The forecasting field is always looking into the future, and especially now, there’s even more going on. So, how to keep up? ?? We’re starting a newsletter! In NextUp: News from Nixtla you'll find what’s new with our releases, guides and how-to’s, and what’s new in forecasting - approaches, implementations, examples and some quick tips and tricks. ?? We hope this helps keep you up to date, learn some things, and have some fun too. ??? Subscribe to the newsletter to get the latest news, and let us know what you'd like to see in our next edition!
关于我们
We are a startup that builds forecasting and time series software for data scientist and developers.
- 网站
-
https://www.nixtla.io/
Nixtla的外部链接
- 所属行业
- 软件开发
- 规模
- 2-10 人
- 总部
- New York
- 类型
- 私人持股
- 创立
- 2021
地点
-
主要
US,New York
Nixtla员工
-
Ludwig Pierre Schulze
Managing Partner @ Alumni Ventures
-
Tracy K. Teal, PhD
Passionate about open source, developing and supporting leaders and teams, and creating system-oriented solutions. Open Source Program Director at…
-
Gautam Krishnamurthi
GreatPoint Ventures
-
Marco Peixeiro
Senior AI Scientist | NLP | Time Series | machine learning & deep learning | Python (TensorFlow, Pytorch, Flask) | MySQL | JavaScript (React)
动态
-
?? for TimeGPT + R!
Last week, I was excited and honored to speak at LatinR, sharing about the ?????????????? ?? ?????????????? that makes ????????????????, Nixtla's state-of-the-art foundation model for time series forecasting, easily accessible to the R community. I love examples, so I put together a demo for retail forecasting with data and code. Full reproducible example here: ?? Let’s imagine you’re the manager of a Corporación Favorita store, a grocery chain in Ecuador. You’re responsible for forecasting the sales of all products at your store for the next two weeks. You have daily sales data for each product: for some, you have years’ worth of data, while for others, only a few weeks. To make things interesting, you need forecasts for all products by tomorrow to start ordering inventory. As an R programmer, you decide to put nixtlar to the test. For comparison, you will also forecast sales using the ARIMA and ETS models from fable, along with Prophet. As a baseline, you will use a seasonal naive forecast. See how they all did! Spoiler alert, TimeGPT outperformed or had comparable accuracy to the widely used R models in a fraction of the time? ?? Here’s the code and the slides from the talk: https://lnkd.in/gzs7TDGv ? Get started with nixtlar here: https://lnkd.in/gdMjaSET Thank you so much to the LatinR organizers and participants for an amazing meeting! ?? ?? #forecasting #timeseries #LatinR2024
-
?? ?? New release of NeuralForecast! ?? ?? NeuralForecast?offers a large collection of neural forecasting models focused on their usability, and robustness. The models range from classic networks like?MLP,?RNNs to novel proven contributions like?NBEATS,?NHITS,?TFT?and other architectures. The v.1.7.6 release from Marco Peixeiro includes: ?? Support for providing DataLoader arguments to optimize GPU usage. ? Set activation function in GRN of TFT.? ?? Updates to conformal predictions. ?? Added ability to load models saved using versions before 1.7. ?? New tutorial on cross-validation. ?? Thanks to community members Jasmine Rienecker, Tyler Nisonoff and JQGoh for their contributions! ?? Get started with neuralforecast: https://lnkd.in/gh3zhHpj See full details in the release notes: https://lnkd.in/grtiAcjS ?? Happy forecasting! #Forecasting #TimeSeries #MachineLearning #Python #DataScience
?? NeuralForecast
nixtlaverse.nixtla.io
-
?????Nixtla named as one of Fast Company’s Next Big Things in Tech ???? We're proud to share that we’ve been named to Fast Company's fourth annual Next Big Things in Tech list, honoring emerging technology that has a profound impact for industries—from education and sustainability to robotics and artificial intelligence. We are honored to be chosen alongside other game-changing organizations for the #FCTechAwards.?https://lnkd.in/e3t5h9Fa Time-series forecasting is a powerful method that leverages time-stamped data to predict future events and remove uncertainty from business conditions. Any organization that wants to perform forecasting or anomaly detection can benefit from a time series model, and we aim to do for time series what LLMs have done for language - making them accessible to anyone. Our nixtlaverse open source ecosystem has more than 18M downloads and is used by forecasters around the world. Our foundation model TimeGPT has made it easier for anyone to forecast using data they already have in Excel or other systems – such as sales, revenue and inventory data - without having to hire a team of engineers to develop custom models. With TimeGPT, we make time-series forecasting available to anyone with just three lines of code. “It’s a tremendous honor to have our time-series forecasting software recognized by Fast Company as groundbreaking technology that will have a profound impact on businesses,” says Max Mergenthaler Canseco Co-Founder and CEO of Nixtla. “Nixtla’s users come from big companies and small, government agencies, nonprofits and academic institutions, and they can create insights and use data to make decisions in ways that were not possible before. Our team has worked hard to bring this technology to market and make time-series forecasting available to everyone.” “The Next Big Things in Tech provides a fascinating glimpse at near- and long-term technological breakthroughs across a variety of sectors,” says Brendan Vaughan, editor-in-chief of Fast Company. “Spanning everything from semiconductors to agricultural gene editing, the companies featured in this year’s list are tackling some of the world’s most pressing and vexing problems.” ?? We’re grateful for this honor and the amazing community who does the work every day that continues to make forecasting the next big thing. ? Give forecasting a try or continue your time-series journey with our open source ecosystem and TimeGPT. https://lnkd.in/gvbF4QRW ?? Happy forecasting! #Forecasting #TimeSeries #MachineLearning #Python #DataScience
-
Thanks for the mention Zeel Thumar. We are curious to know what Clem Delangue ?? thinks about that statement.
This the the last thing you will ever need to know for time-series forecasting [ no PhD required! ??] ?? Step into the future of data-driven decisions with Nixtla TimeGPT! TimeGPT: Access generative AI for predictions without an ML team – upload & predict in seconds! ?? StatsForecast: Lightning-fast statistical forecasting. ?? MLForecast: Scalable machine learning models for any time series. Trained on 100 billion+ data points for unparalleled accuracy. Democratize time series analysis with effortless usability. Fast and scalable solutions for both startups and enterprises. And most of it is open source, so enjoy!!
-
Nixtla转发了
LinkedIn Top Voice | Data Scientist | NTT Data | MS (LJMU'24) | M.Tech (IIEST'20) | B.Tech (NITA'18)
Excited to share that my article "Can machine learning outperform Statistical models for Time Series Forecasting" published in Towards AI has been making waves on Medium , with about 15k reads and 1.1k claps! This shows the rising interest in ML models for time series forecasting and innovative libraries like Nixtla's MLforecast. If you're a data scientist looking to explore this field, you might want to check out the article and Nixtla's library! Link to the article: https://lnkd.in/gWssgniQ #MachineLearning #TimeSeriesForecasting #DataScience #Nixtla #MLforecast #ai #newintech
-
?? ?? New release of Hierarchical Forecast! ?? ?? HierachicalForecast?offers different reconciliation methods that render coherent forecasts across hierarchies. New release v0.4.3 adds the following features: ?? Sparse middle-out reconciliation via?MiddleOutSparse: Efficiently perform middle-out reconciliation in large-scale settings? ?? Support for exogenous variables in utils.aggregate: Aggregate exogenous covariates too when aggregating time series? ? Efficient Schafer-Strimmer for MinT: Perform MinT-shrink reconciliation blazingly fast? ?? Improves residuals-based reconciliation stability and faster ma.cov: Improved stability and NaN-handling, allowing more problems to be solvable with MinT-methods ?? Shout out to our community members?Christopher Titchen?and?Kurai Maingi?for their contributions! Note that as of v0.4.3, hierarchicalforecast no longer officially supports Python 3.8, which is EOL. Our priorities for the next month are (i) adding Polars support and (ii) adding temporal hierarchical reconciliation methods, but open to suggestions otherwise! ?Questions or suggestions for new features? Let us know as a comment or?file an issue on Github.? Get started with Hierarchical Forecast?: https://lnkd.in/gJVT4hBS ?? Happy forecasting! #Forecasting #TimeSeries #MachineLearning #Python #DataScience
Hierarchical Forecast ??
nixtlaverse.nixtla.io
-
?? ?? Presenting our R package for working with TimeGPT ?? ?? We’re excited to present?nixtlar v0.6.1, our R package for working with the?TimeGPT?API. As the R equivalent to?nixtla, the Python SDK for TimeGPT, nixtlar was already available on CRAN and has now been updated to?version 0.6.1. This new version incorporates the new TimeGPT API-v2 and extends support to tibbles. TimeGPT is the first foundation model for time series forecasting. With TimeGPT?you can quickly and easily generate accurate forecasts or conduct anomaly detection in just a few lines of code. nixtlar is designed to be user-friendly, allowing anyone within the R community to leverage generative AI for time series. Capabilities of nixtlar include: ?? Zero-shot inference:?Generate accurate forecasts with no prior model training. ?? Fine-tuning:?Enhance your predictions by fine-tuning the model to your specific dataset. ?? Uncertainty quantification:?Capture the full distribution of your predictions, either via quantile forecasts or prediction intervals. ?? R ecosystem integration:?Works with your favorite R data structures—data frames, tibbles, or tsibbles. nixtlar embodies Nixtla’s core philosophy to democratize access to state-of-the-art predictive insights, making advanced forecasting tools accessible to all. ?? Thank you for all the ideas and feedback from the R community that have contributed to nixtlar! ?? Links in the comments for getting started with nixtlar. And of course nixtlar already has its own hex sticker! Let us know in the comments if you'd like one! #Forecasting #TimeSeries #MachineLearning #AI #RStats #DataScience
-
? 5 ways companies can use time series forecasting Much of the data your company has is already time-stamped. It’s probably sitting in Excel spreadsheets, brimming with potential. Recent advances in AI, such as foundation models, make it possible for smaller companies to access time series forecasting to make predictions, reduce uncertainty, and gain business advantages. In this InfoWorld article Cristian Challu, Ph.D. shares five ways your business can use your data for time series forecasting. ?? * Turn Excel spreadsheets into future knowledge about your business. * Use information about external factors to make predictions. * Measure uncertainty with probabilistic forecasts. * Evaluate different future scenarios. * Detect anomalies that will affect your predictions or identify unusual patterns. https://lnkd.in/gQkpucPy #timeseries #forecasting #machinelearning #ai #python
5 ways companies can use time series forecasting
infoworld.com
-
?? Love all the great resources on time series forecasting. Congratulations on the new edition Manu Joseph & Jeffrey Tackes!
I'm excited to share that the second edition of my book on time series forecasting is now available for pre-order! This is a joint effort between Jeffrey Tackes and me. This edition is packed with new insights and practical tools, including: - A shift to the highly popular StatsForecast library by Nixtla for easier and faster forecasting. - Deep LEarning examples using NeuralForecast by Nixtla, bringing you the latest in forecasting technology. - In-depth exploration of cutting-edge deep learning models like TS Mixer. - A comprehensive and huge new chapter on probabilistic forecasting, covering techniques like probability density functions, quantile regressions. Mc Dropout, and the big one in the lot - Conformal Prediction with a focus on time series forecasting. For those who already own the first edition, this new edition brings a wealth of updated knowledge, making it a must-have companion to your current copy. With cutting-edge libraries like StatsForecast and NeuralForecast, plus new deep learning models and a comprehensive chapter on probabilistic forecasting, this edition ensures you stay at the forefront of the field. It's the perfect opportunity to deepen your understanding and refine your forecasting skills. Don't miss out on this updated resource designed to enhance your forecasting skills. Pre-order your copy today! Link in comments