Back to Programs

Data

sktime

A unified framework for ML with time series

gitgithubmachine learningpythonscikit-learnsphinx

Participation history

2 GSoC years

2024

5 projects

Official year page

Sktime integration with deep learning backends - pytorch and huggingface

Sktime provides a user-friendly interface for a range of forecasting algorithms and time-series functionalities. To enhance its capabilities further,...

Adding Support for Categorical Features

sktime in its current state does not support categorical variables in input data and this has been an often requested feature from its users. This...

Scaling backends, foundation models, PyTorch, darts and pytorch-forecasting with sktime.

With Sktime’s goal of unifying machine learning with time series, I intend to contribute to that mission with the project titled “Scaling backends,...

Scaling Backends : Polars, Hugging Face and Foundation Models

The idea behind this project is to integrate data backends like Polars and deep learning foundation models from libraries like Hugging-Face and...

Deep Learning Forecasters with Global Forecasting API

This project enables global forecasting in sktime. 1. design and implement global forecasting API in sktime. 2. add pytorch-forecasting models with...

2022

3 projects

Official year page

Expanding sktime’s Annotation Capabilities

Real world time series data is not i.i.d, and in fact understanding how the generating distribution shifts over time is part of what makes working...

Deep Learning for Time Series

The M-Competitions are a series of open competitions to evaluate forecasting methods for time-series data. The most recent competition in this...

Anomaly detection in high-dimensional data

This project will focus on broadening the suite of anomaly detection techniques within sktime. Specifically, it will focus on the incorporation of...