* KASI/NIMS machine learning lunch meeting
- Deep regression forests 관련 연구 소개 및 논의
- Deep metric learning과 ranking 방법론
- CAM, Grad-CAM 소개 및 관련 논의
- signal denoising에서의 DNN 활용 사례 소개 및 활용 방안 논의
- Autoencoder를 활용한 clustering 방법 소개 및 논의
- Bayesian parameter estimation using conditional variational autoencoders for gravitational-wave astronomy 논문 소개 및 conditional variational autoencoder 활용 방법 논의
- Machine learning model hyperparameter 탐색을 위한 DeepHyper 활용 가능성 논의
- Deep Mixture Density Network을 이용한 천문학 연구 논의
* Movie files of ram pressure stripping simulations
Runs 0 and 5 in Shin & Ruszkowski 2013.
- ISM with weak turbulent motions : mp4 format. This simulation corresponds to Run 0 in the paper.
- ISM with strong turbulent motions : mp4 format. This simulation corresponds to Run 5 in the paper.
Runs PA1 and PP1 in Shin & Ruszkowski 2014.
- ISM with turbulent motions and ICM with B-field parallel to a ram pressure direction : mp4 format
- ISM with turbulent motions and ICM with B-field perpendicular to a ram pressure direction : mp4 format
* Slide and movie files of reionization simulations (Shin, Trac, and Cen, 2008, ApJ, 681, 756 - 770)
Distribution of mass-weighted ionization fraction
: ionization fraction is 1 for red, about 0.5 for green, and 0 for blue.
- 25 Mpc/h subvolume : slide (download slide files) or mpeg
- 50 Mpc/h subvolume : slide (download slide files) or mpeg
- 100 Mpc/h : slide (download slide files) or mpeg (fast but small size) mpeg (slow but large size)
Distribution of matter density and ionization fraction
: one slice of a simulation box. (shade region = neutral, red = high density)
- 100 Mpc/h : mpeg
* Examples of Splotch visualization
You can find more examples and guides for several visualization parameters in this link.
* Gravitational lensing
* Colors of asteroids
You can see this 3D color distribution in an interactive 3D plot.
* Astronomical Data Analysis Software and Systems
- Applications of multiple DBMSs and algorithms for time-domain astronomy (2019): We present results of using multiple database management systems and algorithms in processing and analyzing time-domain astronomy data. Redis is adopted as our main in-memory spatial data storage for processing time-domain alerts. We also use Google's S2 geometry algorithm and its library to generate light curve products in out time-domain observation programs. SQL-compatible DBMSs store source and object catalogs produced in our analysis, and we have tested three systems: RethinkDB, Vitess, and ClickHouse. Our project uses MongoDB as a main database engine for light curves. We share our experiences with these tools in data analysis for time-domain astronomy.
- Applications of the in-memory database Redis in processing transient event alerts (2018): We present results of using the in-memory database Redis in processing transient event alerts. The Redis works in two different ways for processing alerts. First, the publication-subscription model in the Redis allows us to adopt it as a message delivery system for multiple local alert clients. Second, we use the features of indexing and storing geolocation information in the Redis to enable low-latency matching of transient locations with custom catalogs. The current system collects event alerts by using VOEvent streams and detecting changes in web pages/feeds. We also introduce our efforts of migrating the system from the Redis message delivery environment to the NATS-based message processing configuration as well as application of Uber's H3 spatial indexing model instead of the Redis geolocation support.
- Applications of Open-source NoSQL Database Systems for Astronomical Spatial and Temporal Data (2017): We present our experiences with open-source NoSQL database systems in analyzing spatial and temporal astronomical data. We conduct experiments of using Redis in-memory NoSQL database system by modifying and exploiting its support of geohash for astronmical spatial data. Our experiment focuses on performance, cost, difficulty, and scalability of the database system. We also test OpenTSDB as a possible NoSQL database system to process astronomical time-series data. Our experiments include ingesting, indexing, and querying millions or billions of astronomical time-series measurements. We choose our KMTNet data and the public VVV (VISTA Variables in the Via Lactea) catalogs as test data. We discuss issues in using these NoSQL database systems in astronomy.
- Application of Open-source Spatio-Temporal Database Systems in Wide-Field Time-domain Astronomy (2016): We present our experiences with open-source spatio-temporal database systems for managing and analyzing big astronomical data acquired by wide-field time-domain sky surveys. Considering performance, cost, difficulty, and scalability of the database systems, we conduct comparison studies of open-source spatio-temporal databases such as GeoMesa and PostGIS that are already being used for handling big geographical data. Our experiments include ingesting, indexing, and querying millions or billions of astronomical spatio-temporal data. We choose the public VVV (VISTA Variables in the Via Lactea) catalogs of billions measurements for hundreds of millions objects as the test data. We discuss issues of how these spatio-temporal database systems can be adopted in the astronomy community.