# Realizing Neural Decoder at the Edge with Ensembled BNN

arXiv: https://arxiv.org/abs/2106.09925 · HTML: https://mlsanity.com/paper/2106.09925/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 7 of its 19 references that are in the map.
- First posted on arXiv 18 Jun 2021.

## Builds on (most cited first, 7 of 7)

- [Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation](https://mlsanity.com/paper/1308.3432/) (2013)
- [XNOR-Net: ImageNet Classification Using Binary Convolutional Neural Networks](https://mlsanity.com/paper/1603.05279/) (2016)
- [Ternary Weight Networks](https://mlsanity.com/paper/1605.04711/) (2016)
- [Turbo Autoencoder: Deep learning based channel codes for point-to-point communication channels](https://mlsanity.com/paper/1911.03038/) (2019)
- [Binary Ensemble Neural Network: More Bits per Network or More Networks per Bit?](https://mlsanity.com/paper/1806.07550/) (2018)
- [Design of Communication Systems using Deep Learning: A Variational Inference Perspective](https://mlsanity.com/paper/1904.08559/) (2019)
- [Deep-Learning Based Blind Recognition of Channel Code Parameters over Candidate Sets under AWGN and Multi-Path Fading Conditions](https://mlsanity.com/paper/2009.07774/) (2020)

## Built on by (most cited first, 2 of 2)

- [Binarized ResNet: Enabling Robust Automatic Modulation Classification at the resource-constrained Edge](https://mlsanity.com/paper/2110.14357/) (2021)
- [Rotate the ReLU to implicitly sparsify deep networks](https://mlsanity.com/paper/2206.00488/) (2022)

## Abstract (arXiv)

In this work, we propose extreme compression techniques like binarization, ternarization for Neural Decoders such as TurboAE. These methods reduce memory and computation by a factor of 64 with a performance better than the quantized (with 1-bit or 2-bits) Neural Decoders. However, because of the limited representation capability of the Binary and Ternary networks, the performance is not as good as the real-valued decoder. To fill this gap, we further propose to ensemble 4 such weak performers to deploy in the edge to achieve a performance similar to the real-valued network. These ensemble decoders give 16 and 64 times saving in memory and computation respectively and help to achieve performance similar to real-valued TurboAE.