Data

Tools

Indexes

Applications

Experiments

Open Source Science

neural-additive-models

https://github.com/dili-lab/neural-additive-models

Science Score: 44.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ✓
    CITATION.cff file
    Found CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ○
    DOI references
  • ○
    Academic publication links
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (6.8%) to scientific vocabulary
Last synced: 6 months ago · JSON representation ·

Repository

Basic Info
  • Host: GitHub
  • Owner: DiLi-Lab
  • Language: Jupyter Notebook
  • Default Branch: main
  • Size: 3.82 MB
Statistics
  • Stars: 1
  • Watchers: 3
  • Forks: 0
  • Open Issues: 0
  • Releases: 0
Created about 3 years ago · Last pushed 11 months ago
Metadata Files
Readme Citation

README.md

Using NAMs for eye-tracking data

Get PoTeC data

bash git clone git@github.com:dili-lab/PoTeC PoTeC-data cd PoTeC-data

bash python download_data_files.py

took ~7 minutes

Training and evaluating the NAM model

The model can be trained using three different labels for the three different tasks which are: expertclslabel, allbqcorrect and alltqcorrect. An example call for one label including hyperparemeter tuning is shown below: bash python nam_train.py --hp_tuning --label "expert_cls_label" --dataset_folder PoTeC-data

Training and evaluating the baselines

The configurations for each setting are stored in a separate .json file. See below for an example call.

bash python evaluation.py --config "evaluation_configs/config_baseline_hp_tuning_2_labels_new-reader-split_label_expert_cls.json" --hp-tuning

Analyse features

In order to extract the most important features, extractfeaturecontribution.py can be run. Note that the log directory to the trained model and the label need to be provided manually at the top of the file (lines 30 and 33). The Jupyter Notebook in the folder "feature_analysis" can then be run to analyse the files and create the plots.

Owner

  • Name: Digital Linguistics Lab, Department of Computational Linguistics, University of Zurich
  • Login: DiLi-Lab
  • Kind: organization
  • Email: jaeger@cl.uzh.ch

Citation (CITATION.cff) 

ff-version: 1.2.0
message: "If you use this code, please cite it as below."
preferred-citation:
  authors:
  - family-names: "Jakobi"
    given-names: "Deborah N."
  - family-names: "Reich"
    given-names: "David R."
  - family-names: "Prasse"
    given-names: "Paul"
  - family-names: "Jäger"
    given-names: "Lena A."
  title: "Neural Additive Models Uncover Predictive Gaze Features in Reading"
  type: conference-paper
  year: 2025
  collection-type: proceedings
  collection-title: "2025 Symposium on Eye Tracking Research and Applications"
  collection-location: "Tokyo, Japan"
  collection-series: "ETRA '25"
  publisher:
    name: "Association for Computing Machinery"
    address: New York, NY, USA
  url: "https://github.com/DiLi-Lab/Neural-Additive-Models"

GitHub Events

Total
  • Push event: 1
  • Public event: 1
Last Year
  • Push event: 1
  • Public event: 1

Dependencies

neural_additive_models/setup.py pypi
  • absl-py *
  • numpy >=1.15.2
  • pandas >=0.24
  • sklearn *
  • tensorflow >=1.15
requirements.txt pypi
  • numpy >=1.23.5
  • pandas >=2.0.1
  • scikit-learn *
  • scipy *
  • setuptools *
  • spacy >=3.2.1
  • tensorflow ==2.15.0
  • torch ==2.0.1
  • tqdm *
  • typing_extensions >=4.5.0