Polish British Workshop, 13th PBW 2013, in conjunction with First International Student Workshop
Some random thoughts
Exploration versus Problem solving
Research projects as exploration. We setup rough boundaries, and area,
and conduct exploratory search in a certain parametr set, such as speed,
accuracy, memory usage etc. The problem is typically fixed, and often chosen
arbitrary, in a form of toy-problem or something ad hoc.
Research project as problem solving. We setup very clear boundaries and
objectives, and try to solve it to the best of the state-of-the-art
There is a settle difference between these two: in the latter case
there is a continues, or asymptotic progression towards the final solution.
In the first case, the study is of an open-ended nature, and will never
be concluded, as “more” things can be explored or evaluated. Often, the former
model precedes the former, and is an important step.
The role of mentors and senior academics
For a person, to grow and to go beyond oneself, it is important to have someone that have a different, bigger or more holistic perception of that person. So everyone seems to have a “bubble” or model of oneself, but others, based on observations, interactions, experience and expert knowledge, can have a very different model. If that external model is “bigger” or “better” than person’s own model, good things will happen. Perhaps that’s the most important role of mentors and senior community members.
9:10 Walid Allafi “Fractional order systems”
9:30 Justyna Kulinska (MSc) “Classifiers in body gesture recognition”
Depth sensor hardware based on Asus Xtion Pro, OpenNI/NiTE software
It is more “pose recognition” not “gestures”
15 standardized joints position in coordinates system
based on the center of mass
k-nearest neighbours, SVM, multi-layer perceptron
“Easy” to achieve no-error results with any of the classifiers,
with the exception of the MLP, which yielded 94.6% accuracy
SVM with Gaussian kernel provides best results.
SVM with Polynomial kernel has some issues with non-clean training samples.
9:48 Hector Quintian “Soft computing optimisation of dental…”
various soft comp. algorithms, PCA, SOM, CCA, CMLHL, MLHL, combined with GA algorithm to work as a meta-search technique.
10:03 Piotr Cal, PhD (2nd year)
Distributed Classification of Streaming Data with Concept Drift
Updating classifiers in multi-classifier system. Using DDM (drift detection method).
10:20 Mihail Gorodnichev “On mathematical modelling of totally-connected movement”
Analysis of traffic flow (of cars). Greenshields-Pipes model (linear) 1933, speed and distance between cars. The concept of safe intervals between cars. Drivers prefer to
travel with constant speed, and have preferred distance from other vehicles. Large number of moving particles. Concept of totally connected movement.
10:40 Anna Strzelecka (PhD project) “Towards enhanced sustainability of households:
simulation of utility-service provision”
Electricity, drinking water, etc. Sustainability. UK example, 2016, 80l/person/day (currently 120l). Same for carbon emissions. Devices are connected to Utilities/services and produces products/services/by-products. The housholds are modelled through devices.
The devices can be existing or “currently-under-development”.
Transformation graphs, consists of storage, Device and Service boxes, connected.
Originally part of the project “All in One”.
Invited Talk: Grzegorz Chmaj, UNLV (Las Vegas)
peer-to-peer, NoC, SoC, Reconfigurable Computing
Bitcoin, Litecoin, Namecoin, PPCoin
12:00 Paweł Ksiniewicz MSc “Storm forecasting using NNs”
Preprocessing the data, and arranging it for further processing. Clever tricks
with gathering the data, converting from image data into CSV.
12:20 Zhonghua Shen, Modelling a complex production scheduling problem
Schematic process modelling, and description.
12:30 Roberto Vega “Mobility Solutions in Multidisciplinary Research”
Mobile, and cloud computing, HTML 5, responsive, etc
HighchartsJS, Matlab integration. MIDAS project.
Two stages for people to analyze and process the data:
– Exploratory Projection Pursuit (EPP): PCA, MLHL, CMLHL,
– Modelling: NNs, SVMs.
12:47 Mariusz Hudziak “Comparing different approaches in finding path in multi-obstacle env”
Path finding, meta-heuristics,
13:02 Tomas Burianek “Diversity measure of PSO”
Particles swarm optimisation, calculating the divergence and measure of clustering.
13:15 extra talk, compression based similarity measure, used in spam filtering. Interesting.
Trip to the DonJon fort, and Srebrna Góra fortress.