Graduate

SystemX Seminar: Rethinking the Functional Boundaries of Integrated Radio-Frequency Systems Enables New Wireless Communication Paradigms

Topic: 
Rethinking the Functional Boundaries of Integrated Radio-Frequency Systems Enables New Wireless Communication Paradigms
Abstract / Description: 

Mobile data traffic in 2014 was nearly 30 times the size of the entire global Internet in 2000. Next generation wireless networks are targeting 1000x increase in capacity to meet the insatiable demand for more data. Such a tremendous increase in wireless data will require a complete rethinking of today's wireless communication systems and networks from the physical layer to the network and application layer.

My research program at Columbia University is focused on transformative radio-frequency (RF) and millimeter-wave (mmWave) circuit design techniques that enable new system architectures that challenge the traditional functional boundaries (antenna/RF/analog/digital) of wireless communication systems and specifically enable advanced signal processing techniques at RF, thus enabling new wireless communication paradigms in close collaboration with communications, signal processing and network theorists.

In this talk, I will focus on recent research in CoSMIC lab in this space. I will touch upon our work on energy-efficient and high-power millimeter-wave CMOS circuits that have drawn interest for next-generation 5G cellular communications. The bulk of this talk will focus on our work on enabling full-duplex wireless communication, where transmitters and receivers operate at the same frequency at the same time, thus potentially doubling data throughput, promoting more flexible spectrum usage, and enabling solutions to several network problems. The fundamental challenge in full duplex is the tremendous transmitter self-interference at the receiver, which can be one trillion times more powerful than the desired signal and must be dealt with in all domains. This powerful self-interference is susceptible to uncertainties of the wireless channel (for instance, frequency selectivity and time variance) and the imperfections of the transceiver electronics (nonlinear distortion and phase noise to name a few), making it even harder to deal with. I will discuss several generations of fully-integrated CMOS transceiver ICs with self-interference cancellation that leverage circuit design innovations to enable advanced yet robust signal processing such as noise cancellation, distortion cancellation and wireless channel equalization in the RF and electromagnetic (i.e. antenna) domains and achieve the challenging performance required. I will also briefly touch upon our collaborative work with network theorists to determine the rate gains that are possible under various full duplex scenarios based on realistic physical layer models that we developed. Finally, I will end this talk with a brief description of our ongoing and future work on other emerging wireless communication paradigms.


The SystemX Alliance

Date and Time: 
Thursday, April 9, 2015 - 4:00pm to 5:00pm
Venue: 
Packard 202

Make-a-thon

Topic: 
Make-a-thon
Abstract / Description: 

Fuse is excited to announce our first event of the quarter! We will be partnering with TI and IEEE for a make-a-thon. Starting at 11:00, TI engineers will lead us through a workshop with the TI Launch Pad. This will be followed by a build session and judging round. There will be lunch, snacks, music, and lots of prizes!

This event is perfect for students at the E40 and EE101B level, but will also be fun for more advance circuits students!

You must fill out the google form now to save your spot.
http://goo.gl/forms/YRuDpXEtMa

We hope to see you there!


Please RSVP!

 

Date and Time: 
Saturday, April 11, 2015 - 11:00am to 4:00pm
Venue: 
Packard Atrium

Applied Physics/Physics Colloquium: Tensor Network Renormalization

Topic: 
Tensor Network Renormalization
Abstract / Description: 

The renormalization group (RG), one of the conceptual pillars of statistical mechanics and quantum field theory, explores how the behavior of an extended system depends on the scale of observation. Important notions such as universality, criticality or stability of phases are explained in terms of the existence of fixed-points of the RG flow. On the other hand, studies of entanglement in the context of quantum information have resulted in the development of tensor networks, a framework to efficiently describe many-body wave-functions. In this Colloquium I will review a decade-long research program based on applying tensor networks to the renormalization group. This program has produced a novel non-perturbative, real-space RG approach and the multi-scale entanglement renormalization ansatz (MERA), currently of interest in a wide range of research areas, from statistical mechanics to condensed matter and quantum field theory, from holography in quantum gravity, to error correction in classical and quantum information theory.


Refreshments in the Physics Lobby at 4:00 p.m.

Date and Time: 
Tuesday, March 31, 2015 - 4:15pm to 5:15pm
Venue: 
Hewlett 200

Beyond prosthetics: brain-machine interfaces as common clinical tools

Topic: 
Beyond prosthetics: brain-machine interfaces as common clinical tools
Abstract / Description: 

To date, the scope of brain-machine interfaces (BMIs) has largely been to restore lost function to people with paralysis stemming from conditions such as neurodegenerative disease and spinal cord injury. These systems interface with the brain using neurosurgically implanted electrodes, measure the voltage of individual and groups of neurons, and translate these measurements via a decoding algorithm to control an end effector such as a computer cursor. I will discuss work performed in preclinical rhesus models that led to the highest performing communication BMI demonstrated to date, as well as recent results of an ongoing clinical trial where these preclinical algorithmic innovations have been successfully translated to a human participant, again yielding the highest communication rates of any known clinical BMI.

The example of prosthetics is just one important application leveraging intracortical BMIs as a platform for accurately assessing and acting on the neural state. However, these measurements could play a crucial role in the diagnosis and management of a wide range of neurological and psychiatric diseases and disorders, ranging from stroke and epilepsy to depression and unconsciousness. Just as EEG recordings help localize seizures both temporally and spatially, and MRI imaging provides morphological and gross functional evaluations of the brain, BMI measurements may reveal previously unrecognized disease-specific adulterations in the neural state. Not only could this aid in forming better prognoses, but may also lead to interventions to prevent or alleviate undesirable symptoms and improve rehabilitation. In this manner, the utility of BMIs could extend far beyond communication or motor prosthetics to become an indispensable clinical tool in the treatment of brain disorders. I will discuss the emerging potential and key initial steps of this new class of medical system.


Event is sponsored by the Department of Neurosciences Institute and Bioengineering. View Event on Stanford Neuro Institute.

 

Date and Time: 
Thursday, March 19, 2015 - 10:00am to 11:00am
Venue: 
Clark Center Auditorium

SystemX Seminar: Energy Efficiency and Conversion in 1D and 2D Electronics

Topic: 
Energy Efficiency and Conversion in 1D and 2D Electronics
Abstract / Description: 

We review our recent studies concerning the heterogeneous integration of nanomaterials for low-power electronics and energy harvesting applications. Through careful transport studies of two-dimensional (2D) devices based on graphene and MoS2, we have uncovered details regarding their physical properties and band structure. We have investigated thermoelectric effects in graphene transistors and phase-change memory (PCM) elements for low-power electronics. We found that low-power transistors and memory could be enhanced by built-in thermoelectric effects which are particularly pronounced at nanometer length scales. We have also examined energy harvesting using composites based on one-dimensional (1D) carbon nanotubes, and uncovered both the lower (diffusive) and upper (ballistic) limits of heat flow in 1D and 2D nanomaterials. Our results suggest fundamental limits and new applications that could be achieved through the co-design and heterogeneous integration of 1D and 2D nanomaterials.


The SystemX Alliance: a new forum for university-industry knowledge exchange.

Date and Time: 
Thursday, April 2, 2015 - 4:00pm to 5:00pm
Venue: 
Packard 202

Information Systems Lab Colloquium: Optimization Techniques for Alphabet-Constrained Signal Design

Topic: 
Optimization Techniques for Alphabet-Constrained Signal Design
Abstract / Description: 

The theoretical and computational results in the field of signal design have been of interest to both engineers and mathematicians in the last decades. Signal optimization for active sensing and communications usually deals with various measures of quality (including estimation/detection and information-theoretic criteria), and moreover, the practical condition that the employed signals must belong to a limited signal set. Such diversity of design metrics and signal constraints paves the way for many interesting research works in signal optimization. We study the latest techniques facilitating signal design for optimized actuation, sensing, and communication over constrained sets. In particular, we focus on three different methodologies:

  • Alternating Projections on Converging Sets (ALPS-CS) -- an alternating projections-based approach specialized for constrained alphabets;
  • Power Method-Like Iterations -- a fast approach for alphabet-constrained signal design that resembles power method; and a
  • Monotonically Error-Bound Improving Technique for Optimization (MERIT)-- a novel optimization framework that lays the ground for obtaining computational data-dependent sub-optimality guarantees for the obtained solutions. The new guarantees typically outperform the a priori known guarantees of semidefinite relaxation (SDR) -- a widely used approach for constrained signal design
Date and Time: 
Thursday, March 12, 2015 - 4:15pm to 5:15pm
Venue: 
Hewlett 200

EE380 Computer Systems Colloquium: Dynamic Code Optimization and the NVIDIA Denver Processor

Topic: 
Dynamic Code Optimization and the NVIDIA Denver Processor
Abstract / Description: 

VIDIA's first 64-bit ARM processor, code-named Denver, leverages a host of new technologies to enable high-performance mobile computing. Implemented in a 28-nm process, the Denver CPU can attain clock speeds of up to 2.5 GHz. This talk will outline the Denver architecture and describe some of its technological innovations. In particular this talk will discuss some of the motivations and advantages of dynamic code optimization.

Date and Time: 
Wednesday, March 4, 2015 - 4:15pm to 5:15pm
Venue: 
Gates B03

GSEE Afternoon Tea

Topic: 
EE community event
Abstract / Description: 

Take a break this Wednesday afternoon to enjoy some beverages, snacks, and conversation! All members of the EE community - students, faculty, and staff - are welcome. Don't forget to bring a reusable mug if you have one.

Hope to see you there!

Date and Time: 
Wednesday, March 4, 2015 - 3:00pm to 4:30pm
Venue: 
Packard 2nd Floor lounge

Information Systems Lab Colloquium: Collective computation in nonlinear networks and the grammar of evolvability

Topic: 
Collective computation in nonlinear networks and the grammar of evolvability
Abstract / Description: 

Computation, synchronization, and control are key issues in complex networks. Vast nonlinear networks are encountered in biology, for instance, and in neuroscience, where for most tasks the human brain grossly outperforms engineered algorithms using computational elements 7 orders of magnitude slower than their artificial counterparts. We show that nonlinear systems tools, such as contraction analysis and virtual dynamical systems, yield simple but highly non-intuitive insights about such issues, and that they also suggest systematic mechanisms to build progressively more refined networks and novel algorithms through stable accumulation of functional building blocks and motifs.

Date and Time: 
Wednesday, March 4, 2015 - 4:15pm to 5:15pm
Venue: 
Packard 202

IT-Forum: Searching with measurement dependent noise

Topic: 
Searching with measurement dependent noise
Abstract / Description: 

We consider a search problem in which a target is arbitrarily placed on the unit interval. To acquire the target, any region of the interval can be probed for its presence, but the associated measurement noise increases with the size of the probed region. We are interested in the expected search time required to find the target to within some given resolution and error probability. When the measurement noise is constant (independent of the probed region), this problem is known to be equivalent to standard channel coding with feedback. We characterize the optimal tradeoff between time and resolution (i.e., maximal rate), and show that in contrast to the case of constant measurement noise, measurement dependent noise incurs a multiplicative gap between adaptive search and non-adaptive search. Moreover, our adaptive scheme attains the optimal rate-reliability tradeoff. An extension of this problem into a multi-target setting is also considered. We highlight the equivalence of this extension to coding for a certain multiple access channel and the optimal rate, as a function of the number of targets, is characterized. Finally, we show that as the number of targets increases, the performance gap between adaptive- and non-adaptive search becomes negligible. This talk is based on joint work with Ofer Shayevitz and Tara Javidi.

Date and Time: 
Friday, March 6, 2015 - 1:00pm to 2:00pm
Venue: 
Packard 202

Pages

Subscribe to RSS - Graduate