Teaching
Open lectures, problem sets and notebooks.
Material from courses on Big Data and Astroinformatics, Deep Learning, Basic Electricity and Scientific Methodology, organized by semester.
2021.2
Big Data and Astroinformatics
Notebooks
- Notebook 01 — Open and plot a FITS catalog
- Notebook 02 — Pandas operations
- Notebook 03 — Image operations
- Notebook 04 — Least squares fit of the Hubble constant
- Notebook 05 — Photometric redshifts with a dense neural network
- Notebook 06 — Photometric redshifts with Random Forest
- Notebook 07 — Convolutional neural networks
- 2021.2 project brainstorm document
Slides
- Lecture 01 — Introduction
- Lecture 02 — Pandas and FITS
- Lecture 03 — Pandas and image processing
- Lecture 05 — Image processing
- Lecture 06 — Image processing and segmentation
- Lecture 07 — Segmentation
- Lecture 08 — SExtractor
- Lecture 09 — TOPCAT and AladinUnavailable
- Lecture 10 — Basic statistical conceptsUnavailable
- Lecture 11 — Maximum likelihood estimationUnavailable
- Lecture 12 — Astronomical databases and queriesUnavailable
- Lecture 13 — Maximum likelihood, least squares and logistic regressionUnavailable
- Lecture 14 — Machine Learning and neural networksUnavailable
- Lecture 15 — Machine Learning and neural networks IIUnavailable
- Lecture 16 — Decision trees and Random ForestUnavailable
Exercises
1. Create a new column from a FITS catalog and save the new catalog.
2. Estimate the limiting magnitude of the provided catalog using the relation between mag_err and S/N.
3. Create your catalog using SExtractor
sextractor file_name.fits -c config.sex -PARAMETERS_NAME sex.param -CATALOG_NAME output.fits -CATALOG_TYPE FITS_1.0 -GAIN 8116 -PIXEL_SCALE 0.06 -SEEING_FWHM 0.1 -MAG_ZEROPOINT 25.6651 -PHOT_FLUXFRAC 0.682 -DEBLEND_MINCONT 0.0015 -DEBLEND_NTHRESH 32
- Image file (replace “file_name.fits”): HST image in the f435 band
- Compare with the catalogs from Molino et al. 2017 · catalog for this image
- Create your own sex.param file, referring to the Source Extractor documentation.
4. Derive your Hubble constant from several datasets: hubble_fit_data
5. Show that an ordinary least squares fit is equivalent to the maximum likelihood of a Gaussian variable with a linear model.
6. Show that maximum likelihood for a Bernoulli variable with a linear model is equivalent to minimizing the cross-entropy between predictions and observations.
7. Show that the minimum of the cross-entropy is reached when both probabilities are equal.