Scalable Approaches to Communication and Inference: Minimalistic strategies for measurement and coordination

Emerging engineered systems dwarf their predecessors in scale. As a result, minimalistic design approaches that extract the essential features of the problem at hand have become compelling. Adopting such designs from the very outset enables us to decrease the problem scale to manageable levels. In this talk, we consider two examples which illustrate the benefits of this approach.

We first consider the problem of estimating continuous valued parameters from a few random projections of a high dimensional signal (compressive measurements). A direct application of standard compressed sensing based on discretization suffers from performance loss due to basis mismatch. We show that this is not an inherent limitation of compressive measurements. To this end, we consider lower bounds on estimation error variance, the Cramer Rao Bound (CRB) and the Ziv-Zakai Bound (ZZB) and show that random projections preserve these bounds up to an SNR penalty equal to the dimensionality reduction factor. We show how the convergence of the ZZB to the CRB can be used to tightly predict the number of compressive measurements needed to avoid gross estimation errors. We illustrate these ideas using the example of channel estimation for large 60GHz arrays.

The second problem we consider is that of identifying a user’s interests from just his/her tweet times. By using the known timing of “events” associated with a topic (such as the times when a baseball team plays its games), we are able to identify users interested in this topic (the baseball team). We also show how the timing of these events can themselves be inferred from aggregate Twitter feeds obtained by querying Twitter with a few keywords related to the topic.

Posted in Defenses | Comments Off

ARWU ranks UCSB Engineering at #7 in the World

rank methodology chart
The 2014 Academic Ranking of World Universities (ARWU) places UC Santa Barbara Engineering/Technology and Computer Science as #7 in the world

UCSB also received a perfect score of 100 for engineering in the criteria category of percentage of papers published in the top 20% of journals in engineering fields. According to ARWU’s report on methodology, the top 20% journals are defined as “their impact factors in the top 20% of each ISI category according to Journal Citation Report, 2012” and that the score is calculated as “the number of papers in the top 20% journals of a particular broad subject field to that in all journals of the field.”

ARWU uses six objective indicators to rank world universities, including the number of alumni and staff winning Nobel Prizes and Fields Medals, number of highly cited researchers selected by Thomson Reuters, number of articles published in journals of Nature and Science, number of articles indexed in Science Citation Index – Expanded and Social Sciences Citation Index, and per capita performance of a university.

Posted in News, Press Releases | Comments Off

How Small Can One Make a Semiconductor Laser?

Recently there has been a surge in activities devoted to development of the nano-scale semiconductor lasers. In particular, semiconductor lasers employing surface-plasmon polaritons in metal dielectric structures (“spasers”) with their promise of subwavelength operation have generated a lot of interest in the laser community. Yet it is not clear exactly how small semiconductor laser can be made without sacrificing performance.

In my talk I will review the recent experimental and theoretical results and present a theory that would clearly outline the fundamental limits of how small can the nano-laser actually be. First I will show that in order to go beyond diffraction limit one absolutely must use metallic structures with associated loss. Then I will show that the lasing threshold of the single mode metal-semiconductor nano-laser (spaser) is determined only by the photon absorption rate in the metal and exhibits very weak dependence on the composition, shape, size (as long as it is less than half-wavelength) and temperature of the gain medium. This threshold current is on the order of a few tens of micro-amperes for most semiconductor-metal combinations which leads to unattainably high threshold current densities for a substantially subwavelength in all three dimensions semiconductor laser (spaser). At the same time, lasers that are sub-waveelngth in only one dimensions, particularly in infrared and THz regions can be made operational.

I will also discuss the issues of coherence properties of nano-lasers, and their modulation speed and compare them with those of standard semiconductor lasers . I will also consider surface plasmon emitting diodes, (SPED’s), operating far below “spasing” threshold that may be a more viable option for the chip scale integrated nanophotonics.


  1. J. B. Khurgin, G. Sun, Comparative analysis of Spasers, VCSELs, and Surface plasmon emitting diodes (SPED’s), Nature Photonics (2014)
  2. J. B. Khurgin, G. Sun, “Scaling of losses with size and wavelength in Nanoplasmonics” Appl. Phys. Lett, 99, 211106 (2011)
  3. J. B. Khurgin , G. Sun, “Practicality of compensating the loss in the plasmonic waveguides using semiconductor gain medium”, Appl. Phys. Lett, 100, 011105 (2012)
  4. J. B. Khurgin, G. Sun, “Injection pumped single mode surface plasmon generators: threshold, linewidth, and coherence”, Optics Express 20 15309-15325 (2012)
  5. J. B. Khurgin and G. Sun: “How small can “ Nano ” be in a “ Nanolaser ”?”, Nanophotonics, 1, 3-8 (2012)
Posted in Events | Comments Off

High Speed Integrated Circuits for High Speed Coherent Optical Communications

With the development of (sub) THz transistor technologies, high speed integrated circuits up to sub-THz frequencies are now feasible. These high speed and wide bandwidth ICs can improve the performance of optical components, coherent optical fiber communication, and imaging systems. In current optical systems, electrical ICs are used primarily as driving amplifiers for optical modulators, and in receiver chains including TIAs, AGCs, LPFs, ADCs and DSPs. However, there are numerous potential applications in optics using high speed ICs, and different approaches may be required for more efficient, compact and flexible optical systems.

This dissertation will discuss three different approaches for optical components and communication systems using high speed ICs: a homodyne optical phase locked loop (OPLL), a heterodyne OPLL, and a new WDM receiver architecture.

The homodyne OPLL receiver is designed for short-link optical communication systems using coherent modulation for high spectral efficiency. The phase-locked coherent receiver can recover the transmitted data without requiring complex back-end digital signal processing to recover the phase of the received optical carrier. The main components of the homodyne OPLL are a photonic IC (PIC), an electrical IC (EIC), and a loop filter. One major challenge in OPLL development is loop bandwidth; this must be of order 1 GHz in order for the loop to adequately track and suppress the phase fluctuations of the locked laser, yet a 1 GHz loop bandwidth demands small (<100 ps) propagation delays if the loop is to be stable. Monolithic integration of the high-speed loop components onto one electrical and one photonic IC decreases the total loop delay. We have designed and demonstrated an OPLL with a compact size of 10 x 10 mm2, stably operating with a loop bandwidth of 1.1 GHz, a loop delay of 120 ps, a pull-in time of 0.55 us and lock time of <10 ns. The coherent receiver can received 40 Gb/s BPSK with a bit error rate (BER) of <10-7, and operates up to 35 Gb/s with BER <10-12.

The thesis also describes heterodyne OPLLs. These can be used to synthesize optical wavelengths of a broad bandwidth (optical wavelength synthesis) with narrow linewidth and with fast frequency switching. There are many applications of such narrow linewidth optical signal sources, including low phase noise mm-wave and THz-signal sources, wavelength-division-multiplexed optical transmitters, and coherent imaging and sensor systems. The heterodyne OPLL also has the same stability issues (loop delay and sensitivity) as the homodyne OPLL. In the EIC, a single sideband mixer operating using digital design principles (DSSBM) enables precisely controlled sweeping of the frequency of the locked laser, with control of the sign of the frequency offset. The loop’s phase and frequency difference detector (PFD) uses digital design techniques to make the OPLL loop parameters only weakly sensitive to optical signal levels or optical or electrical component gains. The heterodyne OPLL operates stably with a loop bandwidth of 550 MHz and loop delay of <200 ps. An initial OPLL design exhibited optical frequency (wavelength) synthesis from -6 GHz to -2 GHz and from 2 GHz to 9 GHz. An improved OPLL reached frequency tuning up to 25 GHz. The homodyne OPLL exhibits -110 dBc/Hz phase noise at 10 MHz offset and -80 dBc/Hz at 5 kHz offset.

Finally, the thesis describes a new WDM receiver architecture using broadband electrical ICs. In the proposed WDM receiver, a set of received signals at different optical wavelengths are mixed against a single optical local oscillator. This mixing converts the WDM channels to electrical signals in the receiver photocurrent, with each WDM signal being converted to an RF sub-carrier of different frequency. An electrical IC then separately converts each sub-carrier signal to baseband using single-sideband mixers and quadrature local oscillators. The proposed receiver needs less complex hardware than the arrays of wavelength-sensitive receivers now used for WDM, and can readily adjust to changes in the WDM channel frequencies. The proposed WDM receiver concept was demonstrated through several system experiments. Image rejection of greater than 25 dB, adjacent channel suppression of greater than 20 dB, operation with gridless channels, and six-channel data reception at a total 15 Gb/s (2.5 Gb/s BPSK x 6-channels) were demonstrated.

Posted in Defenses | Comments Off

Automated and Interactive Segmentation Methods for 5D Microscopy Images

Accurate segmentation of cells and tissues in 5D (3D + time + multiple channels) microscopy data is critical in quantitative developmental biology. This is a challenging task due to the large volumes of data (tens of gigabytes for single time lapse 3D sequence) and the inherent characteristics of microscopy image acquisition. In this research we developed new supervised and unsupervised segmentation methods with a focus on application to morphogenesis.

First, we developed a principled approach to unsupervised segmentation and fusion of multiple segmentations. A linear optimization framework is proposed for the joint correction of multiple over-segmentations obtained from different methods. The main idea motivating this approach is that over-segmentations, from a pool of methods with various parameters, are likely to agree on the correct segment boundaries, while spurious boundaries are likely to be method or parameter-dependent. Secondly, we introduced an interactive segmentation and analysis tool for 5D microscopy data, called CellECT. An adaptive confidence measure, called “cell-ness” metric is used to highlight regions of uncertainty in the segmentation. This metric quantifies how much a segment deviates from a typical correct segment. This metric adapts to the dataset and learns from the user interactions. The proposed methods are validated on ascidian time-lapse 3D volume data. CellECT is distributed as an open source software and is used in other quantitative biology applications.

Posted in Defenses | Comments Off

Distributed Tracking and Re-Identification in a Camera Network

This dissertation addresses the challenges in large scale deployment of wide area camera networks and automated analysis of resulting big data. Analysis of such data is limited due to communication bottlenecks and low computational power at individual nodes. The presentation will focus on distributed tracking and search/retrieval in a camera network.

For object tracking in overlapping camera views, we propose a strategy for inducing priors on scene specific information into a multiple camera tracker. Contextual information such as crowd flow, entry/exit points, and known obstacles can be leveraged as scene specific priors. A novel probabilistic multiple camera tracking algorithm with a distributed loss function for incorporating scene priors is proposed, and this leads to a significant increase in the overall tracking accuracy. For non-overlapping views, a novel graph based model is proposed to represent spatio-temporal relationships between objects for search and retrieval tasks. This representation exploits the fact that objects occurring in close spatial-temporal proximity are not completely independent and serve as context for each other. Additional information such as appearance and scene context can also be encoded into the graph model to improve the overall accuracy. A graph ranking strategy is used to order the items based on similarity with an emphasis on diversity. Extensive experimental results on a ten camera network are presented.

Posted in Defenses | Comments Off

Modeling Eye Tracking Data with Application to Object Detection

This research focuses on enhancing computer vision algorithms using eye tracking and visual saliency. Recent advances in eye tracking device technology have enabled large scale collection of eye tracking data, without affecting viewer experience. As eye tracking data is biased towards high level image and video semantics, it provides a valuable prior for object detection in images and object extraction in videos. We specifically explore the following problems in the thesis: 1) eye tracking and saliency enhanced object detection, 2) eye tracking assisted object extraction in videos, and 3) role of object co-occurrence and camera focus in visual attention modeling.

Since human attention is biased towards faces and text, in the first work we propose an approach to isolate face and text regions in images by analyzing eye tracking data from multiple subjects. Eye tracking data is clustered and region labels are predicted using a Markov random field model. In the second work, we study object extraction in videos using eye tracking prior. We propose an algorithm to extract dominant visual tracks in eye tracking data from multiple subjects by solving a linear assignment problem. Visual tracks localize object search and we propose a novel mixed graph association framework, inferred by binary integer linear programming. In the final work, we address the problem of predicting where people look in images. We specifically explore the importance of scene context in the form of object co-occurrence and camera focus. The proposed model extracts low-, mid- and high-level and scene context features and uses a regression framework to predict visual attention map. In all the above cases, extensive experimental results show that the proposed methods outperform current state-of-the-art.

Posted in Defenses | Comments Off

Spatial Pattern Modeling and Discovery in Biological Images

Studying spatial arrangement and relationships in full tissue samples can improve our understanding of the various developmental/pathological processes that underlie proper organ or organism function. In particular, it has been found that neuronal or vascular structures are pervasive in many tissues, and oftentimes are spatially correlated with other cells. This work aims to discover those relationships, by extracting biological knowledge from cellular and sub-cellular imaging using spatial point process methods.

In this defense, I present my discoveries on spatial distributions and attributes of dendritic spines and retinal astrocytes, two crucial elements in the mammalian nervous system. Although little is known about the spatial distributions of either respective to their surroundings and attributes, this thesis attempts to pose some possible biological hypotheses based on strong statistical evidence, as well as further extend the tools used for spatial analysis. In particular, we develop a multi-type version of the linear network K-function, a summary function used for measuring clustering or repulsion of point features existing on a linear network. While the methods used in this work are directly applied to biological tissues and images, they have broad applications in many other domains.

Posted in Defenses | Comments Off

Current Status and Future Prospects of Gallium Oxide-based Optoelectronics

Gallium oxide (Ga2O3) has excellent material properties for power device applications represented by the extremely large breakdown field of 8 MV/cm due to its large band gap of 4.6~4.9 eV. Another important feature in industry is that large single-crystal β-Ga2O3 substrates can be fabricated from a melt-grown bulk. We recently succeeded in fabricating Ga2O3 metal-oxide-semiconductor field-effect transistors (MOSFETs) on single-crystal Ga2O3 (010) substrates by using newly developed technologies for making single-crystal substrates, growing conductivity-controlled epitaxial films, and fabricating devices. The MOSFETs exhibited excellent device characteristics including an off-state breakdown voltage of over 400 V, an extremely low leakage current (below the lower limitation of the measurement instrument), and a high on/off drain current ratio of more than 10 orders of magnitude. The devices also showed good performance at 250°C with no significant permanent degradation. These results indicate that Ga2O3 have comparable or even more potential than Si and typical widegap semiconductors SiC and GaN for power device applications. R&D works on InGaN LEDs on Ga2O3 substrates that a collaborative company Tamura Corporation is conducting are also introduced briefly at the end of this seminar.

Posted in Events | Comments Off

ECE Professor Yasamin Mostofi’s “X-Ray Vision for Robots with Only WiFi” research highlighted in UCSB’s The Current

youtube video
ECE Professor Yasamin Mostofi and graduate student researchers, Saandeep Depatla and Lucas Buckland, enable robots to see through solid walls with Wi-Fi

Imagine unmanned vehicles arriving behind thick concrete walls. They have no prior knowledge of the area behind these walls. But they are able to see every square inch of the invisible area through the walls, fully discovering what is on the other side with high accuracy. Now, imagine robots doing all these with only WiFi signals and no other sensors.

Read UCSB’s The Current article “Now You Can See the Invisible” to learn more about Mostofi’s x-ray vision and wi-fi research.

Posted in News, Press Releases | Comments Off