AutoCP

Implementation of AutoCP - Lightweight Neural Architecture Search for Cerebral Palsy Detection.

Libraries

The code is based on Python >= 3.10 and PyTorch (2.3.0). Run the following command to install all the required packages from setup.py:

pip install .

Dataset

The npy files for the CP dataset have to be stored in the following folder:

data
├── npy_files
│   └── processed
│       ├── cp_19   # put the 19-point npy files here
│       └── cp_29   # put the 29-point npy files here

Additionally, you need to specify your node in the config_node.yaml where you can guide the algorithm. This enables you to run the code on other clusters and put the right paths into place.

Run configuration

The user is provided with the following modes:

• "nas" Mode: Executes code related to neural architecture search (NAS) using AutoCP.
• "buld" Mode: Executes a retraining of a defined architecture inside the config file and executes the test.

To run a model search, check the config file and change the following parameters in the config_node:

work_dir    -> path where you want to store the logs    --> /Users/yourUser/AutoCP/logs
root_folder -> this should be something like           -->  /Users/yourUser/AutoCP/data/npy_files

The modes can either be activated or deactivated by setting the flags to True or False - refer to the different config run files for further information.

Make sure you have the chosen config file ready, which can be given as --config /path/to/your/config. Furthermore, you have to provide the path for the datasets via --node /path/to/your/node/config. The standard config files can be found in the config folder, where you can also specify your node paths.

Run NAS

To run a NAS model search, check the config file and define a Search Space and other parameters.

Afterward, execute:

python main.py -m nas

Run Build

To build an architecture from the obtained weights from the NAS, change the config retrain_... parameters.

Afterward, execute:

python main.py -m build

Results

The results reported in our study are stored in the ./logs folder. There are also predefined configs stored there, which can be used :).

The best architecture reported in the paper is argmax_1003.

Citation and Contact

If you have any questions, feel free to email [email protected].

Please cite our paper if you use this code in your research. :)

@misc{tempelLightweightNeuralArchitecture2024,
	title = {Lightweight {Neural} {Architecture} {Search} for {Cerebral} {Palsy} {Detection}},
	url = {http://arxiv.org/abs/2409.20060},
	publisher = {arXiv},
	author = {Tempel, Felix and Ihlen, Espen Alexander F. and Strümke, Inga},
	year = {2024},
}