Two Statistical Challenges in Classification of Variable Sources by James long

Описание: 20 March 2017 to 25 March 2017

VENUE: Madhava Lecture Hall, ICTS, Bengaluru

This joint program is co-sponsored by ICTS and SAMSI (as part of the SAMSI yearlong program on Astronomy; ASTRO). The primary goal of this program is to further enrich the international collaboration in the area of synoptic time domain surveys and time series analysis of gravitational wave data. The participants will focus on advancing the current understanding of these research topics by incorporating expertise of researchers from India and US who are working on identifying electromagnetic counterparts to gravitational wave sources. In essence, this program would enable US researchers to learn from the expertise of Indian researchers and also enable US researchers to exchange and share the methodologies developed by two of the five working groups of the SAMSI ASTRO program.

The program will begin with a few overview lectures designed to familiarize attendees with current trends in time domain astronomy and modern methodologies in statistics and applied mathematics. The subsequent part of the program will follow with specialized research lectures on the existing subgroups, panel discussions about collaboration possibilities between different groups with specific end-points in mind through collaborative research sessions.

Participation in this program is by invitation only and will involve about 35-40 participants only. If you are interested to participate, please contact one of the organizers.

CONTACT US:
[email protected]

PROGRAM LINK:
https://www.icts.res.in/program/TASSG...

Table of Contents (powered by https://videoken.com)
0:00:00 Time Series Analysis for Synoptic Surveys and Gravitational Wave Astronomy
0:00:04 Two Statistical Challenges in Classification of Variables Sources
0:00:17 Outline
0:00:46 Overview of Statistical Classification
0:02:05 Classifier Construction using CART
0:02:55 Building CART Tree...
0:04:37 Resulting Classifier
0:05:27 Apply Classifier to Test Data
0:06:29 Example CART Classifier Applied to OGLE
0:06:34 Optical Gravitational Lensing Experiment (OGLE)
0:07:07 OGLE Classification Example
0:07:33 First 6 Rows of Feature-Class Dataframe
0:08:23 Feature Distributions
0:08:59 CART Model Fit to Training Data
0:10:26 Confusion Matrix using Test Data
0:14:08 Challenge 1: Controlling Computational Costs in Feature Extraction
0:14:23 Features and Computation Time
0:16:59 Minimizing Features Computations
0:18:27 Multiple Iterations: RR Lyrae in Pan-STARRS
0:20:29 Formalizing this Framework
0:21:35 Controlling Feature Extraction Computational Cost
0:23:11 Challenge 2: Post Classification Inference
0:23:20 What are the distances to these objects?
0:23:43 Standard Candles
0:24:20 RR Lyrae are Variable Stars
0:24:51 Sloan Digital Sky Survey (SDSS) III - Stripe 82
0:25:44 Identifying RRL, Mapping MW halo with SDSS
0:27:42 Mapping the Galactic Halo with DES
0:28:45 Complicated, Multilevel Inference Process