 |
SF Tech
Events Weekly
Newsletter! |
|
|
|
|
| |
|
| |
| BARUG Meetup
|
| With Rami Krispin (Data Science & Engg, Apple), Mariana Menchero (Sr Forecaster, Nixtla). |
| Thermo Fisher Scientific, 180 Oyster Point Blvd, San Francisco
|
|
Mar 14 (Fri) , 2025 @ 01:30 AM
| |
FREE |
|
|
|
|
|
|
|
|
| |
| DETAILS |
|
The meeting is still coming together, but we are excited to announce our first meetup of the year. Please mark Thursday evening, March 13 on your calendar.
Agenda:
6:30 Pizza & networking
7:05 Announcements
7:15 Simon Cawley: Biotech & Data Science at Thermo Fisher
7:30 Mariana Menchero, author of the nixtlar R package will speak about time series forecasting with TimeGPT
8:10 Rami Krispin will talk about time series forecasting with R.
--------------------------------------------------
Mariana Menchero
Time Series Forecasting with TimeGPT
Abstract:
In this talk, I'll explain how TimeGPT works & how it can be used in R via the nixtlar package. I'll introduce foundation models & explain the transformer architecture, the basis of TimeGPT. I'll then show how to use nixtlar effectively with a retail dataset & apply its key features, including fine-tuning, exogenous variables, prediction intervals, & anomaly detection. I'll conclude with a summary of the strengths & weaknesses of foundation models & how TimeGPT can be incorporated into R workflows for effective forecasting.
About Mariana
I'm a Senior Forecaster at Nixtla & the developer of nixtlar, an R package that interfaces with Nixtla's TimeGPT, the first foundation model for time series forecasting. I hold a MIT Micromasters credential in Supply Chain Management & a B.S. in Applied Mathematics. I've been a forecasting enthusiast since 2019 & I'm passionate about building forecasting tools for the R & Python communities.
--------------------------------------------------------------------------------
Rami Krispin
Analyzing Time Series at Scale with Cluster Analysis
Abstract
One of the challenges in traditional time series analysis is scalability. Most of the analysis methods were designed to handle a single time series at a time. In this talk, we will review methods for analyzing time series at scale using unsupervised learning methods. We will demonstrate how to apply cluster analysis & PCA to analyze & extract insights from multiple time series simultaneously. This talk is based on Prof. Rob J Hyndman's paper about feature-based time series analysis.
About Rami
Rami Krispin is a data science & engineering manager who mainly focuses on time series analysis, forecasting, & MLOps applications.
He is passionate about open source, working with data, machine learning, & putting stuff into production. He creates content about MLOps & recently released a course - Data Pipeline Automation with GitHub Actions Using R & Python, on LinkedIn Learning, & is the author of Hands-On Time Series Analysis with R.
|
|
|
|
|
|
|
|
