Strock_SciAdv_2025

Code from the paper: Strock, A., Mistry, P.K., Menon, V. (2025). Personalized deep neural networks reveal mechanisms of math learning disabilities in children. Science Advances.

In this project we use artificial neural networks to model Mathematical Learning Disabilities (MLD) in children as resulting from an higher excitability level of neurons. We use the structure CORnet without any pre-training, and we train it to solve addition and subtraction visually presented.

Initial installation

Make a new environment and install packages from requirements.txt, e.g. pip install -r requirements.txt.

Clone external repositories.

mkdir common/packages
cd common/packages
git clone https://github.com/dicarlolab/CORnet
git clone https://github.com/schung039/neural_manifolds_replicaMFT

Setting up environment

source environment.sh

Generation of the addition/subtraction dataset

submit8c dataset/addsub18_handwritten.py

submit8c dataset/addsub18_font.py

submit1c dataset/behavior_addsub18.py

Model

Training the model

submit1g "model/train.py --scale 1.0"

Testing the model

submit1g "model/test.py --scale 1.0 --saveall --step $(seq -s ' ' 0 100 3800)"

Representational similarity analysis

submit8c analysis/similarity_analysis/addsub_similarity.py  --time 1-00:00:00

Behavioral analysis

submit8c analysis/behavioral_analysis/numberline_entropy.py

Manifold analysis

submit8c analysis/manifold_analysis/step_manifold.py --psteps $(seq -s ' ' 0 100 3800) --time 2-00:00:00 -p owners,normal --pmax 20 --mem 20G -c 32

Manuscript Figures

To obtain all Figures of manuscript:

bash paper/figure_all.sh

To obtain Figure X of manuscript

python paper/figureX.py

To obtain Figure SI X of manuscript

python paper/figureSX.py