40.1%, and resource utilization by up to 98.3% compared to full-precision counterparts.