---

Designing and constructing neural network topologies using nature-inspired algorithms - powered by high performance computer (HPC)

Description 📝

The proposed method NiaNet attempts to pick hyperparameters and AE architecture that will result in a successful encoding and decoding (minimal difference between input and output). NiaNet uses the collection of algorithms available in the library NiaPy to navigate efficiently in waste search-space.

What it can do? 👀

• Construct novel AE's architecture using nature-inspired algorithms.
• • number of solutions = topology shape * layer step * layers * act. func. * epochs * lr * optimizers
• • 3,456,000,000 = 2 * 60 * 60 * 8 * 100 * 100 * 6
• It can be utilized for any kind of dataset, which has numerical values.
• Detect anomalies in data

Installation ✅

Installing NiaNetHPC with pip3:

TODO: Publish it to PyPi

pip3 install nianet-hpc

Documentation 📘

The purpose of this paper is to get an understanding of the first version of the NiaNet approach (without HPC).

Annals of Computer Science and Information Systems, Volume 30: NiaNet: A framework for constructing Autoencoder architectures using nature-inspired algorithms

Examples

Usage examples can be found here.

Getting started 🔨

Create your own example:

• TODO Add description for making your own example.

Change dataset:

Change the dataset import function as follows:

• TODO Add description for dataloader and config file.

Specify the search space:

Set the boundaries of your search space with autoencoder.py.

The following dimensions can be modified:

• Topology shape (symmetrical, asymmetrical)
• Size of input, hidden and output layers
• Number of hidden layers
• Number of neurons in hidden layers
• Activation functions
• Number of epochs
• Learning rate
• Optimizer

You can run the NiaNet script once your setup is complete.

Running NiaNet script with Docker:

docker build --tag spartan300/nianet:dnnae .

docker run \
  --name=nianet-dnnae \
  -it \
  -v $(pwd):/app/logs \
  --gpus all spartan300/nianet:dnnae \
  python ./dnn_ae_run.py

HELP ⚠️

[email protected]

Acknowledgments 🎓

• NiaNet was developed under the supervision of doc. dr Iztok Fister ml. at University of Maribor.

• This code is a fork of NiaPy. I am grateful that the authors chose to open-source their work for future use.

License

This package is distributed under the MIT License. This license can be found online at http://www.opensource.org/licenses/MIT.

Disclaimer

This framework is provided as-is, and there are no guarantees that it fits your purposes or that it is bug-free. Use it at your own risk!