Similarity matching: a bioinspired approach to machine learning
Speaker: Dmitri Chklovskii, Simons Foundation, USA
Date and Time: 3:00-4:00, 12/3 (Thu)
Abstract
Inspired by experimental neuroscience results we developed a family of online algorithms that reduce dimensionality, cluster and discover features in streaming data. The novelty of our approach is in starting with similarity matching objective functions used offline in Multidimensional Scaling and Symmetric Nonnegative Matrix Factorization. We derived online distributed algorithms that can be implemented by biological neural networks resembling brain circuits. Such algorithms may also be used for Big Data applications.
Short bio:
Dmitri “Mitya” Chklovskii is a Group Leader for Neuroscience at the Simons Center for Data Analysis. His research is aimed at reverse engineering the brain by reconstructing comprehensive maps of neural connections called connectomes and developing an algorithmic theory of neural computation using online machine learning. He studied physics and engineering in St. Petersburg, Russia and holds a PhD from MIT. After being a Junior Fellow at the Harvard Society of Fellows he made a transition to theoretical neuroscience and was a Sloan Fellow at the Salk Institute. From 1999 to 2007 he served first as an Assistant, and later an Associate Professor at Cold Spring Harbor Laboratory. From 2007 to 2014 he was a Group Leader at the Howard Hughes Medical Institute's Janelia Farm Research Campus where he initiated and led a collaborative project that assembled the largest-ever connectome.
A Computational Modeling Platform for Biomedical Systems
Speaker: Adriana Compagnoni, Stevens Institute of Technology, USA
Date and Time: 2:00-3:00, 12/3 (Thu)
Abstract
The 2013 Chemistry Nobel Prize winners, Karplus, Levitt, and Warshel, were instrumental in constructing the first computational models able to predict the effects of chemical reactions, by combining classical Newtonian physics and quantum physics. Their contribution was a dramatic demonstration of the long-term trend of computational models becoming an essential tool in engineering and science. These models are used to predict the weather and climate change, estimate drug doses in laboratory experiments, anticipate the behavior of the stock market, and represent signal transduction pathways, to name a few examples. They are part of everyday life from recreation to medicine, from finances to education.
Despite its broad spectrum of applications and the vast number of computational models built, model construction remains primarily an ad-hoc activity, where an expert or small team creates a piece of software describing the behavior of the agents partaking in a specific real world scenario being represented.
An alternative scenario consists of a modeling platform based on a specification formalism where the user builds a model by describing agents' behavior and attributes.
In this talk, I will share our experience in developing a computational modeling platform, its application to simulating antibacterial surfaces, and the fascinating road ahead.
Short bio:
Adriana Compagnoni is a faculty member at Stevens Institute of Technology and holds a PhD in Computer Science from the Netherlands, where she studied under the supervision of Henk Barendregt and Mariangiola Dezani. She was the recipient of two Marie Curie European Fellowships (TMR and EUROFOCS), before joining the department of Computer Science at Stevens Institute of Technology, where she is now an Associate Professor.
Soon after joining Stevens, Adriana received an NSF-CAREER award for her work on language-based security. Her successful research career in cybersecurity and her intellectual curiosity prompted her to explore new ground in computational biology, transferring programming languages technology from one field to the next. Her group developed BioScape, a modeling language for the stochastic simulation of biological processes, currently being used in the modeling of antibacterial surfaces for medical implants and biofilm-related infections, as well as adenovirus traffic.
Living Architectures: Autonomous Self Assembly and Disassembly in Army Ants
Speaker: Simon Garnier, New Jersey Institute of Technology and Rutgers University, USA
Date and Time: 8:00-9:00, 12/5 (Sat)
Abstract
One of the most spectacular examples of construction by social insects are the self-assembling structures formed by New World army ants. In order to conquer the rough and complex terrain of the tropical forests of Central and South America, these nomadic ants create temporary support structures with their own bodies – bridges, pothole covers, and buttresses – forming the backbone of dynamical ant “superhighways”. In particular, bridges formed by the army ants can self-assemble across a wide variety of environments and spanning conditions, reaching spans of over 12 cm (or 12 ant body lengths), and have been shown to recover from damage, adapt their size according to traffic, and even spontaneously disassemble when under-used. The army ants’ living architectures are an existence proof of how complex and dynamical biological structures can be achieved from the cooperation of large numbers of limited individuals.
Over the last few years, such natural systems have inspired the development of a new kind of robotics, where simple, independent agents act together to build large-scale structures as needed, guided only by their reactions to the local situations they encounter. Large robotic swarms that could self-assemble could accomplish remarkable tasks, such as creating bridges to navigate complex terrain, plugs to repair structural breaches, or supports to stabilize a failing structure. Nevertheless, how to achieve complex artificial self-assemblages remains poorly understood. During this talk I will review our latest discoveries on - and current investigations in - the mechanisms of construction in New World army ants, with the goal to provide insight into achieving successful self-assembly in artificial systems.
Short bio:
Simon Garnier is an Assistant Professor in the Federated Department of Biology at the New Jersey Institute of Technology (NJIT). He obtained a PhD from the University of Toulouse (France) under the direction of Dr. Guy Theraulaz and performed his his postdoctoral work with Professor Iain Couzin at Princeton University. Simon is now the head of the Swarm Lab, an interdisciplinary research lab that studies the mechanisms underlying Collective Behaviors and Swarm Intelligence in natural and artificial systems. The Swarm Lab started to operate in July 2012 and its research aims to reveal the detailed functioning of collective intelligence in systems as diverse as ant colonies, human crowds or robotic swarms. The Swarm Lab focuses in particular on the mechanisms of information transfer and integration in large groups that can lead to adaptive (or “intelligent”) collective responses to environmental challenges.
The Topological Field Theory of Data: a program towards a novel strategy for mining data through data language
Speaker: Emanuela Merelli, University of Camerino, Italy
Date and Time: 3:00-4:00, 12/4 (Fri)
Abstract
This talk aims to challenge the current thinking in IT for the 'Big Data' question, proposing a program to construct an innovative methodology to perform data analytics in a way that returns an automaton as a recognizer of the data language: a Field Theory of Data. I will describe the aim steps for building, directly out of probing data space, a theoretical framework enabling us to extract the manifold hidden relations (patterns) that exist among data, as correlations depending on the semantics generated by the mining context. The program, that is grounded in the recent innovative ways of integrating data into a topological setting, proposes the realization of a 'Topological Field Theory of Data', transferring and generalizing to the space of data notions inspired by physical (topological) field theories and harnesses the theory of formal languages to define the potential semantics necessary to understand the emerging patterns. The immune system modelling will be the running example.
Short bio:
Emanuela Merelli, PhD in Artificial Intelligent Systems, is associate professor (with qualification as full professor) and coordinator of the Doctoral program in Computer Science at the University of Camerino. She is the coordinator of the TOPDRIM FP7-FET project and has been unit coordinator of LITBIO (Laboratory for Interdisciplinary Technologies in Bioinformatics) Italian FIRB project. Her research interests are formal methods and computational biology. Currently, her research concerns the study of advanced formal methods driven by algebraic and computational topology. She published many papers with interdisciplinary character in refereed international journals and continuously involved in the organization of interdisciplinary events.
The Internet of the Future: From Codons to Coding
Speaker: Bud Mishra, New York University, USA
Date and Time: 8:00-9:00, 12/4 (Fri)
Abstract
This talk will focus on a game theoretic model of the Internet so as to devise new mechanisms to improve its security, while paying proportional attention to various intertwined issues: namely in the form of deception, privacy, trust, economic utilities and stability. For this purpose, we primarily rely on a realistic formulation of classical information-asymmetric signaling games, in a repeated form, while allowing the agents to dynamically vary their utility functions. To better understand the multi-faceted security concerns in existing and emerging multi-agent interactions over the Internet, we will map, model and analyze various challenging examples of security concerns: namely, those occurring in Search, Ad-exchange, Data Markets, Medical and Finance systems. We also describe a bridge to the future by investigating the extendability of the proposed mechanisms in a specific embodiment, for example, as a simple browser, supported by efficient virtualization technology and with the host operating systems housed in a cloud. New malware that exploit virtualization technology (e.g., BluePills) is of particular interest and relevant in this context, and will be abstracted in a simple game: PWN IT, a novel model of combined repeated information-asymmetric games, first introduced in 2014, at CMU/SEI summer program on games and security, as a generalized FLIP IT game.
The talk will build on our earlier experience in the areas of systems biology (evolutionary models), game theory, data science, model checking, causality analysis, cyber security, insider threat, virtualization and data markets.
Short bio:
Professor Bud Mishra is a professor of computer science and mathematics at NYU's Courant Institute of Mathematical Sciences, professor of human genetics at Mt. Sinai School of Medicine, and a professor of cell biology at NYU School of Medicine. He founded the NYU/Courant Bioinformatics Group, a multi-disciplinary group working on research at the interface of computer science, applied mathematics, biology, biomedicine and bio/nano-technologies. Prof. Mishra has a degree in Physics from Utkal University, in Electronics and Communication Engineering from IIT, Kharagpur, and MS and PhD degrees in Computer Science from Carnegie-Mellon University. He has industrial experience in Computer and Data Science (Genesis Media,Tartan Laboratories, and ATTAP), Finance (Instadat, Tudor Investment and PRF, LLC), Robotics and Bio- and Nanotechnologies (InSilico, Seqster, Abraxis, OpGen, and Bioarrays). He is the author of a textbook on algorithmic algebra and more than two hundred archived publications. He has advised and mentored more than 35 graduate students and post-docs in the areas of computer science, robotics and control engineering, applied mathematics, finance, biology and medicine. He is an inventor of Optical Mapping and Sequencing (SMASH), Array Mapping, Copy-Number Variation Mapping, Model Checker for circuit verification, Robot Grasping and Fixturing devices and algorithms, Reactive Robotics, and Nanotechnology for DNA profiling. He is a fellow of IEEE, ACM and AAAS, a Distinguished Alumnus of IIT-Kgp, and a NYSTAR Distinguished Professor. From 2003-2006, he held adjunct professorship at Tata Institute of Fundamental Research in Mumbai, India. From 2001-04, he was a professor at the Watson School of Biological Sciences, Cold Spring Harbor Lab; currently he is a QB visiting scholar at Cold Spring Harbor Lab.
Adaptive dynamics for shape optimization inspired by the use-and-growth rule in a simple organism of slime mold
Speaker: Toshiyuki Nakagaki, Hokkaido University, Japan
Date and Time: 2:00-3:00, 12/4 (Fri)
Abstract
A kind of huge amoeboid organism named Physarum plasmodium constructs a intricate network of veins for circulating nutrients and signals over the entire body. The networkshape (topology of connectivity and sequence of branching invein network, for instance) is drastically reorganised within an hour in response to external conditions. The past studies showed that the network shape was optimized to maximise possibility of survival, in some sense. So we may extract an algorithm for optimal design of functional network from the primitive organism. The key thing to design is adaptive dynamics of current-reinforcement rule: each vein of network becomes thicker when current is large enough through the vein itself, while it becomes thinner and dies out otherwise. Based on this simple rule, functions and formation of transport network in Physarum is analyzed. We will show that the rule is applicable to the other bio-systems: (1) social dynamics of public transportation, (2) formation of network structure in porous tissues bone (bone remodeling in other words), (3) fibrous tissue of plants and fungi. A tractable perspective to think similarly of a variety of bio-network is given from the viewpoint of current-reinforcement rule.
Short bio:
Toshiyuki Nakagaki graduated from Department of Pharmaceutical Science, Hokkaido University (bachelor, 1987), and obtained Ph.D. (Biophysics) in 1997 from Nagoya University after working for five years in Central Research Center Nagoya, Pfizer Inc. In 2013, he got the current position after the following experiences: a post-doc researcher in RIKEN (1997-2000), an associate professor in Hokkaido University (2000-2010), and a professor in Department of Complex and Intelligent Systems, Future University Hakodate (2010-2013). He is an Adjunct Professor in Graduate School of Frontier BioScience, Osaka University (2010-). His research interests are focused on information processing in cells, physical ethology, active soft matter, bio-inspired algorithms and protozoology.
Dynamic Alignment and Millimeter-scale Vortex Formation of Microtubules Driven by Different Types of Dynein
Speaker: Kazuhiro Oiwa, National Institute of Information and Communication Technology, Japan
Date and Time: 2:00-3:00, 12/5 (Sat)
Abstract
Experimental systems have long been demanded for the study of collective motion often observed in biology (a flock of birds, a shoal of fish, cell migrations during development etc). In vitro motility assays commonly used in biophysical studies on protein-motors now fulfill the demand described above. Using the in vitro motility assays, we report collective motion and vortex emergence of microtubules (MTs) driven by some subspecies of axonemal dyneins and find that under some experimental conditions, the collective motion of MTs can display nematic order, millimeter-scale meandering streams or millimeter-scale vortices. To explore the conditions causing such phase-shifts, we examine the effects of mechanical properties of dyneins on the pattern formation.
Short bio:
Kazuhiro Oiwa studied at the Graduate School of Science, the University of Tokyo and gained a PhD in 1988 on force generating mechanism of protein motor ‘dynein’. He joined the National Institute of Information and Communications Technology (NICT), Kobe in 1993 and has worked there to date. As a staff scientist, he applied his knowledge and techniques to understanding the properties of protein motors and also found applications of protein motors to development of nanometer-scale devices and to understanding of ensemble behavior of self-propelled particles and resultant pattern formation. After he served as the Director General of the Advanced ICT Research Institute, NICT from 2008 to 2013, he is now the Distinguished Researcher and the Fellow of NICT. He was awarded the 23rd Osaka Science Prize in 2005.
Fish' n' robots: not a take out food
Speaker: Maurizio Porfiri, New York University, USA
Date and Time: 9:00-10:00, 12/4 (Fri)
Abstract
Engineering design of robots is often inspired by nature; recently developed bioinspired robots accurately imitate various aspects of their live counterparts. Yet, the relationship between engineering and nature has often been one-directional: engineers borrow ideas from nature to build more efficient, more appealing, and better performing robotic systems for use in traditional human-centered applications. In some cases, these systems are used as proxies for studying the natural system, but whether these devices can be integrated within the ‘ecological niche’ inspiring their design seldom is experimentally tested. An even more elemental research question pertains to the feasibility of modulating spontaneous behavior of animal systems through bioinspired robotics. In this talk, we discuss recent research findings at the Dynamical Systems Laboratory of New York University Polytechnic School of Engineering on the feasibility of integrating robotic fish in experimental paradigms to investigate the behavior of social fish. In this context, autonomous robots can be used to offer consistent, customizable, and controllable stimuli for decomposing the complex nature of information sharing and decision making in fish. This talk will address fundamental scientific questions such as: is fish behavioral response influenced by a robotic fish and, if so, what is the consistency of such response? What are the determinants of attraction or repulsion of a robotic fish, and can they be modulated through the administration of psychoactive compounds? How can we objectively quantify the interactions between fish and robots? What is the role of hydrodynamic effects and visual cues? Does the behavior and motion of a robotic fish influence fish social response? How does fish perceive schools of robotic fish? Do fish interact differently with a robotic fish depending on their ‘personality’?
Short bio:
Maurizio Porfiri is a Professor in the Department of Mechanical and Aerospace Engineering at New York University Polytechnic School of Engineering. He received M.Sc. and Ph.D. degrees in Engineering Mechanics from Virginia Tech, in 2000 and 2006; a “Laurea” in Electrical Engineering (with honours) and a Ph.D. in Theoretical and Applied Mechanics from the University of Rome “La Sapienza” and the University of Toulon (dual degree program), in 2001 and 2005, respectively. He is engaged in conducting and supervising research on dynamical systems theory, multiphysics modeling, and underwater robotics. Maurizio Porfiri is the author of more than 175 journal publications and the recipient of the National Science Foundation CAREER award (Dynamical Systems program). He has been included in the “Brilliant 10” list of Popular Science in 2010 and his research featured in all the major media outlets, including CNN, NPR, Scientific American, and Discovery Channel. Other significant recognitions include invitations to the Frontiers of Engineering Symposium and the Japan-America Frontiers of Engineering Symposium organized by National Academy of Engineering, respectively; the Outstanding Young Alumnus award by the college of Engineering of Virginia Tech; the ASME Gary Anderson Early Achievement Award; and the ASME DSCD Young Investigator Award.
The Puzzles of Task and Resource Scheduling in Cell Networks
Speaker: Yechiam Yemini, Columbia University, USA
Date and Time: 8:15-9:15, 12/3 (Thu)
Abstract:
Cells face difficult resource management challenges: from how to manage the scale, diversity and complexities of their mechanisms, to how to coordinate resource scheduling over time scales ranging by 10-11 orders of magnitude; all these through distributed autonomous elements. Recent research advances have been rapidly shedding new understandings of these mechanisms. I will review some of these findings and their potential implications to computing systems.
Short bio:
Yechiam Yemini (YY) is Professor Emeritus of Computer Science at Columbia University. His career has combined academic and serial entrepreneurship interests. His research pioneered a range of networking technologies, from autonomic self-managing networks, to microeconomic network control and security, to high-speed switches and (what became two decades later) software-defined networks. He co-authored over 100 refereed publications and 30 patents, as well as software tools, used widely in network research. Professor Yemini has also been a serial co-founder of 5 high-tech startups, with the first, Comverse Technology (1983), joining the NASDAQ100 and S&P500, and with the most recent one, VMTurbo.com (2009), listed with Forbe’s “America’s 100 Most Promising Companies” for the past three years.
