eSSENCE/DATA Theme Hack Day
From: 2018-05-30 14:00 To: 2018-05-31 15:30
Place:Pufendorf Institute for Advanced Studies, Biskopsgatan 3, Lund
eSSENCE will host a Hack Day together with the DATA Theme at the Pufendorf Institute of Advanced Studies
Hack days are informal gatherings where people come together to work in groups on particular problems introduced by some of the delegates. The goal is to find viable ways to solve problems of particular interest to those attending and also to go as far as possible in the implementation. The result can be anything from a conceptual approach to the design of an algorithm or workflow to a finished piece of software. In other words, the hack days are a way to connect people with different skill sets to join forces under the heading of e-Science.
Participants will have an opportunity to submit problems before the event. The hack days will begin with brief presentations of the selected problems by those who proposed them. The participants then organise themselves into groups and the rest of the first day is spent in these groups, discussing solutions.
The next day starts with a general gathering where each group presents a status report with the possibility to get input from the rest of the audience. The day continues with the work in groups and ends with another general gathering where the groups report on the progress they have made and what is left to do.
Coffee and tea will be served both days and a light lunch will be served on the second day, free of charge, but the participants will need to organise their own travel and accommodation as well as any other meals at their own expense.
To register, please send an email to mbd [at] astro [dot] lu [dot] se with the subject “Hack Day 30 May Registration” no later than noon on Thursday 24 May. Note that space is limited, so early registration is encouraged to avoid disappointment.
1) Where to put one’s data
Michael Witt, Purdue University
Michael Witt is building a data repository recommendation tool that leverages structured information in a registry of data repositories known as re3data. A researcher has produced a dataset and needs guidance to identify a data repository that is appropriate for their domain and type of data and that will accept it.
The tool’s design document describes a web-based tool that will ask the researcher four questions and based on their answers return results from re3data that are relevant to them.
HACK CHALLENGE: design an algorithm for determining and ranking relevant results.
2) Tracking problem
Kalle Åström, Mathematics
There are many algorithms for detecting and locating objects in one image (one frame) of a video. Typically there are sometimes missing detections, double detections and false detections. How should one connect the correct detections over time, while throwing away false detections and being robust to missing detections? Input varies. For each detection there is a time stamp and there could be information about positions, sizes, classes of objects, possibly other features.
3) Competition for soil water and nutrients between plant individuals root system
David Wårlind, Stefan Olin, Adrian Gustafson, Physical Geography & Ecosystem Science
The individual-based vegetation model LPJ-GUESS simulates competition between plant individuals for light, space and soil resources. Currently soil water is divided equally between individuals according to their demand and root distribution. For nutrient uptake, there is competition, but rather simplistic as soil nutrients gather in a single soil pool. With coming improvements to the representation of the soil column with several soil layers with water and nutrients, these simplistic approaches would need to be updated with a new approach that is more appropriate. The challenge is how to make this as simplistic and flexible as possible, but at the same time computationally efficient and able to capture the complexity of the system.
Individual plants (grasses and different trees) have roots that have different depth and distribution through the soil column. Grasses have a shallower root system and trees have roots that goes deeper. Their uptake strength will depend on the amount of roots they have in each soil layer. Available water and nutrients in different soil layers will change during the course of time, and the plants will try to adopt their root distribution within their limits. The challenging task is to set up a system for competition that accounts for the different ways that plant individuals can meet their demand for water and nutrients. It could be that grasses with shallower roots could meet their water or nutrient demand in the top layers if trees only concentrate on deeper soil layers, but this is not realistic. It could instead be that trees take a portion of their demand from the top layers, resulting in grasses being limited despite resources being available at deeper depths for trees. We need a new framework that will be able to determine the individual plant uptake of both water and nutrients under competition in our new soil column framework with distinct water and nutrient content in each soil layer.
4) Identifying groups in output from N-body simulations
Ross Church, Astronomy & Theoretical Physics
Stars are formed in groups or clusters. We perform N-body simulations following the evolution of such systems. Close encounters between stars in clusters transfer energy, leading to the ejection of some of the stars. Binaries and triples (three stars bound together) may also form through dynamical encounters. We wish to have software which looks at the output of N-body runs of a small number of stars (10-30) and automatically identifies the bound, sub-systems of stars and the multiple stars (binaries, triples) within each sub-system.