# קולוקוויום וסמינרים

כדי להצטרף לרשימת תפוצה של קולוקוויום מדעי המחשב, אנא בקר בדף מנויים של הרשימה.

## קולוקוויום וסמינרים בקרוב

• ### CGGC Seminar: Functional Fluids on Surfaces

דובר:
עומרי איזנקוט (מדעי המחשב, טכניון)
תאריך:
יום ראשון, 30.11.2014, 14:30
מקום:
חדר 337, בניין טאוב למדעי המחשב

Fluid simulation plays a key role in various domains of science including computer graphics. While most existing work addresses fluids on bounded Euclidean domains, we consider the problem of simulating the behavior of an incompressible fluid on a curved surface represented as an unstructured triangle mesh. Unlike the commonly used Eulerian description of the fluid using its time-varying velocity field, we propose to model fluids using their vorticity, i.e., by a (time varying) scalar function on the surface. During each time step, we advance scalar vorticity along two consecutive, stationary velocity fields. This approach leads to a variational integrator in the space continuous setting. In addition, using this approach, the update rule amounts to manipulating functions on the surface using linear operators, which can be discretized efficiently using the recently introduced functional approach to vector fields. Combining these time and space discretizations leads to a conceptually and algorithmically simple approach, which is efficient, time-reversible and conserves vorticity by construction. We further demonstrate that our method exhibits no numerical dissipation and is able to reproduce intricate phenomena such as vortex shedding from boundaries.

• ### CGGC Seminar: Continuous Medial Representation of Raster Images in Image Shape Analysis and Classification

דובר:
ליאוניד מסטטסקי (אונ' מוסקבה)
תאריך:
יום שני, 1.12.2014, 11:00
מקום:
חדר 337, בניין טאוב למדעי המחשב

Medial representation of object shape (skeleton and radial function) is a powerful and widely used tool for image shape analysis. Originally, the concept of skeleton was denoted for continuous objects: the skeleton of a closed region in Euclidean plane is a locus of centers of maximum empty circles in this region. And radial function is defined in every skeleton point and is equal to the radius of inscribed circle centered in this point. However, this concept became the most popular in shape analysis and classification of discrete bitmap digital images. Therefore, the need to generalize the concept of the skeleton for discrete images was raised.

In principle, we can formulate two approaches to extend the concept of the skeleton to discrete images. The first approach, which is the most popular because of ease of implementation, will be called discrete. It consists in a morphological transformation of the original image (Figure a) and construction new image (Figure b), which can be regarded as a skeleton. In this new bitmap medial axis represented by discrete lines one pixel wide. We can say it is a digital image of the skeleton. The discrete approach is implemented in different ways: based on distance maps, thinning etc.

Another approach, which we call continuous, consists of:
- approximation of a discrete object by the polygonal geometrical figure in terms of a continuous geometry (Figure c);
- construction of line segment Voronoi diagram for sides of this figure;
- extract a continuous skeleton of discrete objects from Voronoi diagram (Figure d);
- construction of continuous medial representation as a set of fat Bezier curves.

We present the full implementation of the continuous approach. We propose an original method to obtain continuous skeleton and radial function and represent it as a planar graph whose edges are described by straight lines and quadratic Bezier curves.

• ### Pixel Club: Geometric Normalization for Improved Facial Action Unit Detection

דובר:
אריק יודין (מדעי המחשב, טכניון)
תאריך:
יום שלישי, 2.12.2014, 11:30
מקום:
חדר 337, בניין טאוב למדעי המחשב

We employ a geometric framework to extend the concept of data normalization to the domain of functions that lie on manifolds. We pose normalization in this context as an embedding of all examples into manifolds nearly isometric to one another. Using novel geometric tools, we propose an implementation for the case of discretized functions on triangulated two-dimensional meshes.

We apply the proposed geometric normalization technique to the task of automatic Action Unit (AU) detection. This problem has received much attention in the literature ever since Ekman and Friesen introduced the Facial Action Coding System (FACS) in an attempt to codify facial expressions via modular components called Action Units. By normalizing cross-subject examples to a common template face, we are able to improve the results of state-of-the-art AU detection algorithms.

• ### Principles of Shape Analysis

דובר:
Mooly Sagiv - Colloquium Lecture
תאריך:
יום שלישי, 2.12.2014, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20141202_14_30_Sagiv.html
• ### ceClub: Making Machine Learning Accessible with GraphLab

דובר:
דני ביקסון (GraphLab)
תאריך:
יום רביעי, 3.12.2014, 11:30
מקום:
חדר 337, בניין טאוב למדעי המחשב

GraphLab started as research project at Carnegie Mellon University were our main goal was implementing machine learning methods in large scale, (and writing papers about it!). Despite thousands of users of our open source software (and many papers written), the experience of setting up the system and using a distributed system was quite frustrating, where our target audience was mainly double PhDs with machine learning and graph theory & distributed systems background.

When forming a company around GraphLab, we have decided to switch our focus and make machine learning much more accessible for the masses. In this talk I will summarize our experience in this direction. I will describe data structures we found useful for many commonly repeating computation patterns in machine learning, including support for dense data, graph (sparse) data, text records and even images. I will describe how we tackle the problem of wrapping up machine learning algorithms in a way which is user friendly to non experts, while algorithmic details can be tweaked layer. During the talk I will show many demos of cool applications that run on GraphLab.

About GraphLab: GraphLab is a popular open source big data analytics project initiated in 2009 at Carnegie Mellon University. GraphLab implements numerous machine learning algorithms using a highly efficient multicore machines and over clusters. A year and a half ago we have formed a company in Seattle to backup the open source project, and grown to a machine learning research group of around 12 PhDs. GraphLab is free for academic usages, and has a major open source component.

Bio:
Danny is holding a PhD in distributed systems at the Hebrew University (advised by Prof. Danny Dolev and Prof. Dahlia Malkhi), and was a postdoc at the machine learning department at Carngie Mellon University (advised by Prof. Carlos Guestrin and Prof. Joe Hellerstein from Berkeley). Danny was involved in the GraphLab project from its early stages and is a co-founder of the GraphLab company. His research interests are distributed algorithms and large scale machine learning.

• ### Theory Seminar: Shrinkage of De Morgan Formulae by Spectral Techniques

דובר:
אבישי טל (מכון ויצמן למדע)
תאריך:
יום רביעי, 3.12.2014, 21:30
מקום:
טאוב 401

We give a new and improved proof that the shrinkage exponent of De Morgan formulae is 2. Namely, we show that for any Boolean function $f: \{0,1\}^n \to \{0,1\}$, setting each variable out of $x_1, ..., x_n$ with probability $1-p$ to a randomly chosen constant, reduces the expected formula size of the function by a factor of $O(p^2)$. This result is tight and improves the work of Hastad [SIAM J. C., 1998] by removing logarithmic factors.

As a consequence of our results, we strengthen the best formula size lower bounds for functions in P.

The proof mainly uses discrete Fourier analysis, and relies on a result from quantum query complexity by Laplante et al. [CC, 2006], Hoyer et al. [STOC, 2007] and Reichardt [SODA, 2011].

No prior knowledge about quantum computing is assumed.

• ### Shape and Content: A database-theoretic perspective on the analysis of data structures

דובר:
Tomer Kotek - CS-Lecture -
תאריך:
יום שני, 8.12.2014, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20141208_14_30_Kotek.html
• ### The Cryptographic Lens

דובר:
Shafi Goldwasser - Colloquium Lecture
תאריך:
יום שלישי, 9.12.2014, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20141209_14_30_Goldwasser.html

דובר:
יונג ג'ון קים (מדעי המחשב, טכניון)
תאריך:
יום ראשון, 14.12.2014, 14:30
מקום:
חדר 337, בניין טאוב למדעי המחשב

Generation of precise contact motions for freeform geometry is essential for many applications such as mechanical and manufacturing engineering and motion planning in robotics. Yet contemporary approaches construct the contact motions of geometric objects only for simple geometric objects, typically those bounded by line segments and circular arcs. In this talk, we present an efficient algorithm for constructing precise contact motions (within machine precision) to the case of general freeform planar objects bounded by B-spline curves. We then extend the algorithm to the tool path computation for the CNC machining.

• ### Scalable algorithms for translating natural language to logical form

דובר:
Jonathan Berant - CS-Lecture
תאריך:
יום ראשון, 21.12.2014, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20141221_14_30_Bernat.html
• ### Lessons Learned from and for Requirements Engineering and Building Construction: A Case Study of Requirements Engineering for a Synagogue Kitchen with Use Cases and Scenarios

דובר:
Daniel M. Berry - Colloquium Lecture -
תאריך:
יום רביעי, 24.12.2014, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20141224_14_30_Berry.html
• ### An Algorithmic Approach for Analyzing Social Phenomena

דובר:
Sigal Oren - CS-Lecture
תאריך:
יום ראשון, 28.12.2014, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20141228_14_30_Oren.html
• ### Aspects of Submodular Maximization Subject to a Matroid Constraint

דובר:
Moran Feldman - CS-Lecture
תאריך:
יום חמישי, 1.1.2015, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20150101_14_30_Feldman.html
• ### Parallel Repetition From Fortification

דובר:
Dana Moshkovitz - Colloquium Lecture -
תאריך:
יום חמישי, 8.1.2015, 12:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20150108_12_30_Moshkovitz.html
• ### When Exactly Do Quantum Computers Provide a Speedup?

דובר:
Scott Aaronson - Colloquium Lecture
תאריך:
יום חמישי, 8.1.2015, 14:30
מקום:
חדר 337-8 טאוב.
קישור:
http://www.cs.technion.ac.il/~colloq/20150108_14_30_Aaronson.html