The downfall of transformers for time series forecasting and the rise of time series embedding methods. Plus advances in anomaly detection, classification, and optimal (t) interventions. — It has been quite sometime since I’ve written an update on the state of deep learning for time series. Several conferences have come and gone and the field as a whole has advanced in several different ways. …

Practical examples of using Deep Learning for Time Series Framework Flow Forecast to forecast temporal data — Flow Forecast [FF], is a state of the art deep learning for time series forecasting framework built in PyTorch. In this ongoing series we will use FF to perform forecasts (and classification) on real world time series datasets. In this first example we will use FF to perform forecasts on…

Is data scientist still the “sexiest job of the 21st century” and what that means for you — Over the course of the pandemic (and post-pandemic) we have seen a tremendous growth in the number of data engineering job postings. This has to a large extent outpaced the growth in data science job postings. …

If you’re an aspiring creator or maintainer of open source machine learning frameworks, you might find these tips helpful. — Creating a successful open source project is difficult especially in the data science/machine learning/deep learning space. A large number of open source projects never get used and are quickly abandoned. As the creator of Flow Forecast, an open source deep learning for time series forecasting framework, I’ve had my fair…

Informer is a new transformer model for long sequence time series forecasting that won best paper at the AAAI conference. — Introduction Transformers have revolutionized Natural Language Processing and powered significant improvements in areas such as neural machine translation, classification and named entity recognition. Initially transformers were slow to catch in on in the time series domain. However, over the last year and a half a number of transformer variations for time…

How to leverage Flow-Forecast, Docker, Terraform, Airflow, Kubernetes and ONNX to easily scale your deep time series models to production workloads. — Deploying machine learning models remains a sticking point for many companies and time series forecasting models are no exception. According to VentureBeat around 90% of models never make it into production. While there might be many reasons for this (e.g. models were exploratory in nature and the end goal was…

Version 0.95 of Flow Forecast includes complete support for probabilistic models, multitask forecasting, bug fixes and much more — I’m excited to announce that today we are releasing Flow Forecast version 0.95 (if you are unfamiliar with Flow Forecast please see the introductory article). Although we are still somewhat far away from version 1.0 there are a lot of cool new features now available. To get the latest version…

Lessons and tips learned from training hundreds of PyTorch time series forecasting models in many different domains — Over the past year I’ve used Flow Forecast to train hundreds of PyTorch time series forecasting models on a wide variety of datasets (river flow, COVID-19, solar/wind power, and stock prices). Often beginners come to me looking for info what they should do first. …

A deep dive into the latest literature in time series forecasting and how you can use them for your business use cases — This year at the Neural Information Processing Conference, authors published a number of new papers focusing on time series forecasting and classification. …