Brain encoding is the process of converting texual, visual or any sensory information to neural activity patterns.Language models can be used to predict neural activity. There has been research conducted on the potential of task-specific fine-tuned language models to predict fMRI brain activity. The aim of the project is to develop an effective ensemble of task-specific language models that can improve the accuracy and efficiency of brain encoding.
.
├── results # contains results of all models
├── README.md
├── analysis.ipynb # contains analysis of results and plots
├── encoder.ipynb # contains code for baseline model
├── ensemble.ipynb # contains code for ensemble model
├── ensemble_dyn_wt.ipynb # contains code for dynamic weights model
├── ensemble_evaluation.ipynb # contains code for evaluating ensemble model
├── ensemble_stacking.ipynb # contains code for ensemble stacking model
├── ensemble_weights_getter.py # contains code for getting weights
├── featuer_extractor.ipynb # contains code for extracting features for base models
└── job.sh # script for batch job
Sanjai P (2019112012)
Arvindh A (2019111010)
Jerrin John Thomas (2019114012)