Skip to Content

CSE News

  • CSE Closes Out Successful Hiring Season for 2016-2017

    With a final flurry of announcements, CSE Chair Dean Tullsen marked the end of the current faculty hiring season. He did so in confirming that the last two outstanding offers were accepted. That brings CSE's total new appointments to seven faculty members, as well as two announced in 2015 who are starting this fall.  "The last two appointments officially close the recruiting cycle, and there is no question that it was a great year for recruiting," said Tullsen. "These new faculty members will bolster some of our core strengths and enable us to establish some significant new directions.”

    He added: “This will expand our research focus and enable us to better address the tremendous demand for a computer science education." The CSE chair went on to list many of the people involved in the recruiting effort (see final paragraph at bottom).

    Among the new faculty hires for 2016-2017, CSE previously announced the appointment of Henrik Christensen (right), who is leaving Georgia Tech to be the faculty Director of UC San Diego's new Contextual Robotics Institute (part of which is located initially in the Qualcomm Institute). Christensen earned his Ph.D. from Aalborg University in 1990 in his native Denmark. He does research on robotics, computer vision and artificial intelligence, and is considered one of the leading roboticists in the world. For more on Christensen, see UC San Diego's July announcement (http://www.cs.ucsd.edu/node/3012).

    In addition to Christiensen, CSE is also beefing up its expertise in robotics with the appointment of Laurel Riek (left) as an Associate Professor in the department. Riek is currently the Clare Boothe Luce Assistant Professor at the University of Notre Dame.  She received her Ph.D. in Computer Science from the University of Cambridge, and B.S. in Logic and Computation from Carnegie Mellon University. She designs autonomous robots capable of sensing, responding and adapting to human behavior. Riek's research enables robots to robustly solve problems in real-world, safety-critical human environments, such as hospitals, homes, and factories. Her work tackles fundamental and applied problems that make complex, real-world perception and interaction in these spaces so difficult for machines, and her work has applications in manufacturing, neurorehabilitation, and emergency medicine. In addition to robotics, Riek's research interests include human activity understanding, and healthcare engineering. She has received the NSF CAREER Award, AFOSR Young Investigator Award, a Qualcomm Research Scholar Award, and several best-paper awards. In 2014, she was named one of the American Society for Engineering Education's 20 Faculty under 40.

    Ndapa Nakashole (right) has been appointed an Assistant Professor in CSE, with an official start date of January 2017. She joins UC San Diego from a three-year postdoctoral fellowship in machine learning at Carnegie Mellon. Nakashole's research interests include machine reading, natural language processing, machine learning and data mining. She uses machine learning to build computer systems that intelligently process and understand human language. She received her Ph.D. from Germany's Saarland University, and her M.S. and B.S. degrees are from the University of Cape Town (South Africa).

    Melissa Gymrek (left), has joint appointments in CSE and the Department of Medicine, and she will split her time between an office in the CSE building and her lab in the School of Medicine. Gymrek is known for her work connecting the dots between genetic variations in humans, traits and human health, and she is CSE's first recruit under the Precision Medicine Initiative launched by the campus in 2015. Gymrek completed her Ph.D. in Bioinformatics and Integrative Genomics in the Harvard-MIT Division of Health Sciences and Technology. She is joining UC San Diego from a postdoc in the Analytical and Translational Genetics Unit at Massachusetts General Hospital and the Broad Institute of MIT and Harvard.

    CSE's Database group is getting a new member with the pending arrival of Arun Kumar (right), who joins the department this fall as an Assistant Professor. He works at the intersection of databases and machine learning with a focus on problems related to usability, developability, performance and scalability. Kumar recently completed his Ph.D. at the University of Wisconsin-Madison, where he won the annual Wisconsin CS Graduate Student Research Award for his dissertation research. Other awards include an ACM SIGMOD Best Paper, and an invited paper at ACM TODS. Kumar builds systems and tools, among them: the Hamlet project, which grew out of a machine-learning paper titled "To Join or Not to Join"; as well as project Santoku and Orion which optimize various machine-learning models over normalized data. Some of his tools are also available through the open-source library, MADlib, and there has been substantial tech-company interest in industrial adoption of ideas from his thesis work. Facebook, LogicBlox and Microsoft had adopted some of the ideas from his thesis work, while Cloudera, EMC and Oracle have adopted some components of his other research. Kumar is now on campus, but he will begin teaching (with a graduate-level research seminar) in the winter quarter.

    Assistant Professor Aaron Schulman (left) will also begin teaching in the winter quarter. He officially started July 1, and is transitioning to be full-time in San Diego this November after moving from his current postdoctoral research job at Stanford. Schulman becomes a member of the Systems and Networking group in CSE, but his wide-ranging interests go beyond computer systems and networking to include operating systems, security and even embedded systems. He has improved the efficiency of wireless networks, cellular network flexibility, and the energy efficiency of mobile applications. Schulman also quantifies residential Internet network reliability, made progress in securing the web’s public key infrastructure, and identified privacy leaks in mobile devices. Schulman earned his Ph.D. in Computer Science from the University of Maryland in 2013 (and he subsequently received the SIGCOMM Doctoral Dissertation Award for his doctoral thesis on "Observing and Improving the Reliability of Internet Last-Mile Links"). Schulman also founded Mellow Research, LLC, a startup that markets BattOr, a power monitor for app developers he invented.

    Joe Gibbs Politz (right) joined the UC San Diego faculty in 2016. He worked previously as a visiting instructor at Swarthmore College. Politz studies computer science education, programming languages, compiler design, web programming, and web security.  He has recently had two complementary areas of focus: using peer code review in undergraduate courses, and developing the programming language Pyret for use in computer science curricula from middle school to the undergraduate level (as part of the Bootstrap curricular outreach effort to bring programming ideas to courses such as math and physics at middle and high school level -- targeting a broader base of students with computer science topics.) Politz teaches students to test, document, and explain both their own code and the code of others. Politz received his Ph.D. in computer science from Brown University in May 2016, and his B.S. in computer science from Worcester Polytechnic Institute in 2009.  He has also worked on web security at Google through software engineering internships.

    Other New Arrivals

    Although their appointments were announced a year ago by the department, Assistant Professors Deian Stefan and Manmohan Chandraker requested delays before starting this fall. 

    Professor Manmohan Chandraker (left) worked at NEC Labs America in Cupertino in computer vision before returning to UC San Diego, where he previously completed his Ph.D. in Computer Science in 2009. He did his dissertation under CSE Prof. David Kriegman, and received the 2009 CSE Best Dissertation Award. Subsequently he was a postdoctoral scholar at UC Berkeley in the lab of Prof. Ravi Ramamoorthi, who later moved to CSE at UC San Diego, where he now directs its Center for Visual Computing. Chandraker is now affiliated with Ramamoorthi's center as well as the Contextual Robotics Institute, where his research focuses on self-driving cars and helping robots navigate the real world. The computer vision expert converts 2D images to 3D maps of the world that allow self-driving cars to pinpoint the location of traffic participants such as pedestrians and vehicles. The data are combined with information about the car's surroundings such as lanes, roads and traffic signs. The ultimate goal is to understand the scene, especially in crowded environments. His work on shape recovery from motion cues for complex material and illumination received the Best Paper award at Computer Vision and Pattern Recognition in 2014.

    Deian Stefan (right) accepted his appointment as Assistant Professor last year and deferred his arrival to make time to work on his web security startup company, Intrinsic (formerly GitStar), which provides developers with tools for deploying web applications with minimal trust. The company builds on Stefan's prior research on confinement and information-flow control. His research spans systems, security and programming languages, but he is also interested in building principled and practical secure systems (the subject of Stefan's dissertation for his Ph.D. from Stanford University completed in 2015). (See CSE news release about Deian Stefan: http://www.cse.ucsd.edu/node/2841.)

  • Computer Science at UC San Diego Ranks #14 Worldwide

    In its 2016 university rankings, the Academic Ranking of World Universities (ARWU) for the fourth year in a row ranked UC San Diego the #14 university in the world. At the same time, ARWU used its methology to analyze and rank many of the most important disciplines, including computer science, and CS at UC San Diego came away with an identical ranking, making it the #14 computer science program globally.

    It would have jumped higher if not for CSE receiving no points in two categories that favor older and wealthier departments. Points in the two categories were awarded to campuses with alumni or faculty who have won a Turing Award since 1971. (In other disciplines, Nobel Prizes are required, e.g., in economics, chemistry or physics, while mathematicians must have received a Fields Medal.) As a result, those Turing-dependent categories carry 25 percent of the weight in assessing the strength of computer science, and UC San Diego started with zero.

    On the other hand, CSE did much better on another weighted category, involving highly-cited researchers in computer science. Indeed, its score in this category pegs the UC San Diego program at #3 in the world (after Stanford and MIT) if you look only at highly-cited researchers. The two final categories involve the overall number of computer science research articles, and the percentage of papers published in the top-20% of journals in computer science. On both counts, UC San Diego fared better than many of the high-ranked universities. UC San Diego faculty published more than UCLA, Princeton, Harvard, Cornell or Caltech. For the percentage of papers in top-20% CS journals, CSE did better than UIUC, UT Austin, University of Toronto, USC and Carnegie Mellon -- even though all of those universities placed higher than UC San Diego in the overall computer science rankings. Indeed, UC San Diego has the highest-ranked computer science program of all universities with zero Turing Award winners among faculty and alumni.

  • Barngrover, Hoover Hope to Take Robotics Sequence to Next Level

    CSE lecturers Christopher Barngrover (near right) and Greg Hoover are looking for more M.S. students to take their CSE 291 three-course sequence, which kicks off this fall with Barngrover's Introduction to Robotics Software course. (It's a graduate-level course, but Barngrover says they have allowed upper-class undergraduates in under special circumstances.) It was launched last fall as part of the UC San Diego Master's Program in CSE. Students who wanted to go on to take the final Robotics Project course were required to have taken either the systems course taught by Hoover or the software course with Barngrover. The first project course was taught in Winter 2016 by both Hoover and Barngrover jointly, and the project teams were formed to reflect the students' relative experience in either hardware and systems or software skills.

    "The three courses were designed to teach fundamental hardware and software design skills necessary for building robotic systems," said Barngrover, who is also a research scientist at the Space and Naval Warfare (SPAWAR) Systems Center Pacific working on robotic control and perception in its Unmanned Systems Group. The lecturer is also a CSE alumnus (M.S., Ph.D. '10, '14). "The project course challenged teams of students to complete targeted goals. At each challenge, students had to apply their hardware and software skills as well as their ingenuity to create robotic systems performing specific tasks."

    Hoover's systems course teaches hardware/software interface via integration with sensors, peripheral devices, and the iRobot platform. Students can expect to gain hands-on experience with microcontroller architecture, I/O, communication protocols, task management, bare-metal programming in C, embedded Linux, and more. Barngrover's software course provides a foundation in robotic software frameworks with a focus on the Robot Operating System (ROS) on both laptop and embedded environments. Students learned about modeling, simulation, perception, controls, path planning, route execution, and more, and they gained hands-on experience with the framework in an embedded system with active peripherals. (To view student team presentations at end of Fall 2015 quarter, click here.)

    Over last winter, the Robotics Project course involved the development, construction and testing of autonomous vehicles (pictured) that would be able to compete in a final competition. Each  team was given the same hardware and guidance. The goal: to find, collect and return balls of different colors.

    Each team's robot had to pick up and carry each ball back and drop the ball into the team's bucket in order to gain points. To make the task more interesting, the number of points varied according to the color of the retrieved ball (orange balls were worth twice the points of green ones).  At the end of the winter course, five teams competed for points, and all of each robot's movements had to be done autonomously (see CSE 291 final competition on YouTube). The only exception to the autonomy rule: each team was allowed a set number of times when a team member could 'reset' the robot back on the course (a flexibility that most of the teams took advantage of during the early part of the final competition). All resets, however, came with a minimum 3-minute penalty to make any repairs and return the robot to the starting line.

    All of the teams participating in the winter course came with intriguing titles: Coconuts; SpeedyDug, Double-O-One, Popsicles, and Hand of ROS. "Each team also maintained a wiki on a private bitbucket account where we could see their code and status," noted Barngrover.

    Now that the inaugural series of courses is behind them, the lecturers hope to expand the number of graduate students taking the sequence. They also anticipate that expanded enrollment could make it easier to get approval for a new 'concentration' in Robotics and Vision in connection with CSE's M.S. program.

  • Researchers Devise Proactive Method for Detecting Hardware Trojans at the Gate-Level

    Modern computer chips are made up of hundreds of millions – often billions – of transistors. Such complexity enables the smartphone in your back pocket to perform all manner of powerful computations, but it also provides lots of places for tiny malicious circuits, known as hardware Trojans, to hide. Magnifying this security risk is the increasingly distributed and globalized nature of the hardware supply chain, which makes it possible for a Trojan to be introduced at any point along the way.

    To prevent, detect and combat these hardware Trojans, computer scientists from UC San Diego, together with their collaborators, have devised a new technique that tracks information flow through a circuit’s logic gates, much the way one would track traffic as it flows through an intersection while obeying a series of traffic signals. If information unexpectedly moves to a part of the chip where it shouldn’t be, the method will determine that a security violation occurred, and whether or not a Trojan was the root cause.

    The technique is described in a paper titled “Detecting Hardware Trojans with Gate-Level Information-Flow Tracking,” which is part of the cover story on supply-chain security in cyberinfrastructure in the August 2016 edition of IEEE Computer. The paper’s authors are CSE postdoctoral researcher Wei Hu; his advisor CSE Prof. Ryan Kastner; CSE alumnus Jason Oberg (M.S., Ph.D. '12, '14), who is now CEO of Tortuga Logic, a company he co-founded with Prof. Kastner; and Ph.D. candidate Baolei Mao of Northwestern Polytechnical University. (Pictured l-r: CSE-affiliated co-authors Wei Hu, Ryan Kastner and Jason Oberg. Missing: Baolei Mao, a visiting grad student in Kastner's group from 2013-2015.)

    “Trojans are designed specifically to avoid activation during testing,” explains Kastner, who is head of the Kastner Research Group at UC San Diego and an affiliate of the university’s Qualcomm Institute. “Hardware designs are complex and often consist of millions of lines of code. The standard rule is to expect one ‘bug’ per five lines of code. People with bad intentions – say, a disgruntled employee – can insert these special ‘bugs’ into sequence patterns that are very unlikely to be tested, where they lie dormant and wait for a rare input to happen and then they trigger something malicious, like draining your phone’s battery or stealing your cryptographic key,” (i.e., the key that encrypts sensitive information to keep it secure).

    “The concern these days is that chips are designed and manufactured all over the world, and sometimes in countries that might have a reason to steal intellectual property or other information,” Kastner adds. This concern is so great in the United States, in fact, that government-sensitive technologies are fabricated in trusted foundries (semiconductor fabrication plants) that require security clearance.

    But, notes Kastner, “typically these foundries are not as advanced and not as cheap as those in other countries. Sometimes they’re using technologies that are three- or four-generations old. The hope is that we can continue to send hardware to be manufactured at any foundry, and that this method will make the process more secure.”

    The method uses a technique called GLIFT (gate-level information flow tracking), which works by assigning a label to important data in a hardware design.  If the goal, for instance, is to understand where information about a cryptographic key is flowing, a “confidential” label would be assigned to bits of the key. The test engineer would then write a formal property that asserts that that any confidential information (in this case the key) will be constrained to stay in secure part of the chip. If the key flows outside of that secure area, then the hardware is capable of being compromised.

    Kastner says the previous methods for finding Trojans were mostly statistical and tried to pinpoint inconsistencies and variations in measurable properties in the circuit that would indicate a Trojan, such as how much time it should take to complete a function or how much power it should consume. Because these methods are statistical, they are also susceptible to noise. Smaller Trojan circuits, therefore, are easier to hide in large designs. “It’s like trying to find a needle in a haystack,” says Kastner.

    “The state of the art right now is teams at Qualcomm or Intel, for example, manually inspecting hardware code and the physical characteristics of the chip to determine what they think could happen,” he adds. “It’s a terribly imprecise process, and you could easily overlook a small error which could have large consequences.”

  • Former CSE Chair Looks Back at Enrollment, Student-Faculty Ratio and 'Strong Fundamentals'

    In what may be viewed as a valedictory interview after relinquishing the reins of the CSE department following six years as its Chair, Professor Rajesh Gupta sat down to speak with the San Diego Union-Tribune for a look back at how far the department has come. Science editor Gary Robbins notes that Gupta helped turn the UC San Diego undergraduate computer science program into the largest of its kind in the country.

    He begins the article noting that Gupta brought in a donation of $18.5 million from an anonymous CSE alumnus, allowing Gupta to correct a design flaw in the CSE building when it opened in 2005: the undergraduate computer labs in the building's basement did not link directly to the other floors of labs, seminar rooms and faculty offices, "making it hard for students on the bottom floor... to conveniently mingle with the faculty members working above them," wrote Robbins. Armed with the largest gift ever from a UC San Diego alum, the department is now correcting the problem -- at a cost of roughly $3 million -- and expanding space and facilities on the ground floor. 

    In the interview published on Aug. 11 -- more than a month after Gupta stepped down as Chair as CSE professor Dean Tullsen assumed the job -- Gupta noted that UC San Diego calls itself a student-centered university. "That doesn't happen if you don't put students in the center of things, which has been a problem in our computer science building," said Gupta. "But we're fixing that." Asked about CSE's "explosive growth" in enrollment, especially at the undergraduate level, Gupta confirmed that "the faculty revolted" in response to high teaching loads (at a 44-to-one student-faculty ratio). "We brought in a lot of non-tenure-track lecturers to help, and they're wonderful. But that is only a temporary fix," explained Gupta. "We're now reducing enrollment and increasing the faculty size."

    In response to a question about demand for master's degree programs in computer science, Gupta looked back to when he became Chair in 2010. "We had about 90 or so students in the master's program. This year, the figure was 350," he explained. "This fall, it will go to 450. In six years, the demand for master's degrees has gone up six times -- to 3,500 applicants. Students want the degree because it gives them more specialized knowledge in areas like machine learning, vision and computer systems. It places them in a better position to get jobs." Gupta admits that some faculty worry about taking attention away from Ph.D. students and their research. But he noted that there is a silver lining. "The good news is that [master's programs] account for over 50 percent of the growth in women applicants," said Gupta, "thus helping us in improving diversity in our graduate programs."

    Asked about what corporations say they need from the department, Gupta said "the main thing I hear from industry is, 'Give us people with strong fundamentals'. The fundamentals include the ability to express yourself clearly and to work in groups and teams that promote achievement."

  • CSE, ECE and QI Collaborate on Open Speech Platform

    A research scientist in the Qualcomm Institute, Hari Garudadri (at left)  is principal investigator on a more than $2 million project funded by NIH's National Institute on Deafness and Communication Disorders (NIDCD) that involves a team of researchers and graduate students as well as externl collaborators, including from San Diego State University. The team of engineers from UC San Diego and audiologists from SDSU has set out an ambitious timetable for delivering two new electronic platforms to dramatically improve and accelerate research on better hearing aids. Announced this week at the International Hearing Aid Research Conference in Lake Tahoe, Calif., the Open Speech Platform project at UC San Diego is one of the six groups awarded funding under the NIDCD Open Speech program to develop open-source solutions and development platforms to make it easier for researchers to develop enhanced algorithms and techniques to improve the intelligibility of hearing aids in widespread use today. The UC San Diego-based project will develop a 'real-time, open, portable, extensible speech lab' to be used by audiologists and technology developers. The first platform will be released in early 2018 in a desktop form factor (see image), and by the end of 2017, the researchers will shrink the system for use as a wearable, chip-based embedded system that could be used ito take hearing-aid testing into the field.

    While the two co-PIs on the Open Speech Platform project are ECE faculty members (Patrick Mercier and Bhaskar Rao), CSE is also involved on several counts. CSE Prof. Rajesh Gupta is a collaborating investigator whose research is focused on energy efficiency and mobile computing issues in embedded systems (currently in connection with the NSF-funded RoseLine project on which Gupta is PI). Also representing CSE on the research project is Ph.D. student Sean Hamilton. Hamilton still works part-time as a software engineer at Tortuga Logic, a chip design security startup founded by CSE Prof. Ryan Kastner and CSE alumnus Jason Oberg (M.S., Ph.D. '12, '14). .(Pictured: CSE Prof. Gupta, and Ph.D. student Hamilton.)

    Read the full news release on the Qualcomm Institute website.

  • ThoughtSTEM Lands $750K SBIR Grant; CTO Moves to GitHub

    Another big win for ThoughtSTEM, the San Diego company launched by two recent Ph.D. alumni of the CSE department. The computer science education startup has been awarded a $750,000  grant from NSF's Small Business Innovation Research program to expand use of ThoughtSTEM's LearnToMod software that helps teach kids how to modify ("mod" in the lingo) elements of the popular Minecraft game environment. According to ThoughtSTEM CEO Stephen Foster (Ph.D. '14), LearnToMod has taught computer science to over 50,000 students, and those students used the software to produce over 1.5 million Minecraft mods.

    "The potential impact that Minecraft could have on computer science education in this country is huge," said Foster (above). "Millions of kids who love Minecraft are interested in learning how to mod. With the National Science Foundation's help, we have big plans to make LearnToMod the most cutting-edge platform for CS education."

    While students can mod on their own, Foster said that over 2,000 educators worldwide are already using LearnToMod in their classrooms. Kids who love Minecraft can come up with an idea to change and improve the gameplay by using an easy, drag-and-drop programming interface. Since the software's launch in 2015, ThoughtSTEM received NSF funding to craft new tutorials for teaching students how to mod in Javascript, and Foster created new tools to assist teachers using LearnToMod in classrooms. With the new two-year grant from NSF, ThoughtSTEM developers aim to make it even easier for students and teachers to start using the platform for Minecraft modding and computer science education in general. LearnToMod is an online integrated development environment (IDE), software that provides all the tools necessary to mod Minecraft, including unlockable badges, a free Minecraft server (where students can test their mods using the Vox-L game engine), tutorials, and an online community.

    Meanwhile, ThoughtSTEM co-founder and CSE alumna Sarah Guthals (Ph.D. '14) is now 'emeritus' at the company, but taking her expertise to a new frontier: working with GitHub's Social Impact team. Guthals (at right) is designing, building and testing a new GitHub product that will engage young children online "in really neat ways." It's an online space for under-13-year-old children to build, share, collaborate and contribute to digital artifacts. She joined GitHub as an independent contractor in April, after stepping away from her previous role as chief technology officer for ThoughtSTEM. Earlier this year, Guthals was named to Forbes magazine's latest annual list of 30 Under 30 in Science. 

  • Crowdsourcing the Transformation of Mass Spectrometry Big Data into Scientific Living Data

    In a landmark paper published in the August issue of Nature Biotechnology*, 127 scientists from a consortium of universities and research labs in the U.S. and worldwide report for the first time on the establishment of an online, crowdsourced knowledge base and workbench that could be a game-changer for the study of natural products that could potentially be useful in the development of the next antibiotic, better pesticides, or more effective cancer drugs.

    “The potential of the diverse chemistries present in natural products remains untapped because natural product databases are not searchable with raw data and the research community has no way to share data other than through published papers,” the paper noted. “Mass-spectrometry (MS) techniques are well-suited to high-throughput characterization of natural products, so there is a pressing need for an infrastructure to enable sharing and curation of the data.”

    [Pictured l-r: CSE and Skaggs School of Pharmacy professor Nuno Bandeira; CSE Ph.D. student Mingxun Wang; and Skaggs School professor Pieter Dorrestein.]

    Enter Global Natural Products Social Molecular Networking (GNPS), an open-access knowledge base for community-wide organization and sharing of raw, processed or identified tandem mass spectrometry data.  The platform has operated in beta mode since 2014 under the leadership of the UC San Diego-based Center for Computational Mass Spectrometry (CCMS) and the Collaborative Mass Spectrometry Innovation Center in the Skaggs School of Pharmacy and Pharmaceutical Sciences, also at UC San Diego.

    To date, more than 13,600 researchers from 115 countries have used the GNPS platform, utilizing various features and tools developed by CCMS to enhance or accelerate how researchers handle MS data (by allowing researchers to visualize/group the relationships of related molecules over petabyte-scale data spanning over one million samples). As the user community expands, so will its impact on science: so far, more than 100 different scientific papers have cited their use of the platform, with many different laboratories using it.

    Since its launch, researchers have been developing the service to meet the needs of a broad population of scientists interested in natural products, and not just chemists.  “We’re facilitating making natural product chemical structures accessible to scientists who don’t necessarily have training in chemistry,” said Skaggs School of Pharmacy professor Pieter Dorrestein, one of the lead authors of the Nature Biotechnology paper. “GNPS has created a community that is now willing to share knowledge, and we’re enabling that community to do so, and this is unique in the natural products community.”

    To capture the knowledge of the community, the authors have created a crowdsourced, wiki-like knowledgebase that accepts submissions and revisions of annotated MS data from any scientist. Embracing a spirit of sharing and openness, these submissions are immediately and freely available to the entire community to be used to search MS data. “Just as BLAST finds regions of similarity between biological sequences in the form of nucleotide or protein sequences, the search function in GNPS serves a similar purpose,” said CCMS Executive Director Nuno Bandeira, a UC San Diego professor with joint appointments in Computer Science and Engineering as well as the Skaggs School of Pharmacy, lead author of the Nature Biotechnology paper. “It allows researchers to find identical or similar molecular structures by comparing the chemistry based on spectral signatures captured using mass spectrometry.”

    Why the focus on natural products? Natural product scientists typically do not yet take full advantage of the modern capabilities of mass spectrometry.

    “GNPS is the equivalent of BLAST, but for searching, analyzing and storing chemical signatures of molecules,” said Dorrestein.  “And like BLAST, GNPS is a tool that can be used by many different scientific communities.”

  • Faculty-Affiliate on Cross-Cultural Multimedia Computing

    Shlomo Dubnov is a computer scientist by training (his Ph.D. is from Hebrew University), but his primary appointment at UC San Diego is in the Department of Music. But Dubnov is also an affiliated faculty member in the Computer Science and Engineering department, and his latest work has just been published by the German publishing company Springer as part of its SpringerBriefs in Computer Science series. Titled "Cross-Cultural Multimedia Computing", the book carries the subtitle, "Semantic and Aesthetic Modeling",

    Dubnov and his two co-authors -- MITRE Corp. cognitive scientist Kevin Burns and Japanese electrical engineer Yasushi Kiyoki -- explore ongoing research on computational modeling of visual, musical and textual content, which are described in terms of identifying and mapping their semantic representations across different cultures. They also recap experimental studies attempting to characterize preferences for complexity in abstract, classical and traditional art and music across samples of Western and Far Eastern cultures. The experiments illustrate how aesthetics can be computed in terms of semantic and information measures, highlighting commonalities and uncovering differences in aesthetic preferences across individuals and cultures (highlighted in the varying backgrounds of the three authors).

    Dubnov's work on computational modeling of style and computer audition has led to development of several computer music programs for improvisation and machine understanding of music. He currently directs the Center for Research in Entertainment and Learning (CREL) in Calit2's Qualcomm Institute, and serves as a lead editor of the journal ACM Computers in Entertainment.

    The softcover issue is now on sale, but Springer says that an eBook version of the title will be available soon (ISBN 978-3-319-42873-4).  Professor Dubnov previously teamed with co-author Kevin Burns on an earlier Springer book from its computer science series on database management and information retrieval: they co-edited (with Shlomo Argamon) "The Structure of Style: Algorithmic Approaches to Understanding Manner and Meaning," published in 2010.

  • NOVA Paves the Way for Storage Class Memory File Systems

    Intel expects to start selling 3D Xpoint storage class memory (SCM) before the end of 2016 in the form of its Optane solid-state device (SSD). But experts believe the real payoff from SCM will come when systems connect the SCM directly to the processor, yielding hybrid memory systems that include volatile and non-volatile memories. Computer engineers in CSE and the Center for Networked Systems (CNS) at UC San Diego anticipate that these new memory systems will provide software with sub-microsecond, high-bandwidth access to persistent data.

    But there's a catch: According to CSE Prof. Steven Swanson (far right) and Ph.D. student Jian Xu (near right), existing file systems built for spinning or SSDs create software overheads that can obscure the gains expected from SCM systems.

    Enter NOVA, a log-structured file system proposed by Xu and Swanson in a paper* delivered earlier this year at the USENIX Conference on File and Storage Technologies (FAST) in Santa Clara, and a variation presented at UC San Diego's 2016 Non-Volatile Memory Workshop (organized by Swanson). The proposed NOVA system is designed to maximize performance on hybrid volatile/non-volatile memory systems, while providing strong consistency guarantees.  In the paper’s abstract, Xu and Swanson explained, "NOVA adapts conventional log-structured file system techniques to exploit the fast random access that SCMs provide. In particular, it maintains separate logs for each inode to improve concurrency and stores the file data outside the log to minimize log size and reduce garbage collection costs."

    The system's logs therefore provide metadata, data, and mmap atomicity and focus on simplicity and reliability, keeping complex metadata structures in DRAM to accelerate lookup operations. "NOVA's multi-log design achieves high concurrency, efficient garbage collection and fast recovery," said Xu in his presentation to the FAST audience. "NOVA also outperforms existing file systems while providing stronger consistency and atomicity guarantees." Indeed, experimental results showed that in write-intensive workloads, NOVA provides 22% to 216x throughout improvement compared to state-of-the-art file systems, and 3.1x to 13.5x improvement compared to file systems that provide equally strong data consistency guarantees.

    "Upcoming [SCM] technologies... promise faster writes, higher endurance and greater longevity than the NAND flash used in today's SSDs," wrote storage expert Robin Harris in reviewing the NOVA paper for his blog, StorageMojo. "Systems won't be able to take advantage of [SCM] technology until their DRAM and disk I/O stacks are re-engineered for the specific advantages and quirks of [SCM]."