ABCDFusion

This is the repository to build ML model on multiview data including DTI connectivity data, resting state data, Corticalthichness and other data from ABCD dataset to predict "becoming CHR after 3 years". This README.md assume you're using an Unix/macOS system.

Repo Structure

.
├── abcdfusion              # Source files, define the major modules to build machine learning pipeline with pytorch
├── data                    # Data folder, store raw data (.xlsx) and data after preprocessing (.csv)
├── data_engineer           # Source files, inspect raw data and conduct preprocessing
├── examples                # Source files, perform group-wise (sites) cross validation on single/multiple modalities 
├── notebooks               # Source files, notebooks to get preliminary results with sklearn 
├── configs.yaml            # Config file, define the model architectures and data loading path
├── LICENSE
└── README.md

Install Environment

Install a list of requirements specified in a Requirements File.

foo@bar:~$ python3 -m pip install -r requirements.txt

Force to install all packages even if they are already up-to-date.

foo@bar:~$ pip install -r requirements.txt --force-reinstall

Install the requirments and ignore the installed packages/files

foo@bar:~$ pip install -r requirements.txt --ignore-installed

Install from Source Code

Ensure pip, setuptools, and wheel are up to date

foo@bar:~$ python3 -m pip install --upgrade pip setuptools wheel

Create a virtual environment

foo@bar:~$ python3 -m venv abcd_env
foo@bar:~$ source abcd_env/bin/activate

Install normally from src.

foo@bar:~$ python3 -m pip install .

Install in development mode, i.e. in such a way that the project appears to be installed, but yet is still editable from the src tree.

foo@bar:~$ python3 -m pip install -e .

To deactivate the virtual environment after using it.

foo@bar:~$ deactivate abcd_env

Use Tensorboard to Visualize the Training Procedures and Recorded Statistics

Installation

foo@bar:~$ pip install tensorboard
foo@bar:~$ tensorboard --logdir=runs

Start from a local computer if the stats are saved in folder_to_exps

foo@bar:~$ tensorboard --logdir=folder_to_exps

Start from a remote computer if the stats are saved in folder_to_exps

foo-remote@bar-remote:~$ tensorboard --logdir folder_to_exps --port 6006
foo@bar:~$ ssh -N -f -L localhost:16006:localhost:6006 foo-remote@bar-remote

geshijoker / abcdfusion Goto Github PK

abcdfusion's Introduction