Apply to the BigSkyEarth Training School 2018: GPU-based analytics and data science

Apply to the BigSkyEarth Training School 2018: GPU-based analytics and data science

BigSkyEarth is organising its 2018 Training School on “GPU-based analytics and data science” to be held at the Vicomtech Research Center, Spain, in April 3-9, 2018. Key practitioners from the astronomy, the Earth observation and the computer science domains will be participating in the School contributing the perspectives of both the academic and the industrial sector. Grants will be made available by the Action for a number of participants. The procedure for applying and for the selection of participants is described below.


The objective of COST Action TD1403 BigSkyEarth is to bridge gaps in available methodologies for scalable ‘big data’ analytics and to frame a common long term agenda between astronomy and Earth observation. More and more astronomy and Earth observation scientists and practitioners are looking into using GPUs for image processing and restoration, segmentation and labelling, de-noising, filtering, interpolation and reconstruction. Understanding how to leverage GPUs for big data analytics is crucial to many classes of machine learning and simulation algorithms being introduced into the enterprise and the academia.

Past editions of the Training School

The 1st edition of the BigSkyEarth Training School took place at the German Aerospace Center facilities in Oberpfaffenhofen, Germany, 4-9 April, 2016. It gathered around 30 researchers for hands-on training on Big Data analytics in astronomy and Earth observation with six expert instructors from the EU and the USA. The second edition took place in Preston at the University of Central Lancanshire, UK, on April 3-8 2017 and focused on the creation of effective visualisations of big data bases.

Eligibility Rules

Applicants can be PhD students or PhD holders from any country. However, only applicants from BigSkyEarth COST countries and institutions are eligible for the financial support provided by BigSkyEarth (see the list of countries and institutions below). There will be special considerations with respect to supporting COST policies on promoting gender balance, enabling Early Career Investigators (ECI) and broadening geographical inclusiveness.


The Vicomtech applied research centre is located in San Sebastian in the Basque Country (Spain). From surfing in the centre of the city to gastronomy, from spectacular landscapes to lively film and music festivals, from a rich architectural heritage to sunny beaches, the city is a hub of tourism, lifestyle and culture at the intersection of northern Spain and southern France.
Vicomtech is a research centre innovating in applied Computer Graphics and Computer Vision (Visual Computing), Data Analytics & Intelligence, Interactive Digital Media and Language Technologies solutions. Vicomtech is a founding member of, an International Network for the Cooperation in Applied Research in Computer Graphics, Multimodal-Multimedia Technologies and Visual Interactive Digital Media Technologies.
San Sebastian can easily be reached by plane (airports in Bilbao, Biarritz, Vitoria and San Sebastian are well connected by bus), by train via Paris (with high speed TGV service) or via Madrid (RENFE), and of course by bus or by car.

List of important dates

Dates to be noted include:
– Submit your application no later than: March 20, 2018
– Candidate selection: within a week after submission (until the quota is filled)
– Training School: April 3-9, 2018

Selection Criteria

Candidates will be evaluated based on their proposed research project and their motivation letter. However, there will be special considerations with respect to supporting COST policies on promoting gender balance, enabling Early Career Investigators (ECI) and broadening geographical inclusiveness.

Costs and Fees, Financial Support

There is no registration fee, but participants should anticipate costs of travel, accommodation and meals. The organizers can help with finding accommodation close to the venue. The BigSkyEarth COST Action will provide grants for selected participants who are coming from institutions in: Austria, Belgium, Bulgaria, Bosnia and Herzegovina, Croatia, Czech Republic, Denmark, Estonia, Finland, France, fYR Macedonia, Germany, Greece, Hungary, Ireland, Israel, Italy, Lithuania, Malta, Netherlands, Poland, Portugal, Romania, Serbia, Slovakia, Slovenia, Spain, United Kingdom.
Grants do not necessarily cover all expenses related to attending the Training School, but it should be enough to cover the majority of the overall costs of travel, accommodation and meal expenses.

How to Apply and Acceptance Process

Applicants should prepare:

  • A short CV
  • A list of publications
  • A short (maximum one page) description of the applicant’s research project that would benefit from this Training School
  • A short (maximum one page) description of the background knowledge of topics covered by this Training School and motivation for participating at the school.
  • If applying for a grant support from BigSkyEarth then specify the travel route and an estimate of the travel costs.

Applicants will be evaluated based on their own project proposals. Organisers will cluster, adapt and merge student proposals into manageable units. Students are expected to extensively study online material prior to attending.
The application documents should be sent to Marco Quartulli at <[email protected]>.

Challenges and general goals of the Training School

Challenges are similar in astronomy and Earth Observation, with signal processing, statistics, machine learning, and computer science as the common denominator. BigSkyEarth Training Schools aim at boosting the communication within and between disciplines and applications areas by propagating and advancing relevant common solutions developed within Big Data analysis and management research and industrial environments.
Their goal is to contribute with a valuable know-how encompassing from sensor and data modelling, features extraction and metadata, information representation, data structures, pattern recognition, statistical/machine learning, data analytics, advanced visualisation, to data mining and KDD. Besides, specific computer science topics will be addressed as particular programming techniques, data structures in large databases, cloud computing, and related topics. Their key aspect is addressing all these topics in synergy to set in a logical interdisciplinary framework building bridges between diverse areas.

Topics and general programme

The general structure of the material — with a number of currently pending confirmations — will be as follows:

  • Principles
    – Introduction to analytics software architectures
    – GPU processing basics: software
    – GPU analytics architectures
  • Tools
    – GPU basics: software programming frameworks
    – Web tools for GPU machine learning
    – Tools for running code on GPUs from high-level ‘analytical’ languages and environments, available libraries and ecosystems
    – Python SkLearn on GPUs
    – The GPU Open Analytics initiative
  • Applications
    – GPU-based deep learning
    – Object detection and vision via deep learning on GPUs
    – GPU image processing: restoration, segmentation, labeling, de-noising, filtering, interpolation and reconstruction
    – GPU processing for Earth Observation

What is BigSkyEarth?

With the current emergence of Terabyte(TB)-scale astronomical and Earth observation systems, the traditional approach to basic functions such as data searching, analytics or visualization are becoming increasingly difficult to handle. Simple database queries can result now in data subsets so large that they are incomprehensible, slow (or even impossible) to handle, and impossible to visualize with commodity visualization tools. Astronomy and remote sensing complement each other, as they are on the quest for new Big Data interpretation capabilities: both disciplines have peculiar data, typical data processing and analysis chains, and specific models to be fed with data. However, both disciplines lack the capabilities for easily accessible semantics-oriented browsing (usage of higher level descriptive expressions) in large data archives. Therefore, joint efforts to design and develop innovative Big Data tools should help users in many different fields and set new standards for many communities. This has identified several broad challenges to this line of reasoning that need multidisciplinary approach through international networking of experts and professionals. These challenges are then channelled into Action Objectives:
– Challenge A: Digital curation and data access
– Challenge B: New frontiers in visualisation
– Challenge C: Adaptation to new high performance computing (HPC) technologies
– Challenge D: New generation of scientists in the age of interdisciplinarity
For more detail see the description of the Action in the Memorandum of Understanding.