School of Mechanical Engineering Seminar Wednesday, May 22, 2019 at 14:00
Wolfson Building of Mechanical Engineering, Room 206
Evolving Concepts by On-line Inferiority Estimation and Elimination
Barak Samina
M.Sc. student of Dr. Amiram Moshaiov
This research is motivated by the need to rationalize the allocation of computational resources during a Pareto-based evolution of Set-based Concepts (SBCs). Such an evolution involves the simultaneous exploration of different design spaces, at both the conceptual and particular solution levels, with respect to multiple optimization objectives. Here, several techniques for an on-line estimation of the inferiority of conceptual solutions are suggested and investigated. Such techniques allow on-line intelligent elimination of inferior concepts, which aims at a reduction of the required computational resources.
To evaluate the effectiveness of the proposed techniques, unique concept-based measures and analysis methodology for evolving SBCs are suggested. Extensive statistical study is carried out to assess and compare the suggested techniques. Based on the statistical results, it is concluded that several of the proposed techniques are effective. It is also shown that the effectiveness varies in accordance with the difficulty level of the solved search problems. These conclusions are in accordance with the well-known optimization theorem of no-free lunch.
Estimation Theory with Side Information for Periodic and Constrained Problems
Abstract
In many practical parameter estimation problems some side information regarding the unknown parameters is available. Types of side information that are commonly encountered in signal processing applications include periodicity, parametric equality and inequality constraints, and sparsity. In this research, we address some fundamental topics in estimation theory in the presence of side information. We exploit the side information by choosing proper cost functions, deriving optimal estimation methods, and developing corresponding performance bounds. In the first stage of this work we have investigated the problem of Bayesian parameter estimation in the presence of periodic side information. In periodic parameter estimation, the commonly used mean-squared-error (MSE) risk is inappropriate, since it does not take into account the periodic nature of the problem. We proposed a new class of Bayesian lower bounds on the mean-cyclic-error. This class can be interpreted as the cyclic-equivalent of the Weiss-Weinstein class of MSE lower bounds. This class was extended to provide Bayesian lower bounds for stochastic filtering with periodic side information. In the second stage of this work, we considered non-Bayesian parameter estimation under differentiable equality constraints. For this type of estimation problem, we defined proper unbiasedness in the Lehmann sense and developed new performance bounds that were shown to be more appropriate than the well-known constrained Cramér-Rao bound. In the third stage of this work, we proposed a method for optimal biased estimation. This method is based on combining Lehmann-unbiasedness under a weighted MSE risk and a penalized likelihood approach. It was shown that this method can be useful for parameter estimation under inequality constraints and can lead to estimators that uniformly outperform the minimum variance unbiased estimator and the maximum likelihood estimator.
School of Mechanical Engineering Seminar Monday, April 15, 2019 at 14:00
Wolfson Building of Mechanical Engineering, Room 206
Modeling and Analysis of Biomechanical Interfaces
Dana Solav, PhD
Research Scientist
Massachusetts Institute of Technology
The prosthetic socket constitutes the mechanical interface between the residual limb of amputees and the artificial prosthesis, and is the most critical component of the lower limb prosthetic system. The main challenge is to design a socket that properly distributes the load across the soft tissues, while preventing stress concentrations in vulnerable regions. Conventional socket design is largely artisanal, non-standard, and insufficiently data-driven. A clear necessity in the field is a computational design framework which is automatic, repeatable, data-driven, and based on scientific rationale. In this talk, I will present a novel framework for the computational design of optimized patient-specific sockets. The first and most important step is obtaining a representative biomechanical model of the residual limb, including its accurate shape, deformation, and mechanical properties. This is exceptionally challenging since these measures have to be obtained in-vivo and the soft-tissue material behavior is complex and non-linear. To this end, I have developed a new imaging setup and software based on 3D digital image correlation, which provides accurate full-field measurements of the entire residuum skin. These measurements can be used in combination with data from MRI and a custom indentation device to identify the material parameters of the underlying soft tissues using inverse finite element (FE) analysis. Furthermore, I will demonstrate the capability of a custom software to iteratively execute FE analyses and use the results to inform modifications in the socket design. Finally, I will present future research directions towards optimized design of prosthetic sockets and application to other biomechanical interfaces.
Short Bio
Dana Solav received a BSc in Geophysics from Tel Aviv University (2006) and a PhD in Mechanical Engineering from the Technion Israel Institute of Technology (2016). In January 2017 she joined the MIT Media Lab’s Biomechatronics group as a postdoctoral fellow, where she became a research scientist (2019) to lead the group’s Computational Biomechanics research track. Her research experience includes continuum mechanics, musculoskeletal modeling, soft tissue biomechanics, and imaging methods. Her current research focuses on combined experimental and computational investigation of the mechanical interface between the residual limb of amputees and the prosthetic socket.
School of Mechanical Engineering Seminar Wednesday, March 13, 2019 at 14:00
Wolfson Building of Mechanical Engineering, Room 206
Tumor dormancy and recurrence: theory & experiment for tumor growth, regression & metastasis
Prof. David S. Rumschitzki
Department of Chemical Engineering, The City College of New York
PhD programs in Chemistry and Biology, Graduate School and University Center, City University of New York
Department 0f Medicine (Cardiology), Columbia Presbyterian Medical Center, Columbia University
Collaborators: Adeyinka Lesi (PhD student), Silja Heilmann (Postdoc), Richard M. White (Professor at Memorial Sloan Kettering Cancer Center)
Certain types of cancer including breast cancer and melanoma can recur many years after apparently successful treatment. It is a mystery how tumor cells can remain dormant for many years, avoid eradication by the immune system and then reactivate after many years. We present a population balance model to describe how the size distribution of an ensemble of tumors from many patients evolves in time due to mitosis, cell death and metastasis. A transformation recasts the dynamic interplay between tumor growth and shrinkage in these equations into the form of an advective-diffusion equation in tumor size space. These new equations predict and thus provide a plausible mechanism for tumor dormancy and recurrence for certain relationships amongst the three model parameters. After showing that our model easily fits data sets on tumor size distributions in the literature, we present new, far more refined gender- and immune status-segregated data on the zebrafish melanoma. We show that our model also describes these data very well. Study of gender-segregated cohorts shows gender-dependent parameter only appears in the host-dependent parameter describing tumor shrinkage; it is far more size-dependent in females than in males, which may be relevant for gender differences in human melanoma outcomes. Fortunately for the fish, their model parameters do not predict recurrence over fish lifetimes. However, the model guides our current experiments by instructing us on how to try to perturb fish immunity to bring fish parameters into the range where we should be able to observe fish melanoma dormancy and recurrence. There are claims in the literature that a mouse model of breast cancer can show dormancy and recurrence. We are beginning a collaboration at the Technion Rambam using this mouse model to test our mathematical model and to see we find dormancy and recurrence theoretically, experimentally or both.
תפקידו וייעודו של המהנדס/ת המכני/ת נמצאים בתהליך מתמיד של התחדשות. הכשרת המהנדס/ת המכנית מאפשרת לעסוק במגוון רחב של תפקידים בנושאים הכוללים: מחקר ופיתוח, תכנון, ייצור, ניהול הנדסי ואחזקה. בנוסף לתחומים הקלאסיים של המהנדס/ת המכני, ההכשרה הבסיסית שמקנים הלימודים במגמה להנדסה מכנית מאפשר לבוגרים ולבוגרות לעסוק בקשת רחבה של תחומים, כגון: הנדסה אווירונאוטית, הנדסת אניות ואוקינוגרפיה, הנדסה גרעינית, הנדסה אזרחית, הנדסת חומרים, הנדסת סביבה, הנדסת מכטרוניקה ורובוטיקה, זיווד אלקטרוני, הנדסת מחשבים והנדסה ביו-רפואית.
תחזיות רבות בדבר הטכנולוגיות המובילות בשנים הבאות כוללות בתוכן את אוסף הנושאים הבא:
ננו-טכנולוגיה
רובוטיקה
אנרגיות חלופיות
איכות הסביבה
ביו-הנדסה/ביו-מכניקה
נושאים אלו ואחרים שנלמדים, נחקרים ומפותחים בהנדסה מכנית צפויים להוביל בתרומה לאיכות חיינו, רווחתנו הכלכלית וביטחוננו. נושאים אלו גם משרתים באופן משמעותי את אותן הטכנולוגיות שיותר פופולריות היום כגון מחשבים. לדוגמה, הבעיות הקשורות בהתפתחות מעבדים קטנים ומהירים יותר נוגעות לחומרים, תהליכי ייצור ומעבר חום שהם נושאי הנדסת מכונות קלאסיים. הנדסה מכנית אם כן, היא תחום תשתית לכל הטכנולוגיות האחרות ובעצמה מובילה טכנולוגיות חדשות מלהיבות.
ממה מושפעת תעשיית ההנדסה המכנית?
בעידן המודרני ההנדסה המכנית מושפעת באופן מכריע מהתפתחות אמצעי המחשוב. כלי רכב, מטוסים, כלי שייט, רובוטים, מכונות ייצור, תחנות כוח, מערכות מיזוג אוויר ולעיתים אף מבנים כגון גשרים, סכרים ובניינים חכמים - מופעלים, מבוקרים ומתוחזקים בעזרת מחשב. תהליכים רבים בתעשייה בת זמננו נמצאים תחת בקרה ממוחשבת. חלק גדול מהתיכון והייצור המכני נתמכים כיום בכלים ממוחשבים. במקביל לתכנון המכני הקלאסי מתפתח בשנים האחרונות שטח חדש של תכנון מכטרוני. המכטרוניקה, המשלבת מכניקה, אלקטרוניקה ומחשבים, היא פועל יוצא של מהפיכת המחשוב.
מה הדגש העיקרי בלימודים?
אסטרטגיה של תכניות הלימוד של ביה"ס היא להקנות לסטודנטים כלים בסיסיים אוניברסליים ויכולות לימוד עצמי לאורך זמן, מבוססים על לימודים נרחבים במדעי היסוד כגון מתמטיקה ופיזיקה ועקרונות של מכניקה קלאסית. ביה"ס מציע כיום, יחד עם התכנית בהנדסה מכנית בתואר ראשון, עוד שתי תכניות לתואר כפול - תואר ראשון כפול בהנדסה מכנית ובמדעי כדור הארץ בדגש סביבה. תכניות לתארים מתקדמים מאפשרות לתלמידים להתעמק במחקרים בתחומים של מדעי הנדסה מכנית. תכנית ייחודית למצטיינים לתואר שני ישיר שמאפשרת השלמת התואר בחמש שנים מושכת תלמידים מצטיינים שמשתלבים במחקר של ביה"ס. אנו מאמינים כי תוכניות הלימוד בהנדסה מכנית מספקות את המענה המתבקש לרב-תחומיות תוך כדי שמירה על תחומי ההוראה והמחקר המסורתיים.
מה אפשרויות התעסוקה לבוגרים?
הבוגרים של תכניות אלה משתלבים בהצלחה בתעשייה ובאקדמיה ותורמים לפתרון האתגרים הקשורים בין היתר לצרכים הביטחוניים של מדינת ישראל, ההתפתחות הכלכלית, תגלית הנפט והגז והשינוי האקלימי - כולם מחייבים שילוב הנדסה מכנית לשמירת הביטחון, הסביבה וההתייעלות האנרגטית.
למי זה הכי מתאים?
לימודי הנדסה מכנית מתאימים לכל מעמד ומועמדת בעלי נתוני קבלה נדרשים. התכונות החשובות הן יצירתיות, רצון להתעמק ברמת ההבנה ויכולת לרדת לעומק הדברים, ראיה מערכתית ובינתחומית וגם תפיסה מרחבית.
M.Sc. student under the supervision of Dr. Ofer Shayevitz
Sunday, March 10th, 2019 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering
Distributed Source Simulation With No Communication
Abstract
We consider the problem of distributed source simulation with no communication, in which Alice and Bob observe sequences and respectively, drawn from a joint distribution , and wish to locally generate sequences and respectively with a joint distribution that is close (in KL divergence) to . We provide a new single-letter condition under which such a simulation is asymptotically possible with a vanishing KL divergence. The Gàcs-Körner (GK) common information, which measures the amount of common randomness that can be separately extracted from either marginal of two correlated random variables, plays a crucial part in the problem of distributed source simulation: Our condition is nontrivial only in the case where the Gàcs-Körner (GK) common information between and is nonzero, and we conjecture that only scalar Markov chains can be simulated otherwise.
Motivated by this conjecture, we further examine the case where both and are doubly symmetric binary sources with parameters respectively. This is a private case of zero Gàcs-Körner (GK) common information between and . While it is trivial that in this case is both necessary and sufficient, we show that when is close to then any successful simulation is close to being scalar in the total variation sense.
and 
respectively, drawn from a joint distribution 
, and wish to locally generate sequences 
and 
respectively with a joint distribution that is close (in KL divergence) to 
. We provide a new single-letter condition under which such a simulation is asymptotically possible with a vanishing KL divergence. The Gàcs-Körner (GK) common information, which measures the amount of common randomness that can be separately extracted from either marginal of two correlated random variables, plays a crucial part in the problem of distributed source simulation: Our condition is nontrivial only in the case where the Gàcs-Körner (GK) common information between 
and 
is nonzero, and we conjecture that only scalar Markov chains 
can be simulated otherwise.
and 
are doubly symmetric binary sources with parameters 
respectively. This is a private case of zero Gàcs-Körner (GK) common information between 
is both necessary and sufficient, we show that when 
is close to 
then any successful simulation is close to being scalar in the total variation sense.
