המחלקה למדע והנדסה של חומרים | אוניברסיטת תל אביב

סמינר מחלקתי

Ami Moshiov

29 בדצמבר 2014, 15:00

וולפסון 206

School of Mechanical Engineering Seminar
Monday, December 29, 2014 at 15:00
Wolfson Building of Mechanical Engineering, Room 206

Family Bootstrapping: a Genetic Transfer Learning Approach for Onsetting the Evolution for a Set of Related Robotic Tasks

Amiram Moshaiov

School of Mechanical Engineering, Faculty of Engineering

Tel Aviv University, Israel

Studies on the bootstrap problem in evolutionary robotics help lifting the barrier from the way to evolve robots for complex tasks. It remains an open question, though, how to reduce the need for designer knowledge when devising a bootstrapping approach for any particular complex task. Recent developments in the field of transfer learning may help reducing this need and support the evolution of solutions to complex tasks, through task relatedness. Relying on the commonalities of similar tasks, we introduce a new concept of Family Bootstrapping (FB). FB refers to the creation of biased ancestors that are expected to onset the evolution of "a family" of solutions not just for one task, but for a set of related robot tasks. A general FB paradigm is outlined and the unique potential of the proposed concept is discussed. To highlight the validity of the FB concept, a simple demonstration case, concerning the evolution of neuro-controllers for a set of robot navigation tasks, is provided. The FB concept resulted from an earlier work on bootstrapping the co-evolution of soccer-like players, which is also briefly reported. The presentation is concluded with some suggestions for future research.

הנדסת חשמל - סמינר - אלכס יופיט

~~Electrical Engineering-Systems Department

*** SEMINAR ***

Alex Yufit
M.Sc. student under the supervision of Prof. Yair Be’ery

on the subject:

On Efficient Linear Programming Decoding of HDPC codes

      In our work we propose improved LP decoding techniques for HDPC codes. Our enhanced LP decoder generates several variants of the fundamental polytope for eliminating pseudocodewords and improving decoding performance. On top of the enhanced performance, our decoder reduces the complexity by removing all inactive LP constraints at each iteration. We compare the performance and the complexity of our enhanced decoded to the New Separation Algorithm (NSA) for several BCH codes using AWGN channel. From simulation results we observe that our enhanced LP decoder achieves significant performance gain over the NSA with similar decoding complexity.
      Inspired by the above results we propose three enhanced ADMM message passing decoders and discuss the tradeoffs between them. Our enhancements make the ADMM decoder suitable for HDPC codes. In simulation results our decoders show near-ML performance for BCH and Hamming codes that we’ve tested. Analysis of the decoding complexity implies that in high SNR regime our decoders have negligible overhead of the average complexity relative to the original ADMM decoder.
      Decoding complexity is a very important factor for any practical use. We propose a framework for detailed analysis of the complexity of ADMM message passing decoder. Computational complexity of ADMM decoding algorithm is expressed by the number of arithmetic operations. First, we derive an upper bound on the complexity of a single ADMM message passing iteration. Our upper bound depends linearly on the number of edges in the Tanner graph of a code. Then, relying on simulation results we count the average number of required message passing iterations to get an estimate of the computational complexity of decoding one codeword. We observe that in high SNR regime the most dominant factor affecting total decoder complexity is the complexity of a single ADMM message passing iteration. We extend the analysis to our enhanced ADMM decoders.

12 בנובמבר 2014, 15:00

בניין כיתות-חשמל חדר 011

סמינר מחלקתי Jane O'nill

10 בנובמבר 2014, 15:00

וולפסון 206

School of Mechanical Engineering Seminar
Monday, November 10, 2014 at 15:00
Wolfson Building of Mechanical Engineering, Room 206

ACTIVE CLOAKING OF INCLUSIONS
FOR FLEXURAL WAVES IN KIRCHHOFF PLATES

Miss Jane O’Neill

University of Liverpool

Abstract text.

We present a new method to create an active cloak for a single inclusion in a Kirchhoff plate in the presence of flexural waves. We consider both clamped and free edge boundary conditions. To build an effective cloak, we place control sources exterior to the scatterer and choose their intensities to eliminate propagating components of the scattered wave. Our initial discussion is on the canonical problem of active cloaking of a circular cylinder for which a closed form analytic solution is derived, and exemplified in a number of illustrative cases. We then consider a clamped scatterer of arbitrary shape with a smooth boundary and describe a method for solving the active cloaking problem. Further details can be found at

http://arxiv.org/abs/1403.0816.

Short Bio: Miss Jane O'Neill finished her integrated masters programme in mathematics at the University of Liverpool in June 2013. She is currently a second year PhD student at the same department.

סמינר מחלקתי

Dr. Igor Berisnkii

03 בנובמבר 2014, 15:00

וולפסון 206

School of Mechanical Engineering Seminar
Monday, November 3, 2014 at 15:00
The Wolfson Software Engineering Building, Room 104

Discrete and continuum models to determine the elastic and dynamical properties of materials and structures at micro- and nanoscale

Dr. Igor Berisnkii

Institute for Problems in Mechanical Engineering of Russian Academy of Sciences,

St. Petersburg Polytechnic University (St. Petersburg, Russia)

Solid mechanics concept is widely accepted to model mechanical properties of materials and systems. However, at micro- and nanoscale the microstructure plays an essential role and can not be neglected. At these scales the material can be represented as a discrete medium. This approach gives an opportunity to easily consider the discontinuous effects and determine the properties of the effective elastic media using the knowledge of microparameters such as geometry of crystal lattice and interatomic interaction. Moreover, this approach is also very successful at macroscale to simulate the dynamical processes in cases of material continuality loss.

However, continuum mechanics still is a most convenient instrument for some new tasks of nanomechanics of coupled quasi-stationary magnetic and electric fields. A nano-resonator based on graphene layer suspended over a groove in silicon oxide is considered as an electromechanical system. An electromagnetic field with a high frequency applied to such system generates the mechanical oscillations. As this system is sensitive to change of mass of the oscillating film, it can be used as an ultraprecise mass sensor allowing weighing the single particles. Change of the spectral properties due to a small mass addition is hard to determine experimentally. Due to this, different approaches to improve sensitivity of the nanoresonator are proposed.

A short review of discrete and continuum models at micro- and nanolevel will be presented. Different applications of these models including the simulation of carbon nanostructures and dynamical fracture of materials will be discussed.

EE Seminar: Dr. Tomer Michaeli, Weizmann Institute

~~סמינר מחלקתי
You are invited to attend a lecture by

Dr. Tomer Michaeli
(Weizmann Institute of Science)

on the subject:

Blind deblurring and blind super-resolution
using internal patch recurrence

Small image patches tend to recur at multiple scales within high-quality natural images. This fractal-like behavior has been used in the past for various tasks including image compression, super-resolution and denoising. In this talk, I will show that this phenomenon can also be harnessed for "blind deblurring" and for "blind super-resolution", that is for removing blur or increasing resolution without a-priori knowledge of the associated blur kernel. Our key observation is that the source of the patch recurrence phenomenon is the repetitions of structures at various scales in the continuous scene. Therefore, the way by which this continuous-domain phenomenon is manifested in the discrete image, actually encodes information on the imaging process.
We adopt a continuous-domain model of natural scenes, and use it to study the expected behavior of discrete natural images. This model suggests that the deviations from ideal patch recurrence in a discrete image can be used for recovering the unknown camera blur kernel. More specifically, we show that the correct blur kernel is the one which maximizes the similarity between patches across scales of the image. Extensive experiments indicate that our approach leads to state of the art results, both in deblurring and in super-resolution.

26 בנובמבר 2014, 15:00

בניין כיתות חשמל חדר 011

הדרכה

26 באוקטובר 2014, 10:09

וולפסון 206

בקישור

הדרכת עדכון ארועים

ארוע פתיחת שנה

06 בנובמבר 2014, 13:00

אירוע פתיחת שנת הלימודים תשע"ה לסטודנטים במחזור הראשון לתואר ראשון משולב במדע והנדסה של חומרים ובכימיה ולחברי הסגל. יום חמישי, 6 בנובמבר 2014, שעה 13:00, לובי בניין מעבדות חשמל

סדנא משותפת עם נורת'ווסטרן

22 בפברואר 2015, 9:00

אוניברסיטת תל-אביב, ימים ראשון-רביעי, 22-25 בפברואר 2015

הסדנא המשותפת הראשונה של אוניברסיטת תל-אביב ואוניברסיטת נורת'ווסטרן: מוליכים למחצה, חומרים אלקטרוניים, שכבות דקות וחומרים פוטוניים. ימים ראשון-רביעי, 22-25 בפברואר 2015, אוניברסיטת תל-אביב

אתר הכנס:

תוכנית הכנס

תקצירים לכנס