Today Technical University in Vienna (TU Wien) organised its biennial symposium dedicated to computer graphics: Visual Computing Trends 2011. There were several invited talks, and few hundred participants. The event was a great opportunity for, sometimes exciting and visionary, meeting and discussions concerning visual computing, and all aspects of computer graphics.
The first presenter was prof. Bernt Schiele. He is doing lots of interesting work in the area of semantic scene analysis, segmentation and object detection. However, he is also interested in activity detection from visual cues.
One of the presenters was Pat Hanrahan. Pat has given a very engaging and passionate talk about computer visualisation. He has provided examples from Simon’s work on systems of thought, and the way human cognition partitions and manages information (visual and other). This has particular relevance to visualisation and data analysis. He talked about Simon’s work on isomorphs. The term often used was: visual analytics. The use of visual tools to help with data analysis and understanding. Data analytics is paramount to every area of computing. One of the problems highlighted is the fact that large amounts of highly complicated data, analysed “passively” by a single person may, and often does, provide a single, often skewed perception of the reality and underlying correlations or causal relationships. “A single version of a truth”. Pat stressed the need for collaborative, exploratory approach to complex data analysis. He discussed social aspects and influences on the data analysis and visualisation. He also stressed that data analysis and statistics are two completely different disciplines, and they should not be treated synonymously. In his practice he estimated roughly 10% of data analysis is with the use of statistics.
Business intelligence is in need of good visualisation tools as to close the loop with humans being able to make well-informed decisions. Fine-tuning fully automated algorithms to produce human-like results does not and does not scale. Purely computer-based clustering vs. human-based data analysis and clustering.
Stress that Computer Scientist work with artists, designers and psychologists. Often simple visual tools work better than complex, photo-realistic or 3D visualisations. It is not about appealing appearance, but rather about paradigms and applicability to direct manipulation. The tools we use make us think in certain way, and the visualisation tools should expand our thinking, not constrain it.
VIZ QL – using visual language to generate queries, to generate diagrams straight from SQL-like query;
The biggest strength of Google search is not PageRank algorithm, but the ability to approximate sorting of billions of documents in time <1s. The approximated sorting is what gives the algorithm strength. Being exact is impractical.
Wolfram Mathematica as an example where visual and abstract are easily manipulated and interchanged. One can manipulate abstract and visual elements (e.g. pictures) the same way as one does calculus.
Simon’s example of number scrumble game (each player in turns picks a single digit number as to collect numbers that add up to 15).
The other speaker, Jos Stam talked about the visual space. He stressed that humans always operate in bounded, 4-dimensional, non Euclidean space, that is unique to each individual person. He talked about work by Paul Bach y Rota, on neural plasticity and neurological aspects of human visual perception and cognition. He gave examples of using non visual stimuli to generate “visual” cognition (through stimulating back of non-seeing patients or through stimulating their tongue). He have examples of blind people being able to play ping-pong.
The last speaker of the symposium, prof. Helmut Pottmann from TU Wien. Helmut talked about his mathematics roots, and the remaining challenges in object modeling. In particular, the ability to better understand the structure and function of the 3D models.