AutoGluon Course
An end-to-end practical showcase of automated machine learning with AutoGluon.
Repo: autoGluon-Course
A structured, modular set of Jupyter notebooks covering AutoML with AutoGluon across four real-world paradigms: tabular data, multimodal fusion, time-series forecasting, and computer vision.
Modules
| Module | Focus | Datasets |
|---|---|---|
| Tabular Data | Automated feature engineering, ensemble selection, multi-layer stacking — classification, regression, hyperparameter tuning | Uber fare regression, hotel reservation cancellations |
| Multimodal Fusion | Fusing text, images, and tabular columns into one deep-learning backend | PetFinder adoption prediction, Kindle reviews, sentiment analysis |
| Time-Series Forecasting | Univariate/multivariate forecasting (DeepAR, Temporal Fusion Transformers), known and past covariates | Air quality, alcohol sales, restaurant visitor traffic |
| Computer Vision & Embeddings | Image classification, zero-shot similarity via OpenAI CLIP | Custom image classification datasets |
Models covered
LightGBM, XGBoost, CatBoost, Random Forests, Extra Trees, custom PyTorch networks (tabular) · DeBERTa, ELECTRA transformer backends (text/multimodal) · ResNet, EfficientNet, ConvNeXt, ViT (vision)