README.md

funGCN

Functional Graph Convolutional Networks and Feature Selection

The repo contains the code to perform two models:

1. Functional adaptive Feature Selection with elastic-net penalty (ffs) using a Dual Augmented Lagrangian algorithm for the following models:

   • Function-on-Function (FF)
   • Function-on-Scalar (FS)
   • Scalar-on-Function (SF)
   • Logit (with functional features and categorical response)

2. Functional Graph Convolutional Networks (fgcn). The model estimates a knowledge graph and implements a GCN to work with multi-modal (longitudinal, scalar, and categorical) data and cuncurrently perform different tasks (regression, classification, forecasting).

---

The implemented methodology is described in the following papers:

• A Highly-Efficient Group Elastic Net Algorithm with an Application to Function-On-Scalar Regression
• FAStEN: an efficient adaptive method for feature selection and estimation in high-dimensional functional regressions
• A new computationally efficient algorithm to solve Feature Selection for Functional Data Classification in high-dimensional spaces
• Functional Graph Convolutional Networks: A unified multi-task and multi-modal learning framework to facilitate health and social-care insights

---

FILES DESCRIPTION

expes --------------------------------------------------------------------------------------------------------------

    ffs 
    
        expes/ffs/sim_FF.py:
          file to run feature selection on synthetic data for the Function-On-Function model 

        expes/ffs/sim_FS.py:
          file to run feature selection on synthetic data for the Function-On-Scalar model
        
        expes/ffs/sim_logit.py:
          file to run feature selection on synthetic data for the Logit model

        expes/ffs/sim_SF.py:
          file to run feature selection on synthetic data for the Scalar-On-Function model
    
    fgcn 
        
        expes/fgcn/sim_fgcn.py:
          file to run fungcn on synthetic data


fungcn -------------------------------------------------------------------------------------------------------------
    
    fungcn/ffs:
      folder containing the files to implement ffs. More info is in the folder readme

    fasten/fgcn:
      folder containing the files to implement fmmgcn. More info is in the folder readme

REQUIREMENTS

1. Create a python3.11 environment: conda create -n my_env python=3.11
2. clone funGCN (main branch)
3. install funGCN and required packages by running pip install -e . at the root of the repository, i.e. where the setup.py file is.
4. Lunch the desired experiments, e.g., python expes/ffs/sim_FF.py

For Apple M processors' users:

• To install numpy with the apple library vecLib (which is optimized for Apple processors) run:

  pip install cython pybind11
  pip install numpy cython