Outline of the School

The primary aim of the XXVI Winter School is to introduce young researchers to the use of Bayesian techniques for the analysis of observations.

Our goal as scientists is to decipher the secrets of nature. We do this by developing physical models based on our current knowledge of the Universe. These models are used to make testable predictions, which are then compared with all kinds of observations. Due to the inherent specificities, a large fraction of the work in Astronomy and Astrophysics occurs just at this border, where one has to carry out the comparison between the observations and the models. Observations are always limited in number and accuracy, and the comparison has to be done very carefully. Fortunately, a set of mathematical tools, currently known as “Bayesian inference” have been developed in the last centuries to facilitate and formalize this work at the interface between data and models.

Our view of the Universe is imperfect, because information from observations is always incomplete and uncertain. The presence of noise and limitations in our measurement apparatus, together with our limited knowledge of the phenomena under study produce that comparing models and observations is not an easy task. Uncertainty, degeneracies and ambiguities plague our results and one has to be fully aware of them to give reliable results that help advance the science. The most important consequence of the previous points is that inference has to be carried out probabilistically: we will not be able to be sure of a conclusion with 100% certainty. Additionally, we are not completely blind during inference: we always have some a-priori knowledge about the parameters of the model that we need to make explicit in our line of reasoning. All the previous points are natural part of Bayesian inference and its extreme power has been and is being put in operation in the latest years. Not only on physical sciences the success of Bayesian methods has been enormous, but also in other fields like politics, philosophy, legal reasoning, genomics, medical treatment decision making, etc.

Cosmology, specially the analysis of the cosmic microwave background (CMB), was one of the first fields of Astrophysics in which Bayesian inference was used. The relatively low quality of the first observations (COBE) forced researchers to use Bayesian inference to get information about the parameters of the standard cosmological model with reliable error bars and properly taking into account their correlation. From this moment on, the data has been vastly improved (e.g., with the latest results of Planck and WMAP) but Bayesian methods are heavily used to discard regions of the space of parameters putting in common many different measurements (CMB, baryon acoustic oscillations, supernovae, …) and help better constraint our model for the Universe.

The field of detection and characterization of extrasolar planets, that is now in effervescence due to missions like Kepler, has witnessed one of the most striking uses of Bayesian inference. It is usually difficult to infer the properties of the planet/s orbiting around stars just from the data we have, that is usually contaminated with lots of systematics. Researchers have applied Bayesian ideas to infer the most probable number of planets and their properties around a given star just from photometric data, fully taking into account and marginalizing all contaminating effects.

Future earth-based and space-based telescopes will provide huge amounts of data (not only in night-time astronomy but also in solar physics). The sizes make it impossible to analyze the data in a case-by-case basis. Consequently, robust statistical Bayesian analyses are planned in many missions, many of them based on hierarchical probabilistic models. Future PhD students will probably witness the extinction of catalogs (for instance, galactic and extragalactic sources) as we currently know them. They will be transformed into probabilistic models that will easily facilitate the comparison with models of the large-scale structures.

In a completely different subject, Bayesian methods are at the core of the most advanced image reconstruction methods. Among the current and potential applications, some examples include the generation of super-resolved images from interferometric data (e.g., ALMA) or avoiding the atmospheric perturbation of the atmosphere for astronomical imaging (e.g., going beyond lucky imaging techniques using sparsity constraints).

The Bayesian paradigm is becoming more and more visible every year, as demonstrated by the limited sample cases described above and the number of papers in Astrophysics per year. Furthermore, 2013 was the 250th anniversary of the original paper of Thomas Bayes presented at a meeting of the Royal Society of London by Richard Price. 

This outline names only a few astrophysical contexts where Bayesian methods are of relevance, and they will be covered in this Winter School. To that aim, the Winter School lectures will be given by seven well known and experienced scientists who are actively working on a variety of forefront research projects, and who have played a key role in major advances over the recent years. The format of the School, which includes laboratory sessions, will also encourage direct interaction between the participating students and lecturers. The school is primarily intended for doctoral students and recent postdocs in any field of research in Astronomy. Participants of the Winter School will have the opportunity to display their current work by presenting a short contribution.

The Winter School will take place at San Cristóbal de La Laguna (Tenerife, Canary Islands, Spain), from November 3th to 14th, 2014. The lectures will be delivered in English and will be published by Cambridge University Press in a dedicated monograph. Speakers will present their topics in a series of four lectures. Visits to the IAC's Headquarter in La Laguna and the Teide Observatory in Tenerife will be scheduled as part of the activities.

Those interested in attending the Winter School should register following the instructions provided in the Registration page. Once registered, please login to the personal area of the website and complete the required information. In addition we request a letter of reference from a thesis advisor, or Head of Department, by email to winter@iac.es. These should reach the IAC before June 15th. Selected candidates will be informed by July 1st.