[cs-talks] Upcoming CS Seminars: Gurari PhD Proposal (Thurs) + Erdos PhD Proposal (Fri) + Bassem PhD Proposal (Mon)
Conroy, Nora Mairead
conroynm at bu.edu
Thu Jan 22 08:52:07 EST 2015
Combining Efforts of Crowdsourced Humans and Computers to Collect Accurate Boundaries of Objects in Images
Danna Gurari, BU
Thursday, January 22, 2015 at 9am in MCS 148
Abstract: Advances in image acquisition and storage technologies have led to many image-based experiments that are designed to systematically study basic science processes. This thesis centers on the design of image annotation methods to accelerate the transition from data collection to scientific discovery. The focus is on biomedical images in an effort to contribute to research that addresses society’s health care problems. The first half of the work includes a detailed analysis of the relative strengths and weaknesses of three different approaches to demarcate object boundaries in images: by experts, by crowdsourced laymen, and by automated computer vision algorithms. Results revealed that popular computer vision algorithms and state-of-art human computation crowdsourcing systems come with unique strengths but both methods were insufficient alone to yield the desired accuracy and scale to support quantitative analyses. These results inspired a research trajectory for gaining traction on this problem by examining how to effectively integrate computer vision algorithms and crowdsourced laymen in order to achieve expert-level annotation at scale. Results from two implemented hybrid crowdsourcing-computer methods are compared to annotations of experts, two pure crowdsourcing methods, and two pure algorithmic methods on 405 objects in everyday and biomedical images. Experiments revealed that a hybrid system design yielded the most accurate results among the six system configurations and it yielded an accuracy that is statistically similar to segmentations created by biomedical experts. The final chapter of this proposal includes a case study that will underscore the potential for this hybrid algorithm-crowdsourcing approach to accelerate big data computational biology analyses. A system will be built that incorporates this approach and show how it creates an expert-quality 3D volume reconstruction of fish structures. This work will support analyses that link biological structure and function. To encourage community-wide effort to continue working on developing methods and systems for image-based studies which can have real and measurable impact that benefit society at large, datasets and code are publicly-shared (http://www.cs.bu.edu/~betke/BiomedicalImageSegmentation/).
Group Centrality for Repetition-Aware Content Placement
Dora Erdos, BU
Friday, January 23, 2015 at 10am in MCS 148
Abstract: In this talk I am going to focus on the problem of identifying central nodes in repetition aware environments. Arguably, the most effective technique to ensure wide adoption of a concept (or product) is by repeatedly exposing individuals to messages that reinforce the concept (or promote the product). We propose a novel framework for the effective placement of content: Given the navigational patterns of users in a network, e.g., web graph, hyperlinked corpus, or road network, and given a model of the relationship between content-adoption and frequency of exposition, we define the repetition-aware content-placement problem as that of identifying the set of B nodes on which content should be placed so that the expected number of users adopting that content is maximized. The key contribution of our work is the introduction of memory into the navigation process, by making user conversion dependent on the number of her exposures to that content. This dependency is captured using a conversion model that is general enough to capture arbitrary dependencies. Our solution to this general problem builds upon the notion of absorbing random walks, which we extend appropriately in order to address the technicalities of our definitions. This paper is characteristic of my work on group centrality and will be part of my thesis. In the talk I will also give a brief overview of my thesis which consists of two main topics; centrality measures and the analysis of dot-product graphs.
PhD Proposal Defense
Route Coordination in Brokered Environments
Christine Bassem, BU
Monday, January 26, 2015 at 11am in MCS 148
Abstract: With the recent shift towards the integration of federated commodities to provide high quality services, the adoption of brokered environments has increased. In brokered environments, participants share information about their resources and/or workloads with each other and a broker, which acts as an intermediary between them; with varying degrees of involvement. As many services can be provided in such environments, I focus on the services that require the routing of a set of commodities over a communication/mobility graph, hence the route coordination of autonomous commodities is required.
In this talk, I will briefly present the route coordination models that I study in my thesis, and discuss in details my most recent work on network-constrained packing of brokered workloads in virtualized environments. This work is a form of implicit route coordination on stationary graphs, in which routes’ properties are implicitly defined by their endpoints, and route coordination includes the decision on the endpoints of communication. As we focus on the problem of providing predictability guarantees to data-intensive workloads, we observe the range of data-intensive applications and data center topologies to define an abstraction that allows us to relate the problem to that of bin packing with implicit network properties. Then, we exploit that abstraction to classify the problem model into several special instances, and develop polynomial-time exact algorithms for two of these special instances that occur frequently in reality. Finally, we develop a greedy heuristic to solve the problem in the general model, which we evaluate via extensive simulations.
Azer Bestavros (reader)
Ibrahim Matta (reader)
Evimaria Terzi (reader)
Mark Crovella (chair)
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cs-talks