Tutorials

PyCOMPSs is a task based programming model that enables the parallel execution of sequential Python code on distributed computing platforms. The model is based on the annotation of Python code in the form of method decorators that identify the methods that become tasks at execution time. The programming model is powered by the COMPSs runtime, which at execution time detects the application tasks and the data dependencies between tasks. The runtime builds a task graph that represents the application and expresses the potential concurrency. The tutorial presents the basics of the programming model and of the runtime, based on Jupyter notebooks.

In urban scenarios, citizens travel frequently to engage in social and economic activities. Modern smart cities generate vast volumes of data that help us better understand the dynamics of mobility and design better transport systems. This tutorial introduces participants to the study of urban mobility using a data analysis workflow, with hands on experience in inspecting, cleansing, transforming, modeling and visualizing data using real world mobility datasets.

Syllabus:

• Introduction to urban mobility (1 hour, theory): urbanisation and smart cities, urban mobility and public transport, automatic data collection systems.
• Introduction to urban data analysis (2 hours): data analysis workflow and Jupyter notebooks.
• Pandas (2 hours, practice): data loading, transforming, inspecting, merging and wrangling, basic visualizations.
• Geopandas (2 hours, practice): geospatial analysis, shapes, coordinate reference systems, visualization and geoprocessing.
• Advanced visualization (1 hour, practice): interactive visualization with Bokeh and massive data visualization with Datashader.
• Machine learning toolset (1 hour, practice): clustering analysis and regression models.

Lecturers: Sebastian Bana and Renzo Massobrio, Universidad de la Republica, Uruguay. Coordinator: Sergio Nesmachnow, Universidad de la Republica, Uruguay.

Containerization is gaining traction within the HPC community with the promise of improved reliability, reproducibility and customization. This tutorial educates users on the usability of containers on HPC resources through Singularity, walking through the processes of creating, deploying and running containers on a Linux cluster. It provides a detailed background of containers, introductory hands on experience, and advanced information on running MPI based HPC applications as well as machine learning toolkits.

Instructor: Eduardo Arango, software engineer at Sylabs Inc. and PhD student at Universidad del Valle, Colombia.

COPASI is a widely used software tool for creating, simulating and analyzing models of biochemical reaction networks. It is open source and available for all major operating systems. Its features include deterministic and stochastic simulation, steady state analysis, stoichiometric analysis, optimization and parameter estimation, time scale analysis, sensitivities and metabolic control analysis, Lyapunov exponents and linear noise approximation. The tutorial provides an introduction to the modeling concepts as well as hands on experience using COPASI.

Instructor: Stefan Hoops, University of Virginia.

The use of machine learning in astronomy has become a fundamental technique for pattern recognition. This session introduces the most common machine learning algorithms, the progress achieved so far in astronomy, and a practical session where participants interact with real astronomical data using common Python libraries. Prior basic knowledge of Python is recommended.

Instructor: MSc. Felipe Meza, engineer and astrophysicist, PhD student at TEC and researcher at the Space Research Center (CINESPA, UCR).