# CliqueNet and backward connection neural network

**URL:** https://discourse.julialang.org/t/cliquenet-and-backward-connection-neural-network/95186
**Category:** Machine Learning
**Created:** [February 25, 2023, 2:46pm UTC](https://discourse.julialang.org/t/cliquenet-and-backward-connection-neural-network/95186 "2023-02-25T14:46:14Z")
**Posts on this page:** 2
**Page:** 1

<div class="post-metadata">

### Author: ![bienpierre](https://avatars.discourse-cdn.com/v4/letter/b/8c91f0/32.png) [@bienpierre](https://discourse.julialang.org/u/bienpierre)
#### Post date: [February 25, 2023, 2:46pm UTC](https://discourse.julialang.org/t/cliquenet-and-backward-connection-neural-network/95186/1 "2023-02-25T14:46:14Z")

</div>

Hello,

I would like to implement the [CliquetNet neural networks](https://openaccess.thecvf.com/content_cvpr_2018/papers/Yang_Convolutional_Neural_Networks_CVPR_2018_paper.pdf) with fully connected dense with Flux:

![image](https://global.discourse-cdn.com/julialang/original/3X/2/7/27d25d3aa59a1328e415e67b426b2d7c6761f956.png)

The network has some backward connection, I do not know the best way to implement the connection from the [`Flux.SkipConnection`](https://fluxml.ai/Flux.jl/stable/models/layers/#Flux.SkipConnection) and the [`Flux.Parallel`](https://fluxml.ai/Flux.jl/stable/models/layers/#Flux.Parallel), as I understand only the forward connection is possible.

Does anyone have an idea?

Regards

---

<div class="post-metadata">

### Author: ![ToucheSir](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/touchesir/32/14411_2.png) [@ToucheSir](https://discourse.julialang.org/u/ToucheSir)
#### Post date: [February 25, 2023, 4:15pm UTC](https://discourse.julialang.org/t/cliquenet-and-backward-connection-neural-network/95186/2 "2023-02-25T16:15:05Z")

</div>

As a first step, you could try porting the official implementation at [GitHub - iboing/CliqueNet: Convolutional Neural Networks with Alternately Updated Clique (to appear in CVPR 2018)](https://github.com/iboing/CliqueNet). If Flux’s built-in layers aren’t appropriate, don’t be afraid to create custom ones like the authors did.