aeon

Our latest blog on the development of our forecasting module: this one clearing up our confusion about forecasting and prediction horizons https://lnkd.in/e27Uk98S by Tony Bagnall and Ali El Hadi ISMAIL FAWAZ

🎉New blog article🎉 In this blog, we'll talk about how you can use aeon base classes to accelerate the development of your time series algorithm 📈. By using aeon to handle the input checks and data conversions, we allow you to focus on innovating. ✨ https://lnkd.in/dVWZ_69y

aeon's new forecasting module is taking shape, by the end of the summer we should have a good range of deep learning, machine learning and statistical forecasters

🌐 Visualizing aeon🌐 Today, we'd like to share a fascinating visualization of aeon. This graph, created using Astro Repo, provides a unique insight into the complex web that makes up our project. The visualization maps out the various files and modules within the aeon package, showing how they are interconnected. This can be particularly useful for new contributors looking to understand the project structure, as this tool also provides file summarization to understand their role and content. 📽️ Check out Astro Repo website to see the dependencies in action and gain a deeper understanding of the aeon or any other open-source project: https://lnkd.in/e6bcen3B #Aeon #DataScience #MachineLearning #OpenSource #Python #DataVisualization #SoftwareDevelopment

🚀 Exciting News! Aeon 1.2 is Here! 🚀 This update brings a host of new features, improvements, and contributions from our amazing community: 🔹 Python Version Support: Aeon v1.2.0 now supports Python 3.10-3.13, ⚠️ dropping support for Python 3.9 ⚠️. 🔹 Anomaly Detection: We've implemented a framework for whole-series anomaly detection and refactored into two submodule for collections and single series. 🔹 Classifier Improvements: The ProximityForest and ProximityTree classifiers have been enhanced to support unequal length and multivariate series, and they are significantly faster! 🔹 Forecasting Enhancements: The forecasting module has been boosted with additional methods for direct and recursive forecasting, including new algorithms like ETS. Note that this module is still in early development, so expect more framework changes in future releases. 🔹 Similarity Search: The similarity search module has been significantly reworked to fit the aeon style framework seen in other modules. Extra refinements are planned to prepare this module to leave experimental status. 🔹 Self-Supervised Learning: 🎉 We've added an experimental sub-module for self-supervised learning in aeon 🎉, starting with the TRILITE algorithm. 👏 New Contributors: A big welcome and thank you to our new contributors: @Ahmed-Zahran02 @SomtoOnyekwelu @saadaltohamy @nMaax @AnaghDeshpande 📜 Full Changelog: Check out the full changelog for more details: https://lnkd.in/eNkxVRHE #Aeon #Release #DataScience #MachineLearning #OpenSource #Python

🎉 We welcome our second Google Summer of Code contributor for this year, Balgopal Moharana! 🎉 He is a pre-final year student at the Indian Institute of Technology (IIT), Gandhinagar, India, majoring in Computer Science and Chemical Engineering. During the summer, he will work on the Deep Learning for forecasting project to enhance aeon's forecasting capabilities. Its goal is to provide scalable, efficient, and user-friendly deep learning models tailored for time series applications, expanding aeon's functionality and usability for researchers and practitioners. #OpenSource #TimeSeriesForecasting #GoogleSummerOfCode #DataScience #aeon #MachineLearning #Innovation #NumFOCUS

🚀Exciting news for aeon time series forecasting module!📈 This year, we'll be hosting two Google Summer of Code projects on time series forecasting to improve aeon forecasting module. Let's show support to our new contributors! For our first project, we welcome Jiarong Jin, a PhD student at Shanghai Jiao Tong University from China. She will help develop our framework to work more transparently with ML algorithms, evaluating regression algorithms already implemented for forecasting problems and implementing algorithms from the literature in aeon such as SETAR-Tree. #OpenSource #TimeSeriesForecasting #GoogleSummerOfCode #DataScience #aeon #MachineLearning #Innovation #NumFOCUS

aeon developers will be at ECML/PKDD again this year in Porto, hope its as muh fun as Vilnius last year.

We're proud to announce that aeon v1.1 has been released a few days ago and is now available on PyPi ! 🎉 This version now supports the latest python 3.13 release and raised some dependencies upper bound. We also added the KASBA clustering algorithm and ROCKAD anomaly detector, as well as many documentation fixes along with some bug fixes. We've got 19 new contributors to thank for this new patch, along with our core developers team as always! Full changelog available here : https://lnkd.in/eQHdNUkC

After fruitful collaborations during #GSoC 24, we are pleased to announce that we will be participating to #GSoC 25 projects under the NumFOCUS umbrella. For this year, the following projects will be available : - Clustering: Distance-based time series clustering algorithms - Forecasting: Implementing and evaluating machine learning forecasters - Forecasting: Deep learning for forecasting - Documentation: Improving the aeon documentation interactivity and testing - Maintenance: Modernising the aeon linting and type checking workflows All the project details are available here : https://lnkd.in/ewqqRtDh . Interested in a project? Feel free to ask questions on the dedicated GitHub discussion (https://lnkd.in/eKP9iqSh) or our Slack channel.

We are pleased to announce that aeon 1.0 is out! 🥳 We are adding plenty of new estimators, especially for clustering and anomaly detection, along with a more robust test suite and more visualisation tools. We also introduce the new forecasting module that we will expend upon in the near future.❗With this major release come breaking changes, be sure to check the documentation❗ For this new major version, we are glad to welcome 9 new contributors : - Aryan Ramani (@notaryanramani) - Daniel Robert (@danielroberts20) - Ferdinand Rewicki (@ferewi) - Chuanhang Qiu (@LinGinQiu) - Francesco Spinnato (@fspinna) - Emmanuel Ferdman (@emmanuel-ferdman) - Patrick Müller (@pattplatt) - Kavya Rambhia (@kavya-r30) As always, we would like to thank all contributors involved in this release 🤗. Changelog available here https://lnkd.in/ejj8YgbM

Today aeon hit 1000 stars on github. Steady organic growth was our plan and it seems to be working. We now have 15 core developers of eight different nationalities and over 70 contributors from across the world. We have big plans for version 1.0, coming soon (fyi, the jump after four months was due to a mention by hacker news).

Our paper in JMLR Machine Learning Open Source Software track is finally available. It is a bit out of date now, the toolkit is adapting and growing, but still nice to get the paper. https://lnkd.in/ex-WD-jB

We are glad to introduce another new core developer, welcome to the team Divya Tiwari ! 🙌

We are glad to announce that we are welcoming Aadya Chinubhai as a new core developer! 😎

All of the time series sessions at ECMLPKDD 2024 were very well attended. Our tutorial had around 100 people, standing room only.

We gave a tutorial on time series machine learning at ECML 2024 in lovely Vilnius. Slides and code examples are here https://lnkd.in/eWsZEPQR

A wonderful experience at #ECMLPKDD this week 😎 Thanks again to the organizers and to fellow #AI researchers and #ML practitioners for attending our #timeseries talks and for the interesting discussions! 🤗 Here are some photos of the events with Sebastian Schmidl, Tony Bagnall, Arik Ermshaus, Patrick Schäfer, Ali El Hadi ISMAIL FAWAZ, Antoine Guillaume and Matthew Middlehurst.

#ECMLPKDD is happening next week in #Vilnius. In association with the #AALTD workshop, we'll be giving a tutorial on time series machine learning on Monday at 14:00 in room Zeta 2. We'll cover many subjects, with the intervention of Tony Bagnall, Matthew Middlehurst, Patrick Schäfer, Antoine Guillaume, Ali El Hadi ISMAIL FAWAZ, Arik Ermshaus and Sebastian Schmidl ! All information can be found on the tutorial website : https://lnkd.in/eKMHYeQ5. Once again, we would like to thank people who contributed to this tutorial but won't be able to attend : David Guijo, Germain Forestier, Thorsten Papenbrock and Phillip W. 🙏 Have a lovely weekend and see you on Monday!

Core Developers (Tony Bagnall, Matthew Middlehurst, Ali El Hadi ISMAIL FAWAZ) and friends (Germain Forestier, Geoff Webb) were at #KDD in Barcelona presenting a hands-on tutorial on time series classification and regression with #aeon. The room was packed, with over 100 in attendance! 😎