Twinkle twinkle little star, how I wonder what you are
summary
Uses SDSS spectra rather than vague examples and shows actual pipeline choices: image BRISK descriptors and a non‑linear SVM trained on 8,500 DA and 16,000 non‑DA spectra yields roughly mid‑80s percent accuracy, grid search takes hours, confusion matrix reveals which line profiles confuse the model, then resizing spectra images and balancing classes lets a convolutional network converge quickly and hit 99% with convolutional neural net. Data and slides link provided for replication.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 1:34Searching for white dwarfs
- 7:19Color-color diagrams for candidates
- 10:53Manual to machine learning
- 17:22Neural network classifier
- 20:57Next steps in astronomy
- 22:41Q&A
related talks
37mCounting the Stars
44mSidharth Kumar of Flipkart at Machine Learning Developers Summit 2019
41mExploring Dark Energy & Star Formation w/ ML
37mRecent Solutions Using Deep Learning Models/ The How and Why
45mStars, Planets, and Python
45mData Mining in Astronomical Survey Data of Variable Stars with Python
23mUsing a Bayesian Neural Network in the Detection of Exoplanets
15mGuiding a starship with noise! And blinking!