3D computer graphics and interactive data visualization methods are becoming increasingly important in a wide range of application domains including but not limited to product marketing, entertainment, engineering as well as sciences. In this seminar, we study technologies, methods and use of graphics and visualization methods, comparing and analyzing their algorithms, system implementations as well as applications in software products.
Good knowledge of mathematical foundations, algorithms and data structures as well as programming is necessary. Knowledge of fundamental principles in one or more areas of computer graphics, scientific visualization, image processing, computer vision or multimedia is required. Strong computer science and mathematical skills are beneficial.
The seminar targets MSc students and BSc students in advanced semesters.
This semester's topics concentrate on Dimensionality Reduction in Data and Scientific Visualization, covering mathematical concepts, algorithms and visualization applications.
Successful completion of the seminar requires the following:
Note: There is no requirement for you to implement the method yourself.
Source references must primarily include technical research papers from international conferences or journals. Additional extra references can e.g. include book articles, course materials, case studies, online tutorials, technical blog-posts etc.
The written report is expected to include summarized information of the related works and your own critical analysis of the material. The report is expected to be around six (6) to ten (10) pages in a given format (SIGGRAPH Content Formatting). It is heavily recommended that you use the provided LaTeX template, but you are free to use any other application, as long as you provide a final PDF that matches the formatting of the given template.
Close attention must be paid to proper structure and formatting of the report. Using the appropriate style, placement of figures and tables, as well as correct references and citations is a must.
The report should introduce the technique and provide motivation for its use. You should then precisely state the problem the techniques are attempting to solve, followed by a summary and comparison of each of the methods the different references provide. Finally, conclude with a discussion of the techniques and the individual method's limitations and open problems.
The seminar presentations includes two talks, followed by a discussion of your presentation and the topic. Attendance and active participation in seminar presentations and discussions of other students is mandatory.
You will need to hand in all your presentation materials, such as slides, notes, figures etc.
Close attention must be paid to the structure of the presentation, which should in general include a short introduction and motivation of the topic, a precise statement of the problem, a detailed analysis of the method, a summary of the results and a personal conclusion with discussion of open problems, limitations and ideas.
It is strongly recommended that you rehearse your presentation beforehand and review the presentation with the seminar assistant.
The following list contains a number of predefined topics. Each topic includes one main visualization technique or application paper. These are provided to you as the core topics and as a starting point for your literature research into the topic. Whenever possible, try to find mostly recent works on the topics!
You are allowed to propose an additional topic in the given context of dimensionality reduction and its use in data visualization, however, you need to get it approved. Also you need to make sure that you can find a sufficient number of reference materials (i.e. including 1+1+3 technical references).
Principal Component Analysis (PCA) iPCA: An Interactive System for PCA-based Visual Analytics
Independent Component Analysis (ICA) ISpace: Interactive Volume Data Classification Techniques Using Independent Component Analysis
Linear Discriminant Analysis (LDA) iVisClassifier: An Interactive Visual Analytics System for Classification Based on Supervised Dimension Reduction
Multidimensional Scaling (MDS) Temporal MDS Plots for Analysis of Multivariate Data
Self-Organizing Maps (SOM) SOMFlow: Guided Exploratory Cluster Analysis with Self-Organizing Maps and Analytic Provenance
Locally Linear Embeddings (LLE) Exploratory Analysis and Visualization of Speech and Music by Locally Linear Embedding
t-distributed Stochastic Neighbor Embedding (t-SNE) Approximated and User Steerable tSNE for Progressive Visual Analytics
Spectral Clustering Open-Box Spectral Clustering: Applications to Medical Image Analysis
Spectral EmbeddingLaplacian Eigenmaps for Dimensionality Reduction and Data Representation (note: this is a long seminal paper that includes visualization examples; it can be used for the theory part as well)
A good starting point for finding recent publications (besides) Google are the ACM Digital Library, the IEEE Digital Library or the Eurographics Digital Library where a majority of the relevant publications are hosted. You can access the content from these Digital Libraries from within the UZH (VPN) network.
Further publication venues include the following conferences and symposia: IEEE Visualization/InfoVis/VASR, IEEE Pacific Visualization, EUROGRAPHICS, EuroVis, EuroVA, ACM SIGGRAPH, ACM SIGGRAPH Visualization Symposium, along with the associated journals (ACM Transactions of Graphics, IEEE Transactions on Visualization and Computer Graphics, and Computer Graphics Forum). See Section Links further below for links.
LaTeX template for your report:
Finally, Google is your friend -- most authors put their papers online either on their personal websites or in some University provided space. Further, you might find presentation notes, sample implementations and other notes that can help understanding otherwise technically-advanced papers.