הפקולטה להנדסה ע"ש איבי ואלדר פליישמן

סמינר המחלקה להנדסה ביו רפואית

18 בינואר 2015, 14:15

הבניין הרב תחומי , חדר 315

דר' יעל יניב

The clocks that make our heart tick

Cells within the sinoatrial node, the heart’s pacemaker, generate

spontaneous action potentials (APs) that are conducted to the ventricle

and excite ventricular myocyte contraction.

For several decades, the prevailing viewpoint on the heart’s pacemaker

activity has related heart rate and rhythm only to the surface membrane,

suggesting that activity is driven only by voltage-and time-dependent ion

channels (membrane clock). My recent experimental studies, however,

have shown discovered that the pacemaker membrane clock is tightly

coupled to the intracellular dynamics on a beat-to-beat basis. Moreover,

the mitochondria dynamically buffer cytosolic affecting the sarcoplasmic

reticulum load. The change in sarcoplasmic reticulum load translates into

change in the AP firing rate.

In my talk, I will present bioenergetics and systems biology perspectives

of the new coupled-clock pacemaker paradigm.

סמינר מחלקתי Limor Nizri

10 ביוני 2015, 15:00

206

School of Mechanical Engineering Seminar
Wednesday, June 10, 2015 at 15:00
Wolfson Building of Mechanical Engineering, Room 206

A molecular method for testing the effectiveness of UV systems on-site

Limor Nizri

M.Sc Student of Dr. Hadas Mamane

Ultraviolet (UV) disinfection is a barrier to possible contamination of groundwater pollution. The well-water is drawn and treated with UV to meet the Israeli and internationally accepted standards, for protective radius of well water. Activity in the protective radius of well-water may result in microbial contamination and include sewage lines, barns, landfills, wastewater treatment plants, industrial zone and irrigation with wastewater effluent. Due to filtration of water through the soil, groundwater contains a low concentration of bacteria. Enumerating the bacterial concentration in water via the standard method of ”plate count enumeration” can only quantify bacteria that exist in water and are culturable on nutrient media.
Consequently, it is not possible to examine the effectiveness of UV disinfection of well-water operating in the field via standard culturing methods due to the insufficient concentration of indicator bacteria in the raw water, as determined by conventional methods. In this study, we examined the feasibly of a molecular method for determining the efficacy of UV disinfection systems operating on-site. We have developed a method of using the total DNA from a full array of bacterial species in the water sample without any need of growing or isolating a specific type of bacteria, and succeeded to correlate the total DNA of bacteria to the UV dose in the lab (spiked) in buffered water and in natural well-water without growing or isolating a specific type of bacteria. The "DNA damage" can bridge the gap for water without bacterial counts and may be used to validate the efficacy of full commercial-scale systems. Developing methods for evaluating UV disinfection on-site may serve as the stepping stone for wide acceptance of UV disinfection, especially in such waters.

סמינר מחלקתי

13 בינואר 2015, 14:00

Bootstrapping Semantic Locations from Human Mobility Data

Omer Barak –M.Sc.student

Abstract:

In recent years, the prevalence of mobile phones fitted with GPS and other positioning technologies has made people's exact location in space and time an accessible piece of data. The process of transforming the physical location (e.g. <32.156, 34.69>) of such a mobile user into a semantic location (e.g. "Home" or "Work") is called Semantic Labeling. This semantic representation has numerous advantages: it allows easier discovery of mobility patterns, regardless of the user's geographical setting; it enables straightforward comparison between users; and it does not expose the physical location of private places such as the user's home.

In this talk I will present a simple model for human mobility and a framework for semantic labeling based on that model. The framework uses supervised learning and utilizes spatial, temporal and contextual data. We refer to this process as "bootstrapping" semantic locations since the input data the framework uses is only the time-stamped physical locations, without any a-priori knowledge of the user itself: habits, demographic features, social group and so on. I will discuss the experiments used to evaluate this framework and will present two sample applications of it: for privacy-preserving dataset release and for measuring user similarity.

This work was performed under the supervision of Dr. Eran Toch.

ההרצאה תתקיים ביום שלישי 13.1.15, בשעה 14:00 בחדר 206, בנין וולפסון הנדסה, הפקולטה להנדסה, אוניברסיטת תל-אביב.

סמינר מחלקתי

20 בינואר 2015, 14:00

חדר 206 בניין וולפסון

Online Stochastic Optimization with a Moving Target

Prof. Assaf Zeevi

Abstract :

Stochastic approximation refers to a family of methods whose objective is to sequentially estimate the optimum point of a fixed yet unknown cost function whose observations are confounded by statistical noise. Since its inception in the 1950's, the method has been the subject of a voluminous literature and is widely used in a variety of application settings. In this talk we will develop some theory that attempts to expand the scope of stochastic approximation to non-stationary environments, i.e., when the cost function is allowed to change over time. Some (hopefully interesting) connections will be drawn to a recent strand of literature on online convex optimization, which studies the aforementioned problem in a so-called adversarial setting.

Joint work with Omar Besbes, Columbia University, and Yoni Gur, Stanford University.

ההרצאה תתקיים ביום שלישי 20.1.15, בשעה 14:00 בחדר 206, בנין וולפסון הנדסה, הפקולטה להנדסה, אוניברסיטת תל-אביב.

EE Seminar: Dana Kalinsky

~~Dana Kalinsky,
M.Sc. student under the supervision of Prof. Simon Litsyn

Wednesday, January 21, 2015 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

New Bounds on the Capacity of (d,k)-RLL Codes

Abstract

The usage of run-length limited codes is very wide today in magnetic and optical storage as well as in fiber-optic and wireless communication applications. When dealing with one-dimensional systems we can achieve an exact capacity value of the code, that is the asymptotic information rate as defined by Shannon. The emergence of two-dimensional systems, for example holographic storage systems arises the challenge to calculate the capacity of higher dimensional run-length limited codes. In higher dimension systems exact capacity values are rare. One can find many attempts to overcome the calculation difficulties and to bound the capacity of high dimension run-length limited systems.
In this thesis we attempt to improve estimates of the capacity of two-dimensional run-length limited systems. By further developing the probability approach introduced by Schwartz and Vardy, we suggest a new method to calculate the capacity bound of run-length limited systems. We show that an improvement in the capacity bound is possible in some scenarios.
Specifically, for one-dimensional and two-dimensional run-length limited systems, (0,k)-RLL, we achieve better upper bounds when k is even. For one-dimensional systems we compare our results to the exact capacity values and show that our results converge to the exact capacity values as k grows.

21 בינואר 2015, 15:00

חדר 011, בניין כיתות-חשמל

Seminar 15.1.15

15 בינואר 2015, 15:00

011 KITOT

Physical Electronics Dept.

You are invited to attend a lecture

Dr. Stanislav Derevyanko

A Marie Curie Visiting Fellow at the Weizmann Institute

On the subject:

The Nonlinear Fourier Transform for Optical Communications: Overcoming the Capacity Crunch

The increasing demand for optical telecommunication traffic has put the existing communication systems dangerously close to the currently achievable fiber

Communication capacity. It is projected that within a decade the existing systems will hit this limit which will severely constrain the further growth of the Internet traffic, something which has been named “the capacity crunch”. It is widely believed that one of the most severe current capacity limiting factors in optical telecommunications is fiber (Kerr) nonlinearity. This talk will concentrate on a rapidly developing area of eigenvalue communications and the nonlinear Fourier transform that allow in principle a complete simultaneous compensation of both chromatic dispersion and fiber nonlinearity.

I will give a brief overview of the state of the art of this recently rediscovered technique concentrating on the so called Nonlinear Inverse Synthesis method which combines the benefits of well-known and easily implemented linear modulation formats and waveforms (OFDM, WDM, Nyquist) with all the power of the Nonlinear Fourier Transform.

Thursday, 15 January 2015 at 15:00

Room 011, Kitot Build.

Seminar 12.1.15

12 בינואר 2015, 16:00

Kitot 011

Physical Electronics Dept.

You are invited to attend a lecture

Prof. Jacob B. Khurgin

Dept. of Electrical and Computer Engineering, Johns Hopkins University

On the subject:

Plasmonic sensors – their merits and limitations

Fluorescence, resonance and off-resonance Raman spectroscopy are all precise and versatile techniques for identifying small quantities of chemical and biological substances. One way to improve the sensitivity and specificity of these measurement techniques is to use enhancement of optical fields in the vicinity of metal nanoparticles. The degree of enhancement, however, is drastically different as Raman enhancement of 10 orders of magnitude or more has been consistently measured in experiment, while the enhancement of the seemingly similar process of fluorescence is typically far more modest. While resonance Raman scattering has the advantages of higher sensitivity and specificity when compared with the ordinary, non-resonant Raman process, its plasmon enhancement is far less spectacular. In fact, both fluorescence and resonance Raman measurements are subject to quenching when the molecule is placed too close to the metal surface, such an effect, however, is completely absent from the normal non-resonant Raman process. We shall present an analytical model that reveals the physics behind the strikingly different orders of magnitude in enhancement that have been observed, provide a fundamental explanation for the quenching effect observed in fluorescence and resonance Raman but not in normal Raman, establish limits for attainable enhancement, and outline the path to optimization of all three processes.

Monday, 12 January 2015 at 16:00

Room 011, Kitot-Hachmal Build.

EE Seminar: Dr. Kobi Cohen

~~ (The talk will be given in English)

Speaker: Dr. Kobi Cohen
Coordinated Science Laboratory at the University of Illinois at Urbana-Champaign.
Wednesday, January 28th, 2015
15:00 - 16:00
Room 011, Kitot Bldg., Faculty of Engineering

Active Learning for Dynamic Search Problems in Large-Scale Systems
Abstract
As networks are expected to grow more complex, developing effective algorithms for information gathering becomes more critical for learning tasks. In active learning (first considered by Chernoff in 1959, referred to as sequential design of experiments), a decision maker is required to actively choose among different processes to effectively shape the quality of the observations so as to optimize certain objectives in the system. In this talk I will present several results from my research on active learning for dynamic search problems (referred to as anomaly detection). We consider a system with a large number of processes, among which a few are anomalous. At each time, a subset of the processes can be observed, and the observations from each chosen process follow two different distributions, depending on whether the process is normal or abnormal. The problem is to find a sequential search strategy that optimizes a certain objective in the network (e.g., detection delay, cost) subject to reliability constraints. (Asymptotically) optimal algorithms will be presented to solve the anomaly detection problem under different objectives. The problems considered in this work find applications in anomaly detection in cyber systems, spectrum scanning in cognitive-radio networks, target search, and event detection in sensor networks.
Bio:
Kobi Cohen received the B.Sc. (cum laude) and Ph.D. degrees in electrical engineering from Bar-Ilan University, Ramat Gan, Israel, in 2007 and 2013, respectively. He is currently a postdoctoral research associate at the Coordinated Science Laboratory at the University of Illinois at Urbana-Champaign. Before joining UIUC, he was with the Department of Electrical and Computer Engineering at the University of California, Davis, as a postdoctoral researcher. His main research interests include decision theory, stochastic optimization, multi-agent learning, game theory and statistical inference, with applications in dynamic systems, cyber-security, wireless and wireline networks.

28 בינואר 2015, 15:00

חדר 011, בניין כיתות-חשמל

EE Seminar: Naama Litichever

~~
Speaker: Naama Litichever
M.Sc. student under the supervision of Prof. Simon Litsyn

Wednesday, January 21st, 2015 at 15:30
Room 011, Kitot Bldg., Faculty of Engineering

Sub-Optimal Decoding of Polar Codes
Abstract
Polar codes, introduced by Arikan, are the first known practical codes to achieve the Shannon capacity over symmetric discrete memoryless channels. Their code construction is based on the channel polarization phenomenon. The encoding as well as the decoding procedure of polar codes can be implemented with O(N logN) complexity, where N is the block length of the code.
Arikan used the binary u + v; v mapping to construct polar codes. Later
works showed that other kernel matrices, both binary and non-binary, can be used to construct polar codes. Non-binary kernels have great performance potential, but this comes with a very high price of complexity. In this thesis, we propose a sub-optimal decoding algorithm for polar codes with both binary and non-binary kernels, that lowers their decoding complexity. The reduced complexity makes the use of high-order non-binary kernels feasible.
List decoding algorithms are very common in polar codes' decoding due to their high performance. We suggest here a list decoding algorithm, for binary kernels, which improves the list decoding of polar codes in terms of complexity and latency without any performance degradation.
Moreover, we generalize this algorithm for the non-binary case, providing a non-binary list decoding algorithm with lower latency and complexity in comparison with the well-known polar codes list decoder, and with the same performance.

21 בינואר 2015, 15:30

חדר 011, בניין כיתות-חשמל

Understanding the relations between transcript features and gene expression

11 בינואר 2015, 14:15

הבניין הרב תחומי , חדר 315

שחם גלעד

תלמיד המחלקה להנדסה ביו רפואית לתואר שני ירצה בנושא:

Understanding the relations between transcript features and gene expression

Gene expression is the process in which information encoded in the DNA is used to synthesize new proteins. Understanding the way gene expression is encoded in transcripts should not only contribute to disciplines, such as functional genomics and molecular evolution, but also to biotechnology and human health. Previous studies in the field mainly aimed at predicting protein levels of genes based on their transcript features by assuming that the effect of each transcript feature on gene expression is monotonic.

In this work, we aim to understand, for the first time, if indeed the relations between transcript features and measurements related to the different stages of gene expression are monotonic. To this end, we analyze 5,432 transcript features and gene expression measurements (e.g. mRNA levels, ribosomal densities, protein levels) of
4,367 S. cerevisiae genes. We use the Maximal Information Coefficient (MIC) in order to identify potential relations that are not necessarily linear or monotonic.

Our analyses demonstrate that the relation between most transcript features and the examined gene expression measurements is monotonic. In addition, in the cases of deviation from monotonicity the relation/deviation is very weak. These results should help in guiding the development of computational gene expression modeling and engineering, and improve the understanding of this process. Furthermore, the relatively simple relations between a transcript’s nucleotide composition and its expression should contribute towards better understanding of transcript evolution at the molecular level.

עמודים