הפקולטה להנדסה ע"ש איבי ואלדר פליישמן

EE Seminar: Unsupervised methods for joint segmentation of an image set

~~
Gal Malach,
M.Sc. student under the supervision of Prof. Nahum Kiryati and Prof. Nir Sochen

Wednesday, June 24th, 2015 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Unsupervised methods for joint segmentation of an image set

Abstract

Segmentation of a region of interest in an image is significantly influenced by the availability of prior information. However, in many cases such prior information is not available or not compatible with the image in hand. Luckily, nowadays applications can benefit from the large availability of images with similar or close contents, e.g., multiple overlapping viewpoints of the same object or use of various image acquisition methods. As a result, there is a shift from the classical prior based segmentation to a co-segmentation approach. Under the co-segmentation framework simultaneous segmentation of two images is applied where each segmentation is supported by the other to utilize the large commonality between the two images. In the first part of this work we review the current state-of-the-art co-segmentation methods and present a generalization of the pair-wise methods to account for image ensemble. Furthermore, we set a theoretical framework to draw the connection between our generalized pair-wise co-segmentation method to a probabilistic atlas-based approach.

In the second part of this work, we present a novel method for co-segmentation of common regions of interest (ROIs) in multiple image volumes of possibly different qualities or in the presence of inconsistencies. In contrast to the classical atlas-based approaches, only a single annotated image is used as a prior. The joint segmentation process is supported by the evolving segmentation of each of the individual images while accounting for the varying confidence levels. The proposed approach uses soft segmentation. Labeling uncertainty of a given voxel is determined by its spatial proximity to the estimated ROI's boundaries and the dynamically changing segmentation's confidence, learned throughout the joint segmentation process. Our contribution consists of a robust segmentation method that advances existing co-segmentation algorithms. The proposed algorithm is supported by a theoretical derivation which shows that it is a generalization of previous approaches. Promising results are demonstrated for the joint segmentation of neuroanatomical structures across 50 MR scans of different subjects. The proposed algorithm supports multi-modal data where each modality reveals different features of the ROI. A significant variance in the confidence levels was observed when using the proposed method for cross-modality joint segmentation of brain tumor tissues (BraTS).

24 ביוני 2015, 15:00

חדר 011, בניין כיתות-חשמל

סמינר מחלקתי

09 ביוני 2015, 14:00

חדר 206 בניין וולפסון

Effects of Security Warnings and Blocking on Risk Taking in IT Systems

Omer Dembinsky - Department of Industrial Engineering

Abstract

While working on an IT system (computer, smartphone) a user might choose to perform a risky behavior, such as browsing to an unknown website, downloading a file/application or connecting an external device. These actions are meant to assist the user in trying to achieve a specific goal (work, entertainment) which gives a certain value to the user. The result may be the desired one, but it might also be a negative outcome if a malicious program (virus, Trojan horse) is hidden and activated.

In order to help the user decide whether it is safe to perform a certain action or not a security system can be used. The system monitors the actions performed by the user and when relevant provides a warning or even blocks the action in order to protect the user.

This work examines different security system designs, focusing on the comparison between a Warning and a Blocking system, and their influence on the user’s risk taking, as well as on the productivity of work and the amount of security events (i.e. downloading malware). A normative mathematical model of the human behavior in this situation was developed and an experiment was carried out to examine the model and to compare its predictions to the actual user behavior.

This work was performed under the supervision of Prof. Joachim Meyer.

ההרצאה תתקיים ביום שלישי 09.06.15, בשעה 14:00 בחדר 206, בנין וולפסון הנדסה, הפקולטה להנדסה, אוניברסיטת תל-אביב.

EE Seminar: Integer-Forcing for Multiterminal Communication Problems

~~Speaker: Or Ordentlich
PhD student under the supervision of Prof. Uri Erez

Monday, June 29th, 2015 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Integer-Forcing for Multiterminal Communication Problems

Abstract

Integer-Forcing (IF) is a new framework, based on compute-and-forward, for decoding multiple integer linear combinations from the output of a Gaussian multiple-input multiple-output (MIMO) or multiple-access (MAC) channel. Integer-forcing is applicable when all transmitters use nested linear/lattice codes.

Building on the IF framework, we derive new theoretical results and develop new low-complexity coding schemes for several problems.

We begin by studying the capacity region of the Gaussian MAC under the constraint that all users transmit from a chain of nested lattice codes. Interestingly, the obtained rate-region depends on number-theoretic properties of the channel gains. Then, we apply these results in conjunction with lattice interference alignment to approximate the sum capacity of the symmetric K-user Gaussian interference channel.

We next apply the IF approach to arrive at a new low-complexity scheme, IF source coding, for distributed lossy compression of correlated Gaussian sources under an MSE distortion measure. The performance of the proposed scheme closely follows Berger-Tung's inner bound. Moreover, a one-shot version of IF source coding is described and analyzed.

29 ביוני 2015, 15:00

חדר 011, בניין כיתות-חשמל

סמינר מחלקתי

09 ביוני 2015, 13:00

חדר 206 בניין וולפסון

Efficient algorithms for the time cost tradeoff problem
Dorit S. Hochbaum
UC Berkeley

Abstract :

The time cost tradeoff problem in project management, TCT, is to expedite the durations of activities in order to achieve shorter target project completion time than possible with the normal durations. The linear TCT problem, in which the expediting costs of each activity are linear as a function of the number of time periods reduced, is commonly solved using linear programming. We present here an algorithm, based on a non-polynomial time algorithm by Phillips and Dessouky 1977, PD-algorithm. PD-algorithm repeats iterations in each of which the project duration is reduced by one time unit, at a minimum cost. The activities to expedite, in order to reduce the project duration by one unit, correspond to forward and backward arcs, that reside on a minimum cut in a respective graph.

We present here previously unknown properties of PD algorithm and use these properties to devise a variant of PD-algorithm and that runs in polynomial time for both linear and convex expediting costs and uses a minimum s,t-cut routine at each iteration.

For the uniform costs TCT problem we present here a new algorithm that runs in O(mn) for a project network on n activities and m precedence constraints. The key building block of this algorithm is the generation of all optimal minimum cuts simultaneously, in the same complexity as a single minimum cut. We discuss the implications of this algorithm for maximum weight flow problem with unit capacities.

Time permitting, we will show a dual algorithm for TCT, that solves the problem in time O(n log n (m + n \log n).

ההרצאה תתקיים ביום שלישי 09/06/15, בשעה 13:00 בחדר 206, בנין וולפסון הנדסה, הפקולטה להנדסה, אוניברסיטת תל-אביב

סמינר מחלקתי Prof. Hadas Kress-Gazit

22 ביוני 2015, 15:00

4.6.15

04 ביוני 2015, 15:00

011 Kitot

The Iby and Aladar Fleischman

Faculty of Engineering

You are invited to attend a special department

Seminar and the annual gathering in memory of

Prof. Norman W. Rosenberg

Opening remarks in memory of Prof. Rosenberg will be given by Prof. Mark Shtaif, Chair of the Department of Physical Electronics.

Special Seminar:

Breach of fundamental electrodynamic symmetries in nano-scale optics

Mr. Yarden Mazor

Winner of the Rosenberg Prize for Graduate Students

Abstract:

Reciprocity and symmetry are basic attributes of many physical phenomena involving wave propagation such as sound waves and electromagnetic waves. This fact makes cases where reciprocity is broken interesting from both scientific and practical points of view. Scientifically - unveiling the physical mechanisms involved, and practicallly - using this physical understanding for the design of non-reciprocal components, with the ultimate one being the one-way waveguide. Several designs for 1D and 2D one-way waveguides were recently proposed by our group. All of them are based on using metallic nano-particles in periodic and quasi-periodic arrays and exploiting the magneto-optical effect as a symmetry breaking mechanism. Combining the rotation induced by magnetic fields, and geometrical rotation (or other types of asymmetry) in our systems allows for enhanced non-reciprocity and one-way guiding - which was later extended into sector-way guiding in 2D. These structures offer new ways to control the flow of light in 1D and 2D in the nanoscale and more flexible dispersion engineering which can allow the design of miniature optical isolators, circulators, and other types of optical devices.

____________________

Thursday, June 4^th, 2015, at 15:00 p.m.,

Rooms 011, Kitot build, Tel Aviv University Campus, Ramat Aviv

EE Seminar: Efficient Linear Programming Decoding of HDPC and MDPC codes

~~
Shay Tabiby,
M.Sc. student under the supervision of Prof. Yair Be'ery

Wednesday, June 10, 2015 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Efficient Linear Programming Decoding of HDPC and MDPC codes

Abstract

A new Linear Programming (LP) decoding algorithm for block codes is presented along with detailed performance and complexity analysis for several classes of codes and decoding algorithms. The codes include High Density Parity Check (HDPC) codes, Moderate Density Parity Check (MDPC) codes and Polar codes. The decoding algorithms include Belief Propagation (BP), few versions of LP and the new LP decoder (denoted SGD). The SGD decoder is based on a sparse representation of the Tanner graph which utilizes the more reliable bits to increase the reliability of the less reliable bits in order to make the LP algorithm converge. The new decoder achieves better performance and complexity than the latest most advanced LP decoders for BCH codes and the same performance as Successive Cancellation List decoder for Polar codes.

10 ביוני 2015, 15:00

חדר 011, בניין כיתות-חשמל

EE Seminar: Physical modeling of collective motion in animal groups: carrying ants

~~(The talk will be given in English)

Prof. Niv Gov
Weizmann Institute

Monday, June 8th, 2015
15:00 - 16:00
Room 011, Kitot Bldg., Faculty of Engineering
Physical modeling of collective motion in animal groups: carrying ants
Abstract

Examples of collective motion exist throughout biology, from the collective motion of cells during morphogenesis, wound healing and cancer, and for animal groups that migrate, forage and hunt together. We have been applying physical modeling for such systems, at the different scales. At the level of animal groups, we have been studying the collective load carrying by ants, as they transport food to the nest. In this system we propose a model where the decision making at the level of the single ants is affected by the whole group through the forces transmitted by the load. The model for this system is therefore inherently mean-field in nature, and exhibits a thermodynamic order-disorder transition. Due to the mean-field nature, the critical temperature for the system increases with the system size, which allows to probe it by simply changing the load size. These predictions agree with observations.

08 ביוני 2015, 15:00

חדר 011, בניין כיתות-חשמל

EE Seminar: Nonlinear Interference in Fiber-Optic Communications

~~Speaker: Ronen Dar
PhD student under the supervision of Prof. Meir Feder and Prof. Mark Shtaif

Wednesday, June 3rd, 2015 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Nonlinear Interference in Fiber-Optic Communications

Abstract
The nonlinearity of optical fibers is arguably the most significant factor limiting the capacity of modern wavelength division multiplexed (WDM) communications systems. The Kerr nonlinear effect of the fiber introduces complex distortions into the transmitted optical waveforms, generating nonlinear interference noise (NLIN) between the various WDM channels; this diminishes the receiver's capability of decoding the transmitted data and limits the reach of the system and the achievable data-rates. The modeling of NLIN and the understanding of its impact on system performance are therefore key components in the efficient design of fiber-optic communication systems and in the evaluation of their ultimate reach and capacity.
We formulate a discrete-time channel model for an arbitrarily link and use it to study the statistical nature of NLIN. We examine the accurate extraction of the NLIN variance, its dependence on modulation format, the role of nonlinear phase-noise and polarization rotation, and the existence of temporal correlations. We propose new practical schemes for nonlinear interference mitigation that exploit the temporal correlations and further derive new bounds on the channel capacity. In addition, we present a pulse-collision theory that provides qualitative and quantitative insight into
the build-up of NLIN, offering a simple and intuitive explanation to many of the reported and previously unexplained phenomena.

03 ביוני 2015, 15:00

חדר 011, בניין כיתות-חשמל

סמינר מחלקתי

02 ביוני 2015, 13:30

חדר 206 בניין וולפסון

Locating the Optimal Project Control point based on the Entropy of the project duration

Shiva Cohen-Kashi – M.Sc. student

Abstract:

A project is an undertaking with defined objectives and established starting and ending points that needs to be completed within specified time, resources and budget constraints. Typically during the planning phase a project plan is defined with the project activities, which enables the project manager to derive the expected total project duration. Unfortunately, in most cases the actual duration of the activities differs from the planned duration, leading to inability to anticipate the exact project duration. Enabling the managers to better predict the project duration and completion date is a major advantage which facilitates better decision making, more accurate risk management and higher profitability. This can be achieved by proper monitoring and control policy which will provide more information on the project and specifically the expected duration and completion date. Identifying the optimal control or inspection policy is a difficult task as it requires evaluating each potential policy and depending on the project duration and required number of points can require evaluating hundreds of options even for small scale projects

We propose an analytical approach for identifying the optimal project control points timing using concepts from the Information Theory. The methodology used is based on simulation-optimization using an exhaustive search over all possible control points and selecting the ones that provide the highest potential information on the project duration. The control points are selected based on the amount of information gained on the expected project duration measured by the entropy. The algorithm complexity is not depended on the number of the activities and it can easily process large projects with hundreds of activities.

This work was performed under the supervision of Prof. Irad Ben Gal

ההרצאה תתקיים ביום שלישי 02/06/15, בשעה 13:30 בחדר 206, בנין וולפסון הנדסה, הפקולטה להנדסה, אוניברסיטת תל-אביב

עמודים