Exploring Dark Energy & Star Formation w/ ML
summary
HETDEX uses spectrograph spikes plus shallow imaging to find ~1M distant galaxies and measure cosmic expansion; a labeled subset of 10% labeled training set comes from ancillary telescopes, simple classifiers (decision trees, KNN, random forests, SVM) reach ~90% initial accuracy, and a custom ad hoc polynomial metric mapping precision/recall to distance error becomes the main optimization objective. Convolutional nets ingest 1D spectra to predict eight-bin star formation histories; training time and insufficient feature depth drive recommendations for deeper near‑IR imaging, of which ~70% of observing time is now awarded.
AI-written summary · about
chapters (7)
- 0:00Start
- 0:54Dark Energy Experiment
- 6:18What Is Dark Energy
- 7:33How Dark Energy Has Evolved throughout Cosmic Time
- 9:20Why Machine Learning
- 12:04The Press Shift
- 19:39Support Vector Machines
related talks
36mTwinkle twinkle little star, how I wonder what you are
39mData, Science: Investigating 1 Million Galaxies with Humans and TensorFlow
01h 14mSupernova Detection
27mAI-ML and space exploration at NASA
46mChasing the Dark Universe with Euclid and Python: Unveiling the Secrets of the Cosmos
45mStars, Planets, and Python
44mSidharth Kumar of Flipkart at Machine Learning Developers Summit 2019
23mUsing a Bayesian Neural Network in the Detection of Exoplanets