Select a calendar:
Filter July Events by Event Type:
SUNMONTUEWEDTHUFRISAT
University Calendar
Events for July
-
WEBINAR SERIES: Digital Technologies for COVID-19
Fri, Jul 10, 2020 @ 11:00 AM - 12:00 PM
Information Sciences Institute, USC Viterbi School of Engineering
University Calendar
This week's talk will be by Marjorie Freedman, a Research Team Leader from USC ISI.
Task Specific Data Annotation for COVID-19
Abstract: Information Extraction seeks to transform natural language text into structured records that capture key entities, relations, and events. Typical approaches to information extraction require annotators to construct domain-specific mark-up for each relation, event, and entity type of interest. This limits the applicability of information extraction in new domains, such as organizing scientific literature to suit the needs of COVID researchers. In this work, we explore alternative approaches to creating domain-specific annotation for new entities, relations, and events of interest. We provide annotators with tools to search for and label events of interest to COVID researchers and provide the flexibility in annotation to capture language that is suggestive, but non-definitive for the concepts of interest. We have begun annotation of several relations in the CORD19 (https://allenai.org/data/cord-19) dataset using this approach, and plan to make our results available.
Bio: Marjorie Freedman, a Research Team Lead at ISI, has degrees in linguistics and computer science from Cornell University. At ISI, she serves as PI of DARPA's AIDA, KAIROS, and ASED efforts. Under DARPA's AIDA project, her work has included tailoring speech recognition and optical character recognition systems for use in an information extraction pipeline. Also, under AIDA, she is exploring the impact of uncertainty in anaphora resolution to downstream tasks and working with vision researchers to understand and address the challenges of mapping the output of vision analytics to classic information extraction ontologies. Before joining ISI, she served as PI of IAPRA SCIL and Metaphor efforts; and as co-PI of BBN's DARPA DEFT and LORELEI efforts. As part of DEFT, she provided guidance in API development and served as the task coordinator for NIST's TAC 2014-16 Event Argument evaluations. As a part of this evaluation, she sought to identify a salient unit that could be evaluated and would be useful to downstream knowledge focused tasks. As PI of IARPA SCIL, she developed algorithms to understand the implicit social content of language, for example, identifying persuasive language in online discussion threads. Her work in information extraction has explored how to address limited training data, including fusing rule-based and learned systems, exploring alternative approaches to annotation, and measuring the impact of coreference in bootstrap learning for information extraction.
Co-hosted by:
Craig Knoblock, Executive Director, USC Information Sciences Institute
Bhaskar Krishnamachari, Director, USC Viterbi Center for CPS and IoTWebCast Link: https://usc.zoom.us/webinar/register/WN_bS2IGZDMTw2aymiLaBAzIw
Audiences: Everyone Is Invited
Contact: Bhaskar Krishnamachari
-
PhD Defense - Alana Shine
Tue, Jul 14, 2020 @ 03:30 PM - 05:30 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Defense - Alana Shine
Tues, Jul 14, 2020
3:30 PM - 5:30 PM
Ph.D. Defense - Alana Shine 7/14 3:30 pm Generative graph models subject to global similarity
Ph.D. Candidate: Alana Shine
Date: Tuesday, July 14, 2020
Time: 3:30 pm - 5:30 pm
Committee: David Kempe (chair), Aram Galstyan, Xiang Ren, Kayla de la Haye
Title: Generative graph models subject to global similarity
Zoom: https://usc.zoom.us/j/8333742899
Abstract:
This thesis explores how to build generative graph models subject to global features in order to capture connectivity structure. Generative graph models sample from sets of "similar" graphs according to a probability distribution and are important for simulation studies, anomaly detection, and characterizing properties of real world graphs in areas such as social science and network design. The vague notion of generating "similar" graphs has prompted a vast quantity of generative graph models that define similarity according to various graph features. Graph features used include degree distribution, motif counts, and high-level community structure. Typically, these features are local: the property can be segmented into parts with each part being computed entirely from its own subgraph. For example, node degrees. Instead, this work analyzes graph generation subject to global features that require the entire graph to compute. This thesis focuses on global features that capture connectivity because they are critical in determining how information/diseases spread on graphs and simulations of information/disease spread is a prominent application of generative graph models.
A large class of generative graph models are built from a single real world "target" graph and its features. This thesis presents three new generative graph models that target global similarity through matching (1) connectivity across cuts, (2) random walk behavior, and (3) eigenvalues of the symmetric normalized Laplacian matrix. All three of these global graph features are related to a widely used notion of graph connectivity called conductance. We observe on a number of real world target graphs that the global generative graph models perform superior to benchmark generative graph models on a number of similarity objectives.
WebCast Link: https://usc.zoom.us/j/8333742899
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
-
PhD Defense - Haoyu Huang
Thu, Jul 16, 2020 @ 02:00 PM - 04:00 PM
Thomas Lord Department of Computer Science
University Calendar
Ph.D. Defense - Haoyu Huang 7/16 2:00 pm "Nova-LSM: A Distributed, Component-based LSM-tree Data Store"
Ph.D. Candidate: Haoyu Huang
Date: Thursday, July 16, 2020
Time: 2:00 pm - 4:00 pm
Committee: Shahram Ghandeharizadeh (chair), Murali Annavaram, Jyotirmoy V. Deshmukh
Title: Nova-LSM: A Distributed, Component-based LSM-tree Data Store
Zoom: https://usc.zoom.us/j/99943500149
Google Meet (only if there are issues with Zoom): meet.google.com/ruu-jjiu-fbk
Abstract:
The cloud challenges many fundamental assumptions of today's monolithic data stores. It offers a diverse choice of servers with alternative forms of processing capability, storage, memory sizes, and networking hardware. It also offers fast network between servers and racks such as RDMA. This motivates a component-based architecture that separates storage from processing for a data store. This architecture complements the classical shared-nothing architecture by allowing nodes to share each other's disk bandwidth. This is particularly useful with a skewed pattern of access to data by scattering a large file across many disks instead of storing it on one disk.
This emerging component-based software architecture constitutes the focus of this dissertation. We present design, implementation, and evaluation of Nova-LSM as an example of this architecture. Nova-LSM is a component-based design of LSM-tree using RDMA. Its components implement the following novel concepts. First, they use RDMA to enable nodes of a shared-nothing architecture to share their disk bandwidth and storage. Second, they construct ranges dynamically at runtime to parallelize compaction and boost performance. Third, they scatter blocks of a file across an arbitrary number of disks and use power-of-d to scale. Fourth, the logging component separates availability of log records from their durability. These design decisions provide for an elastic system with well-defined knobs that control its performance and scalability characteristics. We present an implementation of these designs using LevelDB as a starting point. Our evaluation shows Nova-LSM scales and outperforms its monolithic counterpart by several orders of magnitude. This is especially true with workloads that exhibit a skewed pattern of access to data.WebCast Link: https://usc.zoom.us/j/99943500149
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
