הפקולטה להנדסה ע"ש איבי ואלדר פליישמן

Biomedical seminar

26 במרץ 2017, 15:00

Reverse Engineering Insect Flight

Dr. Gal Ribak

Department of Zoology, Faculty of Life Sciences, Tel Aviv University, Israel

Insect flight is the least understood type of aerial locomotion from both the biological and the physical perspectives. The high flapping frequency and relatively low Reynolds number lead to unique unsteady flow phenomena that insects have evolved to exploit during 350 million years of evolution. In my talk I will describe previous and current work carried out at my laboratory on insect flight biomechanics. By studying flight on all spatial, temporal and ecological levels we aim to reverse engineer insect flight, thus getting a true understanding of how nature’s miniature flyers work. The talk will touch on topics related to visual flight control, control of flight muscles, mechanical implications of body miniaturization and effect of wing flexural stiffness on flight performance and aerodynamics.

ההרצאה תתקיים ביום ראשון 26.03.17, בשעה 15:00

בחדר 315, הבניין הרב תחומי, אוניברסיטת תל אביב

סמינר מחלקתי- אלקטרוניקה פיזיקאלית ויטלי קוזלוב

20 במרץ 2017, 11:00

פקולטה להנדסה ביניין תוכנה קומה 5 חדר 512

סמינר מחלקתי- אלקטרוניקה פיזיקאלית ויטלי קוזלוב

~~You are invited to attend a lecture

ELECTROMAGNETIC SCATTERING GOVERNED BY MOVING SCATTERERS AND MAGNETO-ELECTRIC COUPLING
By:
Vitali Kozlov
M.Sc student under supervision of Dr. Pavel Ginzburg

Abstract
Contemporary commercial numerical tools, such as CST, Lumerical and COMSOL are commonly used to solve various electromagnetic, thermal and mechanical problems, to name a few. The above solutions are mostly designed for static problems – yet our world is not static. Scatterers move all the time, they may break into smaller pieces, collide to form bigger ones or otherwise change their shape or other scattering properties in time. To formulate the equations of EM scattering from such dynamic objects, velocities and accelerations must be included alongside the geometry of the objects when applying the boundary conditions of Maxwell’s equations.
This seminar deals with two topics, the first is the micro-Doppler shifts generated by a rotating wire. The wire is effectively a 1D object which makes it relatively simple to solve analytically, allowing physical insight into the phenomena. Intuitively one might expect that since various parts of the wire move at different speeds, the scattered field would have continuous Doppler shifts. Detailed analysis as well as experiments show that the scattered field has discrete frequencies only, forming a micro Doppler comb where the separation between the peaks is an integer multiple of the rotational frequency. The second part of the seminar is related to metamaterials, which have effective characteristics that are interesting for numerous applications in optics and RF, spanning from cloaking, solar cells, optical diodes and more. A hybrid magneto electric meta-atom is presented, made out of a straight wire and a split ring resonator with a unique property of asymmetric backscattering. This means that the meta-atom scatters radiation backwards differently depending on the direction of the incident field. Meta-atoms are the building blocks of metamaterials and it is expected that such meta-atoms could form new and exciting metamaterials with interesting properties.

On Monday, March 20, 2017, 11:00
Room 512, floor 5, Tohna building

EE Seminar: Active Nearest-Neighbor Learning in Metric Spaces

(The talk will be given in English)

Speaker: Dr. Sivan Sabato
Department of Computer Science, Ben Gurion University

Monday, May 22^nd, 2017
15:00 - 16:00

Room 011, Kitot Bldg., Faculty of Engineering

Active Nearest-Neighbor Learning in Metric Spaces

Abstract

In this talk I address the challenges of active learning, an interactive machine learning paradigm that allows using data sources more effectively when data is expensive.
I will present an approach for active learning in general metric spaces, that does not require any input parameters, and obtains a competitive accuracy, compared to the accuracy of a non-interactive algorithm which gets all the data for free.
We prove that the proposed active algorithm can reduce the data costs significantly, and that this type of reduction cannot be achieved using simple subsampling.

Based on joint work with Aryeh Kontorovich and Ruth Urner

22 במאי 2017, 15:00

חדר 011, בניין כיתות-חשמל

EE Seminar: Decoding Linear Codes With Deep Learning

Speaker: Eliya Nachmani,

M.Sc. student under the supervision of Profs. David Burshtein and Yair Be'ery

Wednesday, April 26^th, 2017 at 15:00

Room 011, Kitot Bldg., Faculty of Engineering

Decoding Linear Codes With Deep Learning

Abstract

Designing a practical, low complexity, close to optimal, channel decoder for powerful algebraic codes with short to moderate block length is an open research problem. A novel deep learning method for improving the belief propagation algorithm is proposed with feed-forward neural network. The method generalizes the standard belief propagation algorithm by assigning weights to the edges of the Tanner graph. These edges are then trained using deep learning techniques. A well-known property of the belief propagation algorithm is the independence of the performance on the transmitted codeword. A crucial property of our new method is that our decoder preserved this property. Furthermore, this property allows us to learn only a single codeword instead of exponential number of code-words. Improvements over the belief propagation algorithm are demonstrated for various high density parity check codes.

We also introduce a recurrent neural network architecture for decoding linear block codes. Our method shows comparable bit error rate results compared to the feed-forward neural network with significantly less parameters. We demonstrate improved performance over belief propagation on sparser Tanner graph representations of the codes. Furthermore, we demonstrate that the RNN decoder can be used to improve the performance or alternatively reduce the computational complexity of the mRRD algorithm for low complexity, close to optimal, decoding of short BCH codes.

26 באפריל 2017, 15:00

חדר 011, בניין כיתות-חשמל

School of Mechanical Engineering Ittai Shamir and Lior Shig

10 במאי 2017, 14:00 - 15:00

בניין וולפסון חדר 206

ללא תשלום

School of Mechanical Engineering Ittai Shamir and Lior Shig

School of Mechanical Engineering Seminar
Wednesday, May 10, 2017 at 14:00
Wolfson Building of Mechanical Engineering, Room 206

Investigating the morphometry of cortical layers
in the human brain

Ittai Shamir

MSc. Student of Professor Yaniv Assaf

and Professor Yair Shokef

Several Magnetic Resonance Imaging (MRI) approaches were demonstrated in recent years for visualization of the layer arrangement in the cortex either by using high resolution at high magnetic field or by investigating contrast mechanisms that separate the layers. Still, despite these technological achievements, accurate estimation of whole brain layer composition remains limited due to partial volume effects, since some of the layers are far beyond the image resolution.

In this study, we use a low resolution EPI inversion recovery (IR) scan protocol that provides fast acquisition (~12 minutes) and enables extraction of multiple T1 relaxation time components per voxel. These T1 components are assigned to brain tissue types and the assignment is utilized to extract the sub voxel composition of each of the cortical layers. Spatial analysis is conducted using our unique algorithm, based on sampling a network of spherical volumes dispersed throughout the entire cortical space.

Our suggested protocol offers a simple and accurate method for analysis of cortical layer composition, resolving partial volume effects. While direct visualization of the layers is not possible using our approach, it does offer a robust and powerful tool for investigating the layers and their role in cognition. Using this methodology, studies on the role of cortical thickness in brain function and behavior can be expanded to the cortical layer level, thus providing a new level of detail regarding cortical structure.

School of Mechanical Engineering Seminar
Wednesday, May 10, 2017 at 14:00
Wolfson Building of Mechanical Engineering, Room 206

Active Flow Control of Submerged 3D Air Inlet

Lior Shig

MSc Student of Prof. Avi Seifert

Air ingesting propulsion systems that utilize the oxygen of the atmosphere for combustion require that the air transition with minimum pressure losses from the free stream to the engine's entrance. Small losses in internal flow given the large quantities of air required by jet engines cause serious decrease in the thrust and increase in fuel consumption.

Submerged inlets, are widely used in aircraft and cars as a low drag source of external flow for air conditioning, ventilation and cooling systems. The design criteria of these intakes was established during the 1940’s and 50’s. However due to boundary layer ingestion and distortion are not currently used for critical systems.

Flow analysis and boundary layer control methods were applied to a given submerged inlet design that is efficient from the external aerodynamic point of view. The approach for studying the effectiveness of active flow control methods on this kind of inlet was to significantly shorten the original inlet length, which caused internal flow separations and significant engine face distortions. For simplicity, and as an enabling step, low Reynolds number and low Mach number experiments were performed. The preliminary stages of this study included analysis of the baseline flow features of the chosen original inlet design and shortened version. Passive and active flow control techniques were applied to the short inlet in various locations in order to test the sensitivity of the flow to actuation and try to improve the Aerodynamic Interface Plane flow characteristics. Performance parameters such as volumetric flow rate, pressure recovery and flow distortion were defined and calculated in order to evaluate and quantify the level of success. For each of the applied techniques, the effects on the internal flow and integral parameters were measured and analyzed.

‏

14 מרץ 2017

פרופ' יוסי רוזנוקס בהרצאה מרתקת "להיות הדבר הבא"

ידיעות נוספות בנושא

EE Seminar: Signal Modeling: From Convolutional Sparse Coding to Convolutional Neural Networks

(The talk will be given in English)

Speakers: Vardan Papyan & Yaniv Romano
Computer Sciences Faculty, Technion

Monday, April 3^rd, 2017
15:00 - 16:00

Room 011, Kitot Bldg., Faculty of Engineering

Signal Modeling: From Convolutional Sparse Coding to Convolutional Neural Networks

Abstract

Within the wide field of sparse approximation, convolutional sparse coding (CSC) has gained increasing attention in recent years. This model assumes a structured-dictionary built as a union of banded Circulant matrices. Most attention has been devoted to the practical side of CSC, proposing efficient algorithms for the pursuit problem, and identifying applications that benefit from this model. Interestingly, a systematic theoretical understanding of CSC seems to have been left aside, with the assumption that the existing classical results are sufficient.

In this talk we start by presenting a novel analysis of the CSC model and its associated pursuit. Our study is based on the observation that while being global, this model can be characterized and analyzed locally. We show that uniqueness of the representation, its stability with respect to noise, and successful greedy or convex recovery are all guaranteed assuming that the underlying representation is locally sparse. These new results are much stronger and informative, compared to those obtained by deploying the classical sparse theory.

Armed with these new insights, we proceed by proposing a multi-layer extension of this model, ML-CSC, in which signals are assumed to emerge from a cascade of CSC layers. This, in turn, is shown to be tightly connected to Convolutional Neural Networks (CNN), so much so that the forward-pass of the CNN is in fact the Thresholding pursuit serving the ML-CSC model. This connection brings a fresh view to CNN, as we are able to attribute to this architecture theoretical claims such as uniqueness of the representations throughout the network, and their stable estimation, all guaranteed under simple local sparsity conditions. Lastly, identifying the weaknesses in the above scheme, we propose an alternative to the forward-pass algorithm, which is both tightly connected to deconvolutional and recurrent neural networks, and has better theoretical guarantees.

03 באפריל 2017, 15:00

חדר 011, בניין כיתות-חשמל

14 מרץ 2017

מלגת הנשיא לפוסט דוקטורנטית לנועה מתוקי, שסיימה לאחרונה דוקטורט

ידיעות נוספות בנושא

Biomedical Engineering department seminar

19 במרץ 2017, 14:30

Speaker: Ariel Birenbaum,

School of Electrical Engineering

Multi-View Longitudinal CNN for Multiple Sclerosis Lesion Segmentation

Abstract

This work presents a deep-learning based automated method for Multiple Sclerosis (MS) lesion segmentation. Automatic segmentation of MS lesions is a challenging task due to their variability in shape, size, location and texture in brain Magnetic Resonance (MR) images.

In the proposed scheme, MR intensities and White Matter priors are used to extract candidate lesion voxels, following which Convolutional Neural Networks are utilized for false positive reduction and to obtain the final segmentation result.

The proposed networks use longitudinal data, a novel contribution in the domain of MS lesion analysis.

The method obtained state-of-the-art results on the 2015 Longitudinal MS Lesion Segmentation Challenge dataset, and achieved a performance level equivalent to a trained human rater. Automatic segmentation methods, such as the one proposed, once proven in accuracy and robustness, can help diagnosis and patient follow-up while reducing the time consuming need of manual segmentation.

העבודה נעשתה בהנחיית פרופ' חיית גרינשפן מהמחלקה להנדסה ביו-רפואית,

אוניברסיטת תל-אביב

ההרצאה תתקיים ביום ראשון 19.03.17, בשעה 14:30

בחדר 315, הבניין הרב תחומי, אוניברסיטת תל אביב

EE Seminar: Multi-objective Topology and Weight Evolution of Neuro-controllers and its Application for Robot Navigation

Speaker: Omer Abramovich

M.Sc. student under the supervision of Prof. Emilia Friedman & Dr. Amiram Moshaiov

Wednesday, March 22^nd 2017 at 15:30

Room 011, Kitot Bldg., Faculty of Engineering

Multi-objective Topology and Weight Evolution of Neuro-controllers and its Application for Robot Navigation

Abstract

In the last decades, there has been a substantial increase of interest in the field of evolutionary neural networks with an emphasis on evolutionary robotics. Neuroevolution combines the adaptation power of neural networks with the advantages of evolutionary search to find networks capable of solving highly complex tasks.

Evolving neuro-controllers for robot applications might be a hard challenge for standard evolutionary search techniques. Neuroevolution aims to provide some tailored algorithms to such problems. One example is NEAT, which is an evolutionary algorithm to find not only the optimal connection weights, but also the optimal topology of a network. However, such TWEANN (Topology and Weight Evolution of Artificial Neural-Networks) algorithms are commonly restricted to single objective evolution, and their adjustment for multi-objective problems is in its infancy stage.

In this thesis, we propose a novel fully automated approach to evolve topologies and weights of neuro-controllers for multi-objective problems and in particular for navigation application. The suggested approach dismisses any prerequisites of expert knowledge regarding the network topology and enables to achieve better performances than expert designed neuro-controllers. A key component of this study is the proposed modification to NEAT, which results in an innovative generic multi-objective TWEANN algorithm. The suggested algorithm is proven to outperform state-of-the-art multi-objective TWEANNs in the tested cases of this study.

22 במרץ 2017, 15:30

חדר 011, בניין כיתות-חשמל

Reverse Engineering Insect Flight

Dr. Gal Ribak

בחסות הפקולטה להנדסה אוניברסיטת תל-אביב

ידיעות נוספות בנושא

לנועה מתוקי, שסיימה לאחרונה דוקטורט

ידיעות נוספות בנושא

עמודים