# Stella Nera: A Differentiable Maddness-Based Hardware Accelerator for Efficient Approximate Matrix Multiplication

arXiv: https://arxiv.org/abs/2311.10207 · HTML: https://mlsanity.com/paper/2311.10207/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 35 references that are in the map.
- First posted on arXiv 16 Nov 2023.

## Builds on (most cited first, 20 of 20)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation](https://mlsanity.com/paper/1308.3432/) (2013)
- [Compressing Neural Networks with the Hashing Trick](https://mlsanity.com/paper/1504.04788/) (2015)
- [SLIDE : In Defense of Smart Algorithms over Hardware Acceleration for Large-Scale Deep Learning Systems](https://mlsanity.com/paper/1903.03129/) (2019)
- [Differentiable Product Quantization for End-to-End Embedding Compression](https://mlsanity.com/paper/1908.09756/) (2019)
- [DeepScaleTool : A Tool for the Accurate Estimation of Technology Scaling in the Deep-Submicron Era](https://mlsanity.com/paper/2102.10195/) (2021)
- [Multiplying Matrices Without Multiplying](https://mlsanity.com/paper/2106.10860/) (2021)
- [AI and ML Accelerator Survey and Trends](https://mlsanity.com/paper/2210.04055/) (2022)
- [LUT-NN: Empower Efficient Neural Network Inference with Centroid Learning and Table Lookup](https://mlsanity.com/paper/2302.03213/) (2023)
- [Scalable and Sustainable Deep Learning via Randomized Hashing](https://mlsanity.com/paper/1602.08194/) (2016)
- [CUTIE: Beyond PetaOp/s/W Ternary DNN Inference Acceleration with Better-than-Binary Energy Efficiency](https://mlsanity.com/paper/2011.01713/) (2020)
- [ChewBaccaNN: A Flexible 223 TOPS/W BNN Accelerator](https://mlsanity.com/paper/2005.07137/) (2020)
- [Flex-SFU: Accelerating DNN Activation Functions by Non-Uniform Piecewise Approximation](https://mlsanity.com/paper/2305.04546/) (2023)
- [Bolt: Accelerated Data Mining with Fast Vector Compression](https://mlsanity.com/paper/1706.10283/) (2017)
- [FPnew: An Open-Source Multi-Format Floating-Point Unit Architecture for Energy-Proportional Transprecision Computing](https://mlsanity.com/paper/2007.01530/) (2020)
- [RedMulE: A Compact FP16 Matrix-Multiplication Accelerator for Adaptive Deep Learning on RISC-V-Based Ultra-Low-Power SoCs](https://mlsanity.com/paper/2204.11192/) (2022)
- [PECAN: A Product-Quantized Content Addressable Memory Network](https://mlsanity.com/paper/2208.13571/) (2022)
- [Look-ups are not (yet) all you need for deep learning inference](https://mlsanity.com/paper/2207.05808/) (2022)
- [Yet Another Representation of Binary Decision Trees: A Mathematical Demonstration](https://mlsanity.com/paper/2101.07077/) (2021)
- [Design Automation for Binarized Neural Networks: A Quantum Leap Opportunity?](https://mlsanity.com/paper/1712.01743/) (2017)

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

- [SAIL: SRAM-Accelerated LLM Inference System with Lookup-Table-based GEMV](https://mlsanity.com/paper/2509.25853/) (2025)

## Abstract (arXiv)

Artificial intelligence has surged in recent years, with advancements in machine learning rapidly impacting nearly every area of life. However, the growing complexity of these models has far outpaced advancements in available hardware accelerators, leading to significant computational and energy demands, primarily due to matrix multiplications, which dominate the compute workload. Maddness (i.e., Multiply-ADDitioN-lESS) presents a hash-based version of product quantization, which renders matrix multiplications into lookups and additions, eliminating the need for multipliers entirely. We present Stella Nera, the first Maddness-based accelerator achieving an energy efficiency of 161 TOp/s/W@0.55V, 25x better than conventional MatMul accelerators due to its small components and reduced computational complexity. We further enhance Maddness with a differentiable approximation, allowing for gradient-based fine-tuning and achieving an end-to-end performance of 92.5% Top-1 accuracy on CIFAR-10.