BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Information Systems Group - ECPv6.4.0.1//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Information Systems Group
X-ORIGINAL-URL:https://isg.ics.uci.edu
X-WR-CALDESC:Events for Information Systems Group
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:20200308T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20201101T090000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20200414T110000
DTEND;TZID=America/Los_Angeles:20200414T120000
DTSTAMP:20261108T020004
CREATED:20200425T025505Z
LAST-MODIFIED:20200425T025505Z
UID:1294-1586862000-1586865600@isg.ics.uci.edu
SUMMARY:Babak Salimi: Causal Inference for Responsible Data Science
DESCRIPTION:ABSTRACT: \nScaling and democratizing access to big data promises to provide meaningful\, actionable information that supports decision-making. Today\, data-driven decisions profoundly affect the course of our lives\, such as whether to admit applicants to a particular school\, offer them a job\, or grant them a mortgage. Unfair\, inconsistent\, or faulty decision-making raises serious concerns about ethics and responsibility. For example\, we may know that our training data is biased\, but how do we avoid propagating discrimination when we use this data?  How do we avoid incorrect\, spurious and non-reproducible findings? How can we curate and expose existing data to make it “safe” for informed decision making? \nIn this talk\,I describe how we can combine techniques from causal inference and data management to develop systems and algorithms that help answer some of these questions.  Many existing popular notions of fairness in ML fail to distinguish between discriminatory\, non-discriminatory and spurious correlations between sensitive attributes and outcomes of learning algorithms. I present a new notion of fairness that subsumes and improves upon previous definitions and correctly distinguishes between fairness violations and non-violations. Further\, I describe an approach to removing discrimination by repairing training data in order to remove the effects of any inappropriate and/or discriminatory causal relationships between a protected attribute and classifier predictions.  Analytical SQL queries supported by mainstream business intelligence and analytics environments can lead to perplexing observations and incorrect business decisions. I describe a system that automatically rewrites analytical SQL queries into complex causal queries that support decision-making. \nBio \nBabak Salimi is a postdoctoral research associate in Computer Science & Engineering at the University of Washington\, Seattle\, where he works with Prof. Dan Suciu and the Database Group. He received his Ph.D. from the School of Computer Science at Carleton University in Ottawa\, Canada and his M.Sc. in Computation Theory (2009) and B.Sc. in Computer Engineering (2006) from Sharif University of Technology and Azad University of Mashhad\, respectively. Salimi’s research interests span data management\, causal inference\, decision-making systems\, algorithmic fairness and responsible data science. \n\nZoom Link: https://uci.zoom.us/j/232157494
URL:https://isg.ics.uci.edu/event/babak-salimi-causal-inference-for-responsible-data-science/
LOCATION:https://uci.zoom.us/j/232157494
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20200417T123000
DTEND;TZID=America/Los_Angeles:20200417T140000
DTSTAMP:20261108T020004
CREATED:20191118T211239Z
LAST-MODIFIED:20200117T182717Z
UID:1228-1587126600-1587132000@isg.ics.uci.edu
SUMMARY:David Lomet: Better Database Cost/Performance via Programmable SSD Batched I/O
DESCRIPTION:Abstract: A database storage manager should place data at the most cost/performance-effective tier in the storage hierarchy.  While performance and cost both decrease with distance from the CPU\, the cost/performance trade-off depends on how efficiently a storage manager can move data across tiers.  Log structuring (LS) is designed to improve the cost/performance of secondary storage by writing batches of pages from main memory to secondary storage when using a conventional block-at-a-time I/O interface.  The advent of programmable SSDs changes the game.  A storage manager no longer has to work within the limitations of the block-at-a-time interface.  In this paper\, we re-program an SSD controller’s flash translation layer into one that enables a storage manager to batch pages without itself needing to implement LS.  This removes the CPU computational and I/O load associated with LS recovery and garbage collection.  We compare the Bw-tree key-value store with its LLAMA host-based LS to the same key-value software stack but executing on our batch-interface SSD.  Our experimental results show the benefits of eliminating redundancies\, minimizing interactions across storage layers\, and avoiding the CPU cost of providing LS. \nShort Bio: \nDavid Lomet founded the Database Group at Microsoft Research Redmond in 1995 and managed it for 20 years. His research career began at IBM where\, while on a 1975-76 sabbatical at the University of Newcastle-on-Tyne\, he invented atomic actions (a form of transactions).  He later worked at Wang Institute as a faculty member\, and at Digital Equipment Corporation as a software architect and research staff member. He received a Ph.D. in computer science from the University of Pennsylvania. \nLomet’s primary focus has been the engineering of database systems\, with a focus on database system kernels.  His work on concurrency control and recovery contributed to making DEC’s Rdb and Microsoft’s SQL Server database management systems leaders in cost/performance. His Deuteronomy research project’s latch-free Bw-tree index and log structure store are key elements in Microsoft’s Hekaton main memory database and Azure Cosmos DB cloud data service.  Deuteronomy won the Microsoft Research Redmond “2017 Best Research Project” Award.  Lomet is an author of over 120 papers and over 60 patents. \nLomet has won IEEE awards as well as the ACM SIGMOD Contributions Award and the Computer Society TCDE Service Award  for his 25 year tenure as EIC of the IEEE Data Engineering Bulletin.  HE has also served as editor of ACM TODS\, VLDB Journal and others\, and has been a member of the VLDB Board.  He has been a PC co-chair for ICDE and VLDB.  He is a member of the IEEE Computer Society Board of Governors and society Secretary\, and has been First Vice President and Treasurer.  He is a fellow of IEEE\, ACM\, and AAAS\, and a member of the National Academy of Engineering. \n  \nLunch provided.
URL:https://isg.ics.uci.edu/event/david-lomet-tbd/
LOCATION:DBH 3011
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20200420T110000
DTEND;TZID=America/Los_Angeles:20200420T120000
DTSTAMP:20261108T020004
CREATED:20200222T010700Z
LAST-MODIFIED:20200222T010700Z
UID:1285-1587380400-1587384000@isg.ics.uci.edu
SUMMARY:Redesigning Storage Systems for Future Workloads\, Hardware\, and Performance Requirements (CS Faculty Candidate Seminar)
DESCRIPTION:Speaker: Oana Balmau (University of Sydney) \nAbstract: Cloud storage stacks are being challenged by new workloads\, new hardware and new performance requirements. First\, workloads evolved from following a read-heavy pattern (e.g.\, a static web-page) to a write-heavy profile where the read:write ratio is closer to 1:1 (e.g.\, as in the Internet of Things). Second\, the hardware is undergoing rapid changes.  The divide between fine-grained volatile memory and slow block-level storage is rapidly being bridged by the emerging byte-addressable non-volatile memory devices and the fast block-addressable NVMe SSDs (e.g.\, Intel Optane NVMe SSDs). Third\, performance requirements in storage systems now emphasize low tail latency\, in addition to high throughput.  \nIn this talk I will argue that existing storage systems have fundamental limitations that do not allow them to fully meet these challenges\, and that therefore the storage stack needs to undergo radical change. In particular\, using state-of-the-art key-value stores I will demonstrate that with modern workloads and hardware the bottleneck shifts from I/O to CPU\, invalidating an assumption that has underpinned all past storage system design. In line with this observation I will then present a new design paradigm for key-value stores that departs from the conventional wisdom of optimizing disk usage and instead optimizes CPU usage. To do so\, we keep data unsorted on disk\, reduce contention for shared data structures\, and do away with expensive maintenance operations. \nThis design has been implemented in the KVell key-value store. KVell outperforms state-of-the-art key-value stores such as RocksDB in both read- and write-heavy workloads\, running on modern NVMe SSDs.  Thanks to its novel design\, KVell achieves up to 5x better throughput\, and up to two orders of magnitude lower tail latency. \nBio: Oana Balmau is a PhD candidate in the School of Computer Science at the University of Sydney\, advised by Prof. Willy Zwaenepoel. She earned her Bachelors and Masters degrees in Computer Science from EPFL\, Switzerland. Her research interests are computer systems and storage technologies. Currently\, she is working towards redesigning storage systems\, aimed at future workloads and hardware. (Homepage: oanabalmau.com)
URL:https://isg.ics.uci.edu/event/redesigning-storage-systems-for-future-workloads-hardware-and-performance-requirements-cs-faculty-candidate-seminar/
LOCATION:DBH 3011
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20200427T110000
DTEND;TZID=America/Los_Angeles:20200427T120000
DTSTAMP:20261108T020004
CREATED:20200424T192059Z
LAST-MODIFIED:20200424T192059Z
UID:1291-1587985200-1587988800@isg.ics.uci.edu
SUMMARY:Lei Cao: Toward an End-to-end Anomaly Discovery Paradigm
DESCRIPTION:ABSTRACT: \nAnomaly detection is critical in enterprises\, with applications ranging from preventing financial fraud\, and defending network intrusions\, to detecting imminent device failures. \nAlthough previously developed research offers a plethora of stand-alone methods for detecting particular types of anomalies\, there is no end-to-end solution for data scientists to effectively discover anomalies over large volumes of varied data. To build such a system\, several critical challenges have to be solved: How to determine which among many alternative anomaly detection algorithms is the best for a given task and to find the proper parameter settings? How to leverage a small amount of end-user feedback to improve the anomaly extraction process? How to best present the anomaly detection results such that users do not have to evaluate the potentially large number of anomaly candidates one by one? \nThis talk will present our ADP solution that solves all above problems. ADP supports all stages of anomaly discovery by seamlessly integrating anomaly-related services within one integrated platform. It enables the tuning free anomaly detection and achieves the power of sense making afforded by anomaly summarization and explanation services\, while allowing the users to easily steer the discovery process with human ingenuity. \nBio: \nDr. Lei Cao is a Postdoc Associate at MIT CSAIL\, working with Prof. Samuel Madden and Prof. Michael Stonebraker. Before that he worked at IBM T.J. Watson Research Center as a Research Staff Member. He received his Ph.D. in Computer Science from Worcester Polytechnic Institute\, supervised by Prof. Elke Rundensteiner. His recent research is focused on developing end-to-end tools for data scientists to effectively make sense of data. \n\nZoom Linj: https://uci.zoom.us/j/808173003
URL:https://isg.ics.uci.edu/event/lei-cao-toward-an-end-to-end-anomaly-discovery-paradigm/
LOCATION:CA
END:VEVENT
END:VCALENDAR