# LegoNet: Memory Footprint Reduction Through Block Weight Clustering

arXiv: https://arxiv.org/abs/2603.06606 · HTML: https://mlsanity.com/paper/2603.06606/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 23 references that are in the map.
- First posted on arXiv 18 Feb 2026.

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

- [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://mlsanity.com/paper/1905.11946/) (2019)
- [MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications](https://mlsanity.com/paper/1704.04861/) (2017)
- [Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding](https://mlsanity.com/paper/1510.00149/) (2015)
- [Knowledge Distillation: A Survey](https://mlsanity.com/paper/2006.05525/) (2020)
- [Learned Step Size Quantization](https://mlsanity.com/paper/1902.08153/) (2019)
- [Pruning neural networks without any data by iteratively conserving synaptic flow](https://mlsanity.com/paper/2006.05467/) (2020)
- [Do Wide and Deep Networks Learn the Same Things? Uncovering How Neural Network Representations Vary with Width and Depth](https://mlsanity.com/paper/2010.15327/) (2020)
- [Ternary Weight Networks](https://mlsanity.com/paper/1605.04711/) (2016)
- [Compressing Deep Convolutional Networks using Vector Quantization](https://mlsanity.com/paper/1412.6115/) (2014)
- [Low-bit Quantization of Neural Networks for Efficient Inference](https://mlsanity.com/paper/1902.06822/) (2019)
- [And the Bit Goes Down: Revisiting the Quantization of Neural Networks](https://mlsanity.com/paper/1907.05686/) (2019)
- [CHIP: CHannel Independence-based Pruning for Compact Neural Networks](https://mlsanity.com/paper/2110.13981/) (2021)
- [AutoQ: Automated Kernel-Wise Neural Network Quantization](https://mlsanity.com/paper/1902.05690/) (2019)
- [An Overview of Neural Network Compression](https://mlsanity.com/paper/2006.03669/) (2020)
- [Learning Filter Basis for Convolutional Neural Network Compression](https://mlsanity.com/paper/1908.08932/) (2019)
- [Towards Compact CNNs via Collaborative Compression](https://mlsanity.com/paper/2105.11228/) (2021)
- [Deep k-Means: Re-Training and Parameter Sharing with Harder Cluster Assignments for Compressing Deep Convolutions](https://mlsanity.com/paper/1806.09228/) (2018)
- [Permute, Quantize, and Fine-tune: Efficient Compression of Neural Networks](https://mlsanity.com/paper/2010.15703/) (2020)
- [FleXOR: Trainable Fractional Quantization](https://mlsanity.com/paper/2009.04126/) (2020)
- [Structural Pruning in Deep Neural Networks: A Small-World Approach](https://mlsanity.com/paper/1911.04453/) (2019)

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

- [A Multimodal Framework for Aligning Human Linguistic Descriptions with Visual Perceptual Data](https://mlsanity.com/paper/2602.19562/) (2026)

## Abstract (arXiv)

As the need for neural network-based applications to become more accurate and powerful grows, so too does their size and memory footprint. With embedded devices, whose cache and RAM are limited, this growth hinders their ability to leverage state-of-the-art neural network architectures. In this work, we propose LegoNet, a compression technique that constructs blocks of weights of the entire model regardless of layer type and clusters these induced blocks. Using blocks instead of individual values to cluster the weights, we were able to compress ResNet-50 trained for Cifar-10 and ImageNet with only 32 4x4 blocks, compressing the memory footprint by over a factor of 64x without having to remove any weights or changing the architecture and no loss to accuracy, nor retraining or any data, and show how to find an arrangement of 16 4x4 blocks that gives a compression ratio of 128x with less than 3% accuracy loss. This was all achieved with no need for (re)training or fine-tuning.