Ori Zitzer - Multi-Modal Multi-Objective Evolutionary Optimization with Solutions of Variable Length

סמינר מחלקת מערכות - EE Systems Seminar

10 בינואר 2024, 14:00 
זום  
 Ori Zitzer - Multi-Modal Multi-Objective Evolutionary Optimization with Solutions of Variable Length

Electrical Engineering Systems Zoom Seminar

 

Join Zoom Meeting

https://zoom.us/j/94236001944?pwd=RmI1TWZlWXo5cHdVd2tSZnRWRWdTdz09
Meeting ID: 942 3600 1944
Passcode: 6kzM4R

 

Speaker: Ori Zitzer

M.Sc. student under the supervision of Dr. Amiram Moshaiov

Wednesday, 10th January 2024, at 14:00

Multi-Modal Multi-Objective Evolutionary Optimization with Solutions of Variable Length

Abstract

Multi-modal optimization aims to provide decision-makers with alternative solutions, possibly near optimal, and not just one optimal solution. In the past few years there is a need to solve a special kind of multi-modal multi objective optimization problems (MMOPs) in which solutions belong to decision-spaces of various dimensions (i.e., solutions of variable length). We propose a new evolutionary algorithm to solve such problems and save computing resources in compare to the existing algorithms.

 

השתתפות בסמינר תיתן קרדיט שמיעה = עפ"י רישום בצ'אט של שם מלא + מספר ת.ז.

 

 

ASIC Digital IP Infrastructure student

Qualifications

  • 3rd year Computer/ Electrical Engineering student.
  • Knowledge of Unix env, csh/sh/bash, Python, perl - advantage.

סמינר מחלקה של איתי גריניאסטי - פונקציונליות מורכבת מתהווה במכונות מיקרוסקופיות ובמודלים חישוביים

15 בינואר 2024, 14:00 - 15:00 
פקולטה להנדסה  
0
סמינר מחלקה של איתי גריניאסטי - פונקציונליות מורכבת מתהווה במכונות מיקרוסקופיות ובמודלים חישוביים

 

SCHOOL OF MECHANICAL ENGINEERING SEMINAR
Monday January 15.1.2024 at 14:00

Wolfson Building of Mechanical Engineering, Room 206

 

Emergent complex functionality in microscopic machines and computational models

Itay Griniasty

Itay Griniasty is a Schmidt AI in science postdoc fellow at Cornell university

 

Systems composed of many interacting elements that collaboratively generate a function, such as meta-material robots, proteins, and neural networks are notoriously difficult to design.

Such systems elude traditional explicit design methodologies, which rely on composing individual components with specific subfunctions, such as cogs, springs and shafts, to achieve complex functionality. In part the problem stems from the fact that there are few principled approaches to the design of emergent functionality.  In this talk I will describe progress towards creating such paradigms for two canonical systems: I will first describe how bifurcations of the system dynamics can be used as an organizing principle for the design of functionality in protein like machines with magnetic interactions. I will then introduce a computational microscope that we have developed to analyze emergent functionality, and its application to machine learning. There we uncovered compelling evidence that the training of neural networks is inherently low dimensional, suggesting new paradigms for their design.

References

1. T. Yang et al. Bifurcation instructed design of multistate machines. Proceedings of the National Academy of Sciences, 120(34):e2300081120, 2023

2. J. Mao et al. The training process of many deep networks explores the same low-dimensional manifold. arXiv preprint arXiv:2305.01604, 2023.

3. R. Ramesh, et al. A picture of the space of typical learnable tasks. Proc. of International Conference of Machine Learning (ICML), 2023.

A diagram of a geodesic system

Description automatically generated

 

 

Short bio

Itay Griniasty is a Schmidt AI in science postdoc fellow at Cornell university, studying the design of microscopic and soft machines, non newtonian fluids and computational tools to analyze deep neural networks and multiphysics simulations.

 

Itay was trained as a mechanical designer in the technological unit in the IDF intelligence corps.

He studied mathematics and physics at the Hebrew university for his BSc, where his minor thesis led to a long collaboration on developing novel mathematical tools for the integration of non linear partial differential equations. He went on to a PhD in physics at the Weizmann institute, studying the propagation of waves in inhomogeneous media.

 

Itay has been awarded an Amirim merit scholarship for his BSc, an Azrieli excellence scholarhship for his PhD,  the Chaim Mida Prize for an excellent PhD student, a Fulbright postdoctoral fellowship (which he declined) and a Schmidt AI in science fellowship towards his postdoc

 

Dr. Tomer Galanti (MIT) - Fundamental Problems in AI: Transferability, Compressibility and Generalization

סמינר מחלקת מערכות - EE Systems Seminar  

08 בינואר 2024, 15:00 
זום  
  Dr. Tomer Galanti (MIT) - Fundamental Problems in AI: Transferability, Compressibility and Generalization

(The talk will be given in English)

Speaker:     Dr. Tomer Galanti

Postdoctoral Associate at the Center for Brains, Minds, and Machines at MIT

011 hall, Electrical Engineering-Kitot Building

 

Monday, January 8, 2024

15:00 - 16:00

 

Fundamental Problems in AI: Transferability, Compressibility and Generalization

 

Abstract

 

In this talk, we delve into several fundamental questions in deep learning. We start by addressing the question, "What are good representations of data?" Recent studies have shown that the representations learned by a single classifier over multiple classes can be easily adapted to new classes with very few samples. We offer a compelling explanation for this behavior by drawing a relationship between transferability and an emergent property known as neural collapse. Additionally, we explore why certain architectures, such as convolutional networks, outperform fully-connected networks, providing theoretical support for how their inherent sparsity aids learning with fewer samples. Lastly, I present recent findings on how training hyperparameters implicitly control the ranks of weight matrices, consequently affecting the model's compressibility and the dimensionality of the learned features.
 
Additionally, I will describe how this research integrates into a broader research program where I aim to develop realistic models of contemporary learning settings to guide practices in deep learning and artificial intelligence. Utilizing both theory and experiments, I study fundamental questions in the field of deep learning, including why certain architectural choices improve performance or convergence rates, when transfer learning and self-supervised learning work, and what kinds of data representations are learned with Stochastic Gradient Descent.
 
Short Bio
 
Tomer Galanti is a Postdoctoral Associate at the Center for Brains, Minds, and Machines at MIT, where he focuses on the theoretical and algorithmic aspects of deep learning. He received his Ph.D. in Computer Science from Tel Aviv University and served as a Research Scientist Intern at Google DeepMind's Foundations team during his doctoral studies. He has published numerous papers in top-tier conferences and journals, including NeurIPS, ICML, ICLR, and JMLR. His work, titled "On the Modularity of Hypernetworks," was awarded an oral presentation at NeurIPS 2020.
 
Zoom Link:
 

 

סמינר מחלקה של אנדי טוואקו - להבות קרירות וטכנולוגיות בעירה חדשות

12 בפברואר 2024, 14:00 - 15:00 
פקולטה להנדסה  
0
סמינר מחלקה של אנדי טוואקו - להבות קרירות וטכנולוגיות בעירה חדשות

 

SCHOOL OF MECHANICAL ENGINEERING SEMINAR
Monday February 12.2.2024 at 14:00

Wolfson Building of Mechanical Engineering, Room 206

 

Towards an Environmentally Sustainable Future: Cool flames and Novel Combustion Technologies

 

Andy Thawko, Ph.D.

Postdoctoral Research Fellow, Mechanical and Aerospace Engineering Department,

Princeton University, USA

Email: andyth@princeton.edu

 

As global consensus on the critical need to mitigate greenhouse gas emissions and combat anthropogenic climate change grows, there is an urgent imperative to study the fundamentals of low-temperature combustion. This research is essential not only to improve the thermal efficiency of systems in the energy and transportation sectors but also to pave the way for the development of innovative technologies grounded in low-carbon and carbon-neutral fuels. While high-temperature combustion and hot flames have been extensively studied for decades, a new frontier in combustion science has emerged—low-temperature combustion and cool flames, with only a few research groups worldwide actively investigating this field. Our research on high-pressure cool flames led to the discovery of a new pressure-dependent relation for the cool flame heat release rate. This finding, distinct from the well-established pressure-independent relation of hot flames, emphasizes the profound influence of pressure on cool flames. Furthermore, I will introduce the radical index theory for high-pressure cool flames, offering a quantitative measure of the low-temperature reactivity of fuels. This measure serves to assess the suitability of existing or newly synthesized fuels for advanced propulsion technologies based on low-temperature combustion. Finally, I will present a new understanding of the kinetic enhancement effect in the deflagration to detonation transition (DDT), allowing acceleration of the shock-ignition coupling and the detonation transition. The insights gained from this research are significant to further develop new methods based on ignition enhancers such as plasma-assisted DDT because DDT acceleration is crucial for eliminating detonation stability and reducing heat losses, leading to improved combustion efficiency and enhanced thermodynamic cycles by up to 30%. This seminar aims to shed light on the pivotal role of low-temperature combustion in the ongoing global effort to address climate change. Through our research, we contribute valuable insights that have implications for both fundamental combustion science and the practical development of environmentally sustainable technologies.

 

 

Andy Thawko has been Postdoctoral Research Fellow at Princeton University since 2022. His research interests include thermofluids and combustion science with a focus on low-temperature combustion, energy decarbonization, and kinetic enhancement of processes. Andy completed his Ph.D. in 2021 at the Grand Technion Energy Program, conducting research in the Faculty of Mechanical Engineering. Prior to this, he earned his M.E. in Energy Engineering in 2013, and his B.S. in Mechanical Engineering in 2009, both from the Technion.

סמינר מחלקה של פרופ' אולג גנדלמן

20 במאי 2024, 14:00 - 15:00 
פקולטה להנדסה  
0
סמינר מחלקה של פרופ' אולג גנדלמן

פרטים יפורסמו בהמשך

סמינר מחלקה של בני בר און - איך בונים "גשר"? האסטרטגיה של הטבע לחיבור חומרים קשים ורכים

18 במרץ 2024, 14:00 - 15:00 
פקולטה להנדסה  
0
סמינר מחלקה של בני בר און - איך בונים "גשר"? האסטרטגיה של הטבע לחיבור חומרים קשים ורכים

 

 

Monday 18.03.2024 at 14:00

Wolfson Building of Mechanical Engineering, Room 206

 

How to build a "bridge"?

Nature's strategy for connecting hard and soft materials

 

Benny Bar-On

Department of Mechanical Engineering, Ben-Gurion University of the Negev, Israel

 

Load-bearing biological materials employ specialized bridging regions to connect material parts with substantially different mechanical properties (hard vs. soft). While such bridging regions have been extensively observed in diverse biomaterial systems that evolved through distinctive evolutionary paths—including arthropod parts, dental tissues, and marine threads—their mechanical origins and functional roles remain vague.

In my talk, I introduce a hypothesis that these bridging regions have primarily formed to minimize the near-interface stress effects between the connected material parts, preventing their splitting failure, and obtain a simple theoretical law for the optimal mechanical properties of such bridging regions. I demonstrate this principle through Finite Element simulations and physical experiments on a model synthetic-material system and verify its predictability for different biomaterial systems. The bridging principles of biological materials can be implemented into advanced material designs—paving the way to new forms of architected materials and composite structures with extreme load-bearing capabilities.

 

תמונה שמכילה טקסט, צילום מסך, עיצוב גרפי, כחול חשמלי

התיאור נוצר באופן אוטומטי 

 

Biography: Benny Bar-on is a professor at the Department of Mechanical Engineering at the Ben-Gurion University of the Negev. He received his Ph.D. in Mechanical Engineering from the Technion and was a postdoctoral fellow at the Weizmann Institute of Science and the Max-Planck Institute of Colloids and Interfaces. Prof. Bar-On's research aims to identify structural–mechanical relationships in load-bearing biological materials, including plant organs, mineralized tissues, and arthropod cuticles.

עמודים

אוניברסיטת תל אביב עושה כל מאמץ לכבד זכויות יוצרים. אם בבעלותך זכויות יוצרים בתכנים שנמצאים פה ו/או השימוש שנעשה בתכנים אלה לדעתך מפר זכויות
שנעשה בתכנים אלה לדעתך מפר זכויות נא לפנות בהקדם לכתובת שכאן >>