TeamCerebro

Julio Esparza, Angelica Parra, Maria Royo Cano, Zihao Wang
Dept. of Bioengineering, Imperial College London, London, UK

This repository is part of BMI class at Imperial College of London. The aim of its functions are to evaluate different methods to decode a monkey's hand position from neural activity. The repository is subdivided into 3 main folders:

1. Select_parameters: This folder contains the functions used to tune the different parameters the algorithms contain as to select the most appropiate ones ad hoc.
   a. CNB_kfold.m: evaluate CNB accuracy through 10-kfold.
   b. SVM_del_kfold.m: evaluate SVM (with delete cells as the preprocessing step) accuracy through 10-kfold.
   c. SVM_gauss_kfold.m: evaluate SVM (with gaussian as the preprocessing step) accuracy through 10-kfold.
   d. parameters_performance.m: general function that calls for the different classifiers parameter testing. This function may take several hours to run due to the iterative nature of the parameter optimization.
   e. fillformat.m & plotFill1.m: functions used for plotting. Developed by Sara Mederos and Julio Esparza at Cajal Institute (Spanish Research Council), Madrid, Spain.
   
2. Method_evaluation: This folder contains the functions used to test the different algorithm combinations (direction classifier + trajectory predictor).
   a. evaluate_methods.m: evaluation across different combinations of direction (classification) and trajectory (regression) prediction methods. This requires the positionEstimatorTraining.m & positionEstimator.m functions located in "Train&Predict" folder.
   
3. Train&Predict: This folder contains the functions used to train and predict the data for each algorithm (i.e. this folder contains the functions in the competition format).
   a. positionEstimatorTraining.m: train different models for neural decoding.
   b. positionEstimator.m: predicts reaching direction and movement trajectory using the trained model.
   c. testFunction_for_students_MTb.m: splits dataset, call training and testing functions.
   

Last Updated: 17/05/2021 (DD/MM/YYYY)

Happy training!