12-07, 19:00–19:30 (UTC), Machine Learning Track
We live in a real time world, where information and consumer preferences can change multiple times per day. This requires machine learning algorithms that can be trained and updated frequently and cost effectively. This talk will demonstrate how data scientists can use new frameworks to develop ML models that can be easily updated with new data, without requiring retraining on the full dataset.
This talk will demonstrate how data scientists can use existing tools to easily develop classic machine learning models and deep learning models that can be frequently updated with new data without requiring retraining on the full dataset. We will demonstrate the importance of frequent model updates by providing several examples. We will present the framework that enables this and provide a technical and a theoretical explanation of the data structure behind it. Finally, we will demonstrate the results that can be achieved with more frequent retraining and explain how data scientists can get started without requiring significant changes to their stack or pipelines and without incurring significant training costs.
No previous knowledge expected
Oren is Co-Founder and CEO of DataHeroes. Prior to starting DataHeroes, Oren was Co-Founder and CEO of cClearly and Co-Founder and CEO of DoubleVerify (NYSE: DV). Oren was named to the a Silicon Alley 100 list and is a winner of the Technology Pioneers Award from the World Economic Forum in Davos.
Chief Scientist at DataHeroes.