Welcome to the 2024 Kaggle Playground Series! Happy New Year! This is the 1st episode of Season 4. We’re continuing the tradition of providing interesting and approachable datasets for the community to sharpen their machine learning skills — with a new challenge each month.

Your Goal: In this episode, you’ll build a model to predict whether a customer will churn — that is, close their account or stay.

But here’s the twist: alongside traditional models like XGBoost, we invite you to try out NdLinear — a new drop-in replacement for nn.Linear that improves generalization while reducing parameters, FLOPs, and memory. It works across MLPs, Transformers, RNNs, and more.

Get started here:
🔗 https://github.com/ensemble-core/NdLinear

⭐️ Star the repo and test NdLinear in your neural network implementations — we’d love to see creative approaches that challenge the usual tree-based baselines!

Good luck, and have fun experimenting!