PhD Talk/Colloquia

The PhD colloquium is a one-hour format, designed for practicing longer conference presentations or job talks. Participants in the spring colloquium will be candidates for the Best PhD Colloquium award.

**Seminars and Talks will be held in Towne 337 on Wednesdays at 12:00 PM unless otherwise specified.**

Spring 2019 Semester

Chamon Luiz Chamon
Wednesday, March 6, 2019
"Functional nonlinear sparse models"
Read the Abstract and Bio

Abstreact: Signal processing and statistics are rich in inherently continuous and often nonlinear problems, either because these are properties of the physical world (e.g., direction-of-arrival for mmWave or super-resolution) or because linear systems do not have the desired properties (robustness in classification problems). In many of these cases, sparsity plays a key role in dealing with indeterminacies to obtain state-of-the-art results. Coping with the infinite dimensionality and non-convexity of these problems typically involves discretization and convex relaxations, e.g., using atomic norms. Though successful, these approaches are not without issues. Discretization often leads to high dimensional, potentially ill-conditioned optimization problems and due to grid mismatch issues, a sparse signal in the continuous domain need not be sparse when discretized. On top of that, relaxing the sparsity objective does nothing to address the non-convexity of dealing with nonlinear measurements. What is more, existing performance guarantees for atomic norm relaxations in the linear case hold under assumptions that may be hard to meet in practice and cannot be checked. In this talk, we show how to address these issues by directly tackling the continuous, nonlinear problem cast as a sparse functional optimization program. We show that despite their infinite dimensionality and non-convexity, they can be solved exactly and efficiently using duality and classical convex optimization tools. We illustrate the range of applications for this new approach by formulating and solving nonlinear spectral estimation and robust classification problems.

Bio: Luiz Chamon is a Ph.D. candidate in the Department of Electrical and Systems Engineering at the University of Pennsylvania. He received the B.Sc. and M.Sc. degree in electrical engineering from the University of São Paulo, Brazil, in 2011 and 2015 and was an undergraduate exchange student at the Masters in Acoustics of the École Centrale de Lyon, France, in 2009. In 2018, he was recognized by the IEEE Signal Processing Society for his distinguished work for the editorial board of the IEEE Transactions on Signal Processing. His research focuses on optimization theory with applications to signal processing, control, and statistics.

Hans Riess
Wednesday, March 13, 2019
"Sheaves, Lattices and Optimization"
Read the Abstract and Bio

Abstract: Sheaf theory is a rich mathematical theory invented in the 1940's by Henri Cartan. More recently, sheaf theory has found its way into the field of Topological Data Analysis (unsupervised learning) and beyond. In this talk, we will describe a new extremely general framework for thinking about networked systems via cellular sheaves, a tool to describe (in)consistency in a system. Our novel contribution is that we consider situations where our data is not vector-valued, but may take values in other datum such as a lattice. After covering the basics, we quickly move on to describing message passing algorithms over networks via cellular sheaves. Along the way, we describe two Laplacians; the first of which vastly generalizes the ubiquitous graph Laplacian, and the second of which (novel) approaches the combinatorial setting of Lattice sheaves. Next, we take a look at how distributed optimization techniques come into play. Finally, we end by suggesting applications to clustering. Comments/questions about how cellular sheaves might fit into their research are especially welcome.

Bio: Hans Riess is a second year Ph.D. candidate under the supervision of Robert Ghrist. Hans received a B.S. in Mathematics from Duke University in 2016. His research brings together areas of mathematics such as algebraic topology, lattice theory, and homological algebra with areas of engineering such as optimization, machine learning, and networked systems.

Markos Epitropou
Wednesday, March 27, 2019
"Dynamic Mechanisms with Verification"
Read the Abstract and Bio

Abstract: We consider a principal who allocates an indivisible object among an infinite number of agents who arrive on-line, each of whom prefers to have the object than not. Each agent has access to private information about the principal's payoff if he receives the object. The decision to allocate the object to an agent must be made upon arrival of an agent and is irreversible. There are no monetary transfers but the principal can inspect agents' reports at a cost and punish them. I will first present a reformulation of this dynamic problem as a compact linear program. Using the formulation we characterize the form of the optimal mechanism and reduce the dynamic version of the inspection problem with identical distributions to an instance of the secretary problem with one fewer secretary and a modified value distribution. This reduction also allows us to derive a prophet inequality for the dynamic version of the inspection problem.

Bio: Markos Epitropou received his Diploma in Electrical and Computer Engineering from the National Technical University of Athens, Athens, Greece in 2012. He later received a Ms. S. in Logic, Algorithms and Computation from the National and Kapodistrian University of Athens, Athens, Greece in 2014. Since August 2014, he has been working towards the Ph.D. degree in the Department of Electrical and Systems Engineering, University of Pennsylvania. His research interests include market design, algorithm design, and combinatorial optimization.

Mohammad Fereydounian
Wednesday, April 3, 2019
"Non-asymptotic Coded Slotted ALOHA"
Read the Abstract and Bio

Abstract: Coding for random access communication is a key challenge in Internet of Things applications. In this paper, the well-known scheme of Coded Slotted Aloha (CSA) is considered and its performance is analyzed in the non-asymptotic regime where the frame length and the number of users are finite. A density evolution framework is provided to describe the dynamics of decoding, and fundamental limits are found on the maximum channel load (i.e., the number of active users per time slot) that allows reliable communication (successful decoding). Finally, scaling laws are established, describing the non-asymptotic relation between the probability of error, the number of users, and the channel load.

Bio: Mohammad is a second year PhD student in the Department of Electrical and Systems Engineering at the University of Pennsylvania. He is also a M.Sc. student of statistics in Wharton school at the same time. Before joining Penn, he received two separated B.Sc. degrees in electrical engineering and pure mathematics as well as a M.Sc. degree in pure mathematics, all from Sharif University of Technology, Tehran, Iran. His research interests include coding and information theory, network and graph theory, optimization, and machine learning. His research interests in pure math include algebraic combinatorics, algebraic graph theory, Galois theory, advanced field and module theory, measure theory, topology, manifolds, and functional analysis.

Yung Huang
Wednesday, April 10, 2019
"Towards Compact Diamond Quantum Memory"
Read the Abstract and Bio

Abstract: Solid-state quantum emitters have emerged as building blocks for quantum technologies and network systems. In particular, the nitrogen-vacancy (NV) center in diamond is a leading platform in quantum communication. While some key principles and techniques for quantum memory networks have been demonstrated, collection efficiency, scalability, and overhead management remain critical roadblocks in fully realizing a compact, efficient quantum register. Building on recent advances in metasurfaces and photonics, we demonstrate the first steps towards a fiber-coupled, chip-scale quantum memory with an immersion metalens coupled to a single NV center in diamond. We will discuss the design and practical challenges and lay out a roadmap for continuing to improve this work.

Bio: Yung received their B.S.E in electrical engineering from Princeton University in 2015. They are now pursing their Ph. D. in Electrical and Systems Engineering at the University of Pennsylvania, where they work in the Quantum Engineering Laboratory led by Professor Lee Bassett. Yung's current research interests are in designing and engineering solid-state systems for quantum sensing and communication.

Lin Du
Wednesday, April 17, 2019
"CMOS Compatible Hermetic Packages Based on Localized Fusion Bonding of Fused Silica"
Read the Abstract and Bio

Abstract: Hermetic packages compatible with complementary metal-oxide-semiconductor (CMOS) processes and capable of transmitting RF and optical energy are in high demand. A hermetic packaging process based on the localized fusion bonding of patterned and stacked fused silica wafers to encapsulate a CMOS chip has been successfully demonstrated. The process also allows for feedthroughs to be fabricated within the package for wired interconnection to encapsulated devices. A carbon dioxide laser is applied to achieve localized fusion bonding of the fused silica stack. Exploiting vector motion of the laser, simultaneous bonding and dicing is achieved within 2 seconds. The low thermal conductivity of fused silica thermally isolates the CMOS chip from the bond area. The heat diffusion during the process is examined both experimentally and using finite element analysis. The temperature falls to CMOS compatibility levels (under 400°C) within a distance of 660 mm from the bond line. The viability of the packaging process and the hermeticity of the package are simultaneously evaluated by encapsulating a commercial MEMS humidity sensor within the package and monitoring the internal package humidity as a function of time. The packaged sensor was placed in an external relative humidity environment of 85% at 24°C and the relative humidity change inside the package was measured over a time span of 300 days. A water vapor leakage rate less than 4.6 x 10-14 atm cm3/s was observed. These preliminary results suggest that the hermeticity of the package can meet the criterion of implantation in human body for more than 70 years for a package with an inner volume of 0.0012 cm3.

Bio: Lin Du received the B.S. degree in Mechatronical Engineering and the M.S. degree in mechanical engineering, both from Beijing Institute of Technology, Beijing, China, in 2012 and 2015, respectively. She is currently a Ph.D. candidate in the Department of Electrical and Systems Engineering at University of Pennsylvania, where she works in the MicroSensors and MicroActuators Group led by Professor Mark Allen. Lin's research interests include implantable micro-electro-mechanical (MEMS) packages, micromachining, and its applications for human health monitoring systems with wireless transmission.

Fernando Gama
Wednesday, April 24, 2019
"Graph Convolutions in Neural Networks"
Read the Abstract and Bio

Abstract: Neural networks are information processing architectures that consist of a concatenation of layers, each of which applies a linear transform followed by an activation function. The parameters of the linear transform are learned by minimizing a task-related cost function on a given training set. Neural networks do not scale since the number of parameters to learn depends on the size of the data. Convolutional neural networks overcome this issue by regularizing the linear transform to exploit the regular structure present in many datasets such as time signals or images. By forcing the linear transform to be a bank of filters, the computationally-efficient operation of convolution can be used to compute the output. Network data, however, does not exhibit a regular structure to which a convolution can be readily applied. We exploit graph signal processing to define a graph convolution that effectively exploits the underlying irregular structure of network data, and that can be efficiently implemented by local exchanges among nodes. The regularization of the linear transform by means of a graph convolution gives rise to graph neural networks (GNNs). Furthermore, we discuss two key properties of GNNs, namely permutation invariance and stability to topology changes, that help explain the success of these architectures in processing network data.

Bio: Fernando Gama received the electronic engineer degree from the University of Buenos Aires, Argentina, in 2013, and the M.A. degree in statistics from the Wharton School, University of Pennsylvania, Philadelphia, PA, USA, in 2017. He is currently working towards the Ph.D. degree with the Department of Electrical and Systems Engineering, at the University of Pennsylvania. He has been a visiting researcher at TU Delft, the Netherlands, in 2017 and a research intern at Facebook Artificial Intelligence Research, Montreal, Canada, in 2018. His research interests are in the field of information processing and machine learning over network data. He has been awarded a Fulbright scholarship for international students.

Fall 2018 Semester

Mark Eisen
Wednesday, October 3rd
"Control-Aware Radio Resource Allocation in Low-Latency Wireless Systems"

Read the Abstract and Bio

Abstract: We consider the problem of allocating radio resources over wireless communication links to control a series of independent wireless control systems. Low-latency transmissions are necessary in enabling time-sensitive control systems to operate over wireless links with high reliability. Achieving fast data rates over wireless links thus comes at the cost of reliability in the form of high packet error rates compared to wired links due to channel noise and interference. However, the effect of the communication link errors on the control system performance depends dynamically on the control system state. We propose a novel control-communication co-design approach to the low-latency resource allocation problem. We incorporate control and channel state information to make scheduling decisions over time on frequency, bandwidth and data rates across the next-generation Wi-Fi based wireless communication links that close the control loops. Control systems that are closer to instability or further from a desired range in a given control cycle are given higher packet delivery rate targets to meet. Rather than a simple priority ranking, we derive precise packet error rate targets for each system needed to satisfy stability targets and make scheduling decisions to meet such targets while minimizing total transmission time. The resulting Control-Aware Low Latency Scheduling (CALLS) method is tested in numerous simulation experiments that demonstrate its effectiveness in meeting control-based goals under tight latency constraints relative to control-agnostic scheduling.

Bio: Mark Eisen received the B.Sc. degree in electrical engineering from the University of Pennsylvania, Philadelphia, USA in 2014. He is now working towards his PhD in the Department of Electrical and Systems Engineering at the University of Pennsylvania. His research interests include distributed optimization and machine learning. In the summer of 2013, he was a research intern at the Institute for Mathematics and its Applications at the University of Minnesota, Minneapolis, MN. Mr. Eisen was awarded Outstanding Student Presentation at the 2014 Joint Mathematics Meeting, as well as the recipient of the 2016 Penn Outstanding Undergraduate Research Mentor Award. He spent the summer of 2018 as an intern at Intel Labs in Hillsboro Oregon, working in the area of wireless industrial systems.

Anastasios Tsiamis
Wednesday, October 10th
"Secrecy Codes for Wireless Control Systems"

Read the Abstract and Bio

Abstract: Because of its broadcast nature, the wireless medium is susceptible to eavesdropping. This raises confidentiality concerns in networked control systems, where many sensors and devices communicate wirelessly while carrying sensitive data about the system's operation. In this talk, we focus on eavesdropping attacks in a remote estimation scenario. An authorized user estimates the state of a linear system, based on the packets received from a sensor. Meanwhile, the packets may also be intercepted by an eavesdropper. Our goal is to design a coding scheme at the sensor which hides the state information from the eavesdropper. We present a new class of codes, termed "state-secrecy codes", which are specialized for dynamical systems. By applying properly designed linear transformations to the current and previously received states, they impose artificial unstable dynamics to the eavesdropper’s estimation scheme. As a result, under minimal conditions, they achieve secrecy in the estimation theoretic sense: the eavesdropper’s minimum mean square error converges to the maximum possible value, i.e. the open-loop prediction one when no message is received. Those conditions require that at least once, the user receives the corresponding packet while the eavesdropper fails to intercept it.

Bio: Anastasios Tsiamis received the Diploma degree in Electrical and Computer Engineering from the National Technical University of Athens, Greece, in 2014. Currently, he is a Ph.D. student in the Department of Electrical and Systems Engineering, University of Pennsylvania, working with prof. George Pappas. His research interests include control systems security, system identification, and learning for control.

Jacob Seidman
Wednesday, October 24th
"A Chebyshev-Accelerated Primal-Dual Method for Distributed Optimization"

Read the Abstract and Bio

Abstract: We consider a distributed optimization problem over a network of agents aiming to minimize a global objective function that is the sum of local convex and composite cost functions. To this end, we propose a distributed Chebyshev accelerated primal-dual algorithm to achieve faster ergodic convergence rates. In standard distributed primal-dual algorithms, the speed of convergence towards a global optimum (i.e., a saddle point in the corresponding Lagrangian function) is directly influenced by the eigenvalues of the Laplacian matrix representing the communication graph. In this paper, we use Chebyshev matrix polynomials to generate gossip matrices whose spectral properties result in faster convergence speeds, while allowing for a fully distributed implementation. As a result, the proposed algorithm requires fewer gradient updates at the cost of additional rounds of communications between agents. We illustrate the performance of the proposed algorithm in a distributed signal recovery problem. Our simulations show how the use of Chebyshev matrix polynomials can be used to improve the convergence speed of a primal-dual algorithm over communication networks, especially in networks with poor spectral properties, by trading local computation by communication rounds.

Bio: Jacob is a third year PhD student in the AMCS program at Penn advised by Victor Preciado and George Pappas. His current research interests include optimization and generalization in machine learning problems.

Matthew Malencia
Wednesday, October 31st
"Rethinking Scientiifc Presentations: The Assertion-Evidence Approach"

Read the Abstract and Bio

Abstract: Defaults have a powerful influence on human behavior. When companies ask employees to elect not to be enrolled in a retirement plan, rather than asking them to opt-in, enrollment nearly doubles. Default options are also seen to effect organ donorship, email marketing, HIV testing, and, unfortunately, research presentations.
PowerPoint defaults promote a topic-subtopic approach with word-filled slides and small figures. Cognitive scientists and psychologists have shown that these practices hinder audience understanding. And engineers and scientists know this; when asked to identify problems in peer presentations, they consistently identify these problems. Despite knowing the power of presentations for networking and promoting technical work, and easily identifying these problems in others’ slides, researchers consistently follow the defaults.
This presentation teaches the Assertion-Evidence approach to slide design. This approach is based on psychology and neuroscience research on learning and seeks to combat the consequences of default slide design. Experimental testing shows that this slide design approach improves material recall for the audience, drastically reduces audience misconceptions, and even improves presenter understanding.

Short Bio: Matthew received a B.S. in Mechanical Engineering from Penn State in 2016, with honors in Electrical Engineering. During his undergraduate career, he was trained in technical presenting as part of Engineering Ambassadors, an organization that conducts STEM outreach. After graduating, he worked for MIT Lincoln Laboratory doing research in the Homeland Protection. While at Lincoln Laboratory, he studied cognitive robotics with a professor in the Aeronautics and Astronautics Department at MIT. He is now a PhD student in the Department of Electrical and Systems Engineering at University of Pennsylvania with Vijay Kumar and George Pappas.


Andreea Alexandru
Wednesday, November 7th
"Cloud-based MPC with Encrypted Data"
Read the Abstract and Bio

Abstract: This talk explores the privacy of cloud outsourced Model Predictive Control (MPC) for a linear system with input constraints. In our cloud-based architecture, a client sends her private states to the cloud who performs the MPC computation and returns the control inputs. In order to guarantee that the cloud can perform this computation without obtaining anything about the client's private data, we employ a partially homomorphic cryptosystem. We propose protocols for two cloud-MPC architectures motivated by the current developments in the Internet of Things: a client-server architecture and a two-server architecture. In the first case, a control input for the system is privately computed by the cloud server, with the assistance of the client. In the second case, the control input is privately computed by two independent, non-colluding servers, with no additional requirements from the client. We prove that the proposed protocols preserve the privacy of the client's data and of the resulting control input. Furthermore, we compute bounds on the errors introduced by encryption. We present numerical simulations for the two architectures and discuss the trade-off between communication, MPC performance and privacy.

Bio: Andreea Alexandru received the B.Eng. degree in Automatic Control and Systems Engineering from “Politehnica” University of Bucharest, Romania, in 2015. She is currently pursuing a Ph.D. degree in the Department of Electrical and Systems Engineering, University of Pennsylvania, working with prof. George Pappas and prof. Ali Jadbabaie. Her research interests lie in the security of control systems and private computations.


David Hopper
Wednesday, November 7th
"Optimizing Spin Readout of the Nitrogen-Vacancy Center in Diamond with Spin-to-Change Conversion"

Read the Abstract and Bio

Abstract: The nitrogen-vacancy center in diamond is a mature platform for quantum technology, enabling sophisticated quantum information protocols as well as versatile quantum sensors operating in previously unreachable size and field regimes. The standard photoluminescence-based spin readout is fast (300 ns) but typical measurements yield only a few hundredths of a photon on average, necessitating tens of thousands of repeats to overcome shot noise in detecting the spin state. Spin-to-charge conversion (SCC) offers an alternative readout with significantly improved single-shot information. However, this benefit comes at the expense of orders of magnitude longer readout durations. Here, we present a framework for optimizing the SCC readout parameters that leads to dramatic reductions in overall measurement acquisition times. The improvements arise from the combination of an analytical charge readout model with numerical optimization of the overhead durations. In addition, our current work on incorporating real-time logic and classical signal processing for optimal sensor performance will be outlined. Finally, we discuss relevant applications such as spin relaxometry and control of nuclear registers and outline how other spin readout methods can benefit from this framework.

Bio: David received his B. Sc. In Physics from The Pennsylvania State University in 2014. He is now pursing his Ph. D. in Physics at the University of Pennsylvania, where he works in the Quantum Engineering Laboratory lead by Professor Lee Bassett. David’s research interests are in semiconductor quantum dynamics and designing systems that enable the utilization of quantum technology.


 

Maria Peifer
Wednesday, November 14th
"Learning Sparse Multi-Kernel Representations of Reproducing Kernel Hilbert Space Models"

Read the Abstract and Bio

Abstract: Reproducing kernel Hilbert spaces (RKHSs) have been at the core of successful non-parametric tools in signal processing, statistics, and machine learning. Despite their success, the computational complexity of these models often hinders their use in practice. Fitting RKHS models typically relies on the representer theorem to express the solution space as a combination of kernels evaluated at the training samples. Thus, the computational cost of evaluating these models is proportional to the number of training samples, which in many applications is prohibitively high. Additionally, in functions with heterogeneous degrees of smoothness, the complexity is artificially kept high by the parts of the function that vary most. In this work we propose a method, which addresses both problems by obtaining a sparse representation which adapts the smoothness and location of each kernel.

Bio: Maria Peifer is currently a PhD student in Prof. Alejandro Ribeiro's lab. She received her Master's degree in Electrical Engineering from Rutgers University and her bachelor degrees in Computer Engineering and Electrical Engineering from Drexel University. Her research interests are in optimization, and multi-kernel learning.


Mohammad Fereydounian
Wednesday, November 28th
"Channel Coding at Low Capacity"

Read the Abstract and Bio

Abstract: Low-capacity scenarios have become increasingly important in the technology of Internet of Things (IoT) and next generation of mobile networks. Such scenarios require efficient, reliable transmission of information over channels with extremely small capacity. Within these constraints, the performance of state-of-the-art coding techniques is far from optimal in terms of either rate or complexity. Moreover, the current non-asymptotic laws of optimal channel coding provide inaccurate predictions for coding in the low-capacity regime. In this paper, we provide the first comprehensive study of channel coding in the low-capacity regime. We will investigate the fundamental non-asymptotic limits for channel coding as well as challenges that must be overcome for efficient code design in low-capacity scenarios.

Bio: Mohammad is a second year PhD student in the Department of Electrical and Systems Engineering at the University of Pennsylvania. He is also a M.Sc. student of statistics in Wharton school at the same time. Before joining Penn, he received two separate B.Sc. degrees in electrical engineering and pure mathematics as well as a M.Sc. degree in pure mathematics, all from Sharif University of Technology, Tehran, Iran. His research interests include coding and information theory, network and graph theory, optimization, and machine learning. His research interests in pure math include algebraic combinatorics, algebraic graph theory, Galois theory, advanced field and module theory, measure theory, topology, manifolds, and functional analysis.


Vasant Iyer
Wednesday, December 5th
"Non-Invasive Cancer Monitoring with CMOS-graphen Sensor Arrays"

Read the Abstract and Bio

Abstract: In cancer diagnosis and monitoring, “liquid biopsy” – searching for signatures of cancer in the blood – has been proposed as a non-invasive alternative to imaging and tissue sampling. Liquid biopsy allows cancer to be monitored more frequently and precisely. Many liquid biopsy strategies focus on circulating tumor cells (CTCs), which carry information about their parent tumor; however, they are extremely rare and must be distinguished from billions of other cells in the blood. Finding CTCs currently requires complex instrumentation and data analysis, often by highly trained experts. In this talk, we present an automated diagnostic that uses magnetic flow cytometry to find CTCs in whole blood, significantly reducing cost, size, and sample preparation requirements. An array of microscale graphene Hall sensors is integrated with parallelized CMOS circuitry and microfluidics to enable sensitive detection at high throughput. This talk will focus on the new fabrication, circuit, and system-level techniques developed in the course of designing the heterogeneous platform.

Bio: Vasant received a B.S. in electrical engineering from Caltech in 2017. He is now a PhD student in the Department of Electrical and Systems Engineering at the University of Pennsylvania, jointly supervised by David Issadore and Firooz Aflatouni. His research interests are in the design of new biomedical and clinical tools using approaches from microelectronics, photonics, and microfluidics.



Tzu-Yung (Yung) Huang
Wednesday, December 5th
"Engineering Platforms for Diamond Quantum Sensing and Photonics"

Read the Abstract and Bio

Abstract: The nitrogen-vacancy (NV) center in diamond is the basis for emerging quantum technologies including sensing, quantum communication, and quantum computing. Many sensing modalities possible with the NV center, such as magnetometry and electrochemical potential sensing, have been demonstrated, paving the way to more targeted and dynamical sensing platforms. On the other hand, while the NV center has been heralded as one of the building blocks for quantum network systems, collection efficiency and scalability remain critical roadblocks. In this presentation, we examine ways to improve on these diamond-based systems through engineering. First, we explore the viability of site-specific, dynamic quantum sensing systems using nanodiamonds, leveraging their low toxicity and apparent availability for functionalization. We then investigate pathways to better collection efficiency and scalability in diamond systems through photonic structures. Finally, we provide an outlook on engineering NV- and diamond-based platforms in quantum sensing and communication.

Bio: Yung received their B.S.E in electrical engineering from Princeton University in 2015. They are now pursing their Ph. D. in Electrical and Systems Engineering at the University of Pennsylvania, where they work in the Quantum Engineering Laboratory lead by Professor Lee Bassett. Yung's current research interests are in designing and engineering solid-state systems for quantum sensing and communication.


Raj Patel
Wednesday, December 12th
"Creation and Controls of Single Spins Hexagonal Boron Nitride"

Read the Bio
Abstract: Hexagonal boron nitride (hBN) is a van der Waals material which hosts visible fluorescent emitters at room temperature. These fluorescent emitters have been shown to be single-photon sources. Recently, some of these quantum emitters have exhibited spin-dependent fluorescence. The presence of single spins and single-photon sources at room temperature combined with low dimensionality makes hBN a unique material for applications such as quantum sensing and communication. However, it is necessary to identify and characterize the spin dynamics of these emitters. In fluorescent emitters, Optically Detected Magnetic Resonance (ODMR) can be used for understanding the spin dynamics. An understanding of how to create the emitters on-demand is also necessary for practical applications. These questions remain to be answered. In this talk, we present recent results and on-going experiments geared towards identifying and characterizing spin-based quantum emitters in hBN. We present results from recent observation of spin-dependent quantum emission in hBN, techniques to create and characterize these emitters, devices developed to perform ODMR and on-going experiments.

Bio: Raj received his M.Sc. in Physics and B.E. in Mechanical Engineering from BITS Pilani - Goa, India in 2015. He received an M.S.E. in Materials Science & Engineering from University of Pennsylvania in 2017. He is now pursuing his Ph.D. in Electrical & Systems Engineering at the University of Pennsylvania, where he works in the Quantum Engineering Laboratory lead by Prof. Lee Bassett. His research interests are in understanding spin-based quantum emitters in two-dimensional materials and other solid-state systems for quantum sensing and computation applications.

Clark Zhang
Wednesday, February 20th
"Machine Learning in Motion Planning"

Read the Abstract and Bio

Abstract: Motion planning has had many algorithmic and theoretical developments and can reliably work for many robotic systems. However, high dimensional systems with new and interesting dynamics are still difficult to plan with. For these cases, we seek to combine machine learning techniques with traditional planning algorithms to either improve their speed or to plan with systems that are hard to model analytically. We will first present work in learning heuristics for random sampling based planners. These planners are more effective in high dimensional spaces but can depend on lot on heuristics to guide the search. An appealing approach is to learn these heuristics with data gathered from past experiences. We then present work on learning and using dynamics models. Learning a system model from data is highly useful for modeling systems that are hard to model analytically or require too many precise physical measurements. A new objective function for learning these models are presented that incorporate constraints on "sufficient accuracy."

Bio: Clark is a third year PhD student in ESE at Penn working with Alejandro Ribeiro and Daniel Lee. His research interests include incorporating machine learning techniques into traditional robotic algorithms to address their weaknesses. He received his B.S.E in Computer Engineering at the University of Michigan.