New York Database Day is a workshop targeting the database researchers and research-oriented practitioners in the New York area. New York has an incredibly strong database presence, and it is prime time for folks to get together. The Database Group at Columbia University will host the inaugural NYDBDay on November 10th at Columbia University. The purpose is for researchers in the area to learn about each others’ research, have discussions on important database directions, and look for ways to grow and foster the research+industrial community.
Abstract: Crowdsourcing has raised several security concerns. One of the concerns is how to assign sensitive tasks in the crowdsourcing market, especially when there are colluding participants in crowdsourcing. In this presentation, I will talk about our work on sensitive task assignments in crowdsourcing markets with colluding workers.
Abstract: Digital traces of our lives are now constantly produced by various connected devices, internet services and interactions. These “personal digital traces” are different from traditional personal files; they are typically (but not always) smaller, heterogeneous, and accessible through a wide variety of portals and interfaces or directly stored on our devices. Users have very few tools to organize, understand, and search the digital traces they produce. The DigitalSelf project at Rutgers University proposes models and techniques to integrate, aggregate, organize, and search personal information within a collection of a user’s personal digital traces. First, I will present our current work –based on an intuitive data model which classifies personal digital traces along the basic six dimensions: what, when, where, who, why, and how– on searching personal data using dynamic frequency- and learning-based ranking functions. Then, I will discuss how we integrate and connect traces through the use of episodic scripts to create personal narratives. Finally, I will discuss possible extensions of this work in various real-world applications such as senior care, or privacy-preserving personalization.
Abstract: In this talk, we will present an overview of the research projects undergoing at the Big Data Analytics Lab (BDaL), at the New Jersey Institute of Technology. Broadly speaking, at BDaL, we focus on an alternative computational paradigm that involves ““humans-in-the-loop”” for large scale analytics problems . These problems arise at different stages in a traditional data science pipeline (e.g., data cleaning, query answering, ad-hoc data exploration, or predictive modeling), as well as from emerging applications. We study optimization opportunities that come across because of this unique man-machine collaboration, and address data management and computational challenges to enable large scale analytics with humans-in-the-loop. These research projects are funded by National Science Foundation, Office of Naval Research, National Institute of Health, and Microsoft Research. The talk will have two parts - in the first part, we will describe how we can substitute expensive domain expertise without compromising much on the quality or latency by leveraging multiple human workers (such as students, amateur scientists, or even lay crowdsourced workers) in problems that arise in a traditional data science pipeline. We will highlight our solutions to address computational challenges and optimization opportunities. In the second part of the talk - we will describe problems that come from emerging applications to enable collaboration among a group of individuals and machine algorithm. We will present an overview of our proposed approach and conclude by outlining some research challenges.”
Abstract: My research group does work in verified concurrent algorithms for data structure, a meta-algorithm for machine learning that allows the refusal of a predication in order to guarantee a correctness rate, energy-efficient blockchains, and a few other topics. My talk will give an overview of the research areas and focus on the energy-efficient blockchain design under fail-stop and traitorous failures.
Title: Algorithms and Systems for Responsible Data Science
Abstract: In this talk I will give my perspective on data science, a field that employs elegant algorithmic insights and generalizable systems advances to meet important societal challenges. I will give this perspective through the lens of the ongoing projects in my research group. These include the Portal project – a formal framework for analysis of evolving graphs, instantiated in an open-source system based on Apache Spark, and DB4Pref – algorithms and systems for analysis of preferences and votes that branches into computational social choice. The main focus of my talk will be on the Data, Responsibly project that frames managing data in accordance with ethical and moral norms, and legal and policy considerations, as a core system requirement. I will highlight some of our recent algorithmic advances, and will discuss the over-all framework in which responsibility properties are managed and enforced through all stages of the data lifecycle.
Abstract: My research group focuses on web search engines, and in particular on efficiency issues and on search engine architecture. Current interests include basic indexing and query processing techniques, efficiency in cascading search architectures used to implement complex ranking functions, and load balancing and query routing in distributed (sharded) search architectures. In this talk, I will give a brief overview of current research directions in my group.
Abstract: Specialized hardware offers the potential for higher performance with a lower power and transistor budget than general purpose hardware. For common workloads such as data analytics, it may pay to invest in a Database Processing Unit (DPU) analogous to a GPU for graphics workloads. In this talk, I will present the Q100 architecture, designed to support SQL query processing. I will outline the Q100 design and evaluation, including the selection of individual tiles to perform component operations such as sorting, joins, and aggregations. The internal network connecting the components can be specialized to favor common data paths for better area and/or performance.
A walk through of what Bluecore is and what data products we use. I’ll briefly talk about our data problems we have run into, which mostly relate to “manageability” and connecting data between different systems.
Many corporations often have a large corpus of databases created for different purposes. A data analyst, given a task at hand, needs to find the right dataset(s) from this corpus to use for her task. This often turns into a challenging exercise due to several reasons. One such reason is that technical metadata, such as table and column names, are often abbreviated, e.g. “cust_acct” instead of “customer_account” for a column name. At the same time, column “cust_acct” might not have any business metadata, i.e., human-generated column descriptions, associated with it. Hence, the analyst searching for “customer_account” might not find “cust_acct” column due to the keyword mismatch.
In our Automated Keyword Generation (AKG) project at AT&T we have developed a machine learning approach that learns “tag expansions” from a small corpus of business metadata. The approach is based on supervised learning from noisy data, as column descriptions can have a tangible amount of noise. When a tag has multiple meanings, the approach is able to leverage context to better choose the most likely meaning. The approach has been deployed at AT&T as part of a larger system that allows the user to navigate and search the metadata for a very large corpus of databases.
Julia will moderate a broad discussion of human in the loop data analytics.
Below is a list of registered posters. If you did not register a poster, you are still welcome to bring one to present. We will provide easels, pins, and posterboard.
Algorithmic decisions often result in scoring and ranking individuals to determine credit worthiness, qualifications for college admissions and employment, and compatibility as dating partners. While automatic and seemingly objective, ranking algorithms can discriminate against individuals and protected groups, and exhibit low diversity. Furthermore, ranked results are often unstable — small changes in the input data or in the ranking methodology may lead to drastic changes in the output, making the result uninformative and easy to manipulate. Similar concerns apply in cases where items other than individuals are ranked, including colleges, academic departments, or products.
Data lineage describes the relationship between individual input and output data items of a workflow and is an integral ingredient for both traditional (e.g., debugging or auditing) and emergent (e.g., explanations or cleaning) applications. The core, long-standing problem that lineage systems need to address is to quickly capture lineage across a workflow in order to speed up future queries over lineage. Current lineage systems, however, either incur high lineage capture overheads, high lineage query processing costs, or both. In response, developers resort to manual implementations of applications that, in principal, can be expressed and optimized in lineage terms. In this poster, we describe Smoke, an in-memory database engine that provides both fast lineage capture and lineage query processing. To do so, Smoke tightly integrates the lineage capture logic into physical database operators; stores lineage in efficient lineage representations; and employs optimizations if future lineage queries are known up-front. Our experiments on microbenchmarks and realistic workloads show that Smoke reduces the lineage capture overhead and lineage query costs by multiple orders of magnitude as compared to state-of-the-art alternatives. On real-world applications, we show that Smoke meets the latency requirements of interactive visualizations (e.g., <150ms) and outperforms hand-written implementations of data profiling primitives.
Amir Pouya Aghasadeghi
In recent years there has been significant interest in evolutionary analysis of large-scale networks. Researchers study network evolution rate and mechanisms, the impact of specific events on evolution, and spatial and spatio-temporal patterns. To support data scientists who are studying network evolution, there is a need to develop scalable and generalizable systems. Tangible systems progress in turn depends on the availability of standardized datasets on which performance can be tested. In this work, we make progress towards a data generator for evolving property graphs, which represent evolution of graph topology, and of vertex and edge attributes. We propose an attribute-based model of preferential attachment, and instantiate this model on a co-authorship network derived from DBLP, with attributes representing publication venues of the authors. We show that this attribute-based model predicts which edges are created more accurately than a structure-only model. Finally, we demonstrate that synthetic graphs are indeed useful for evaluating performance of evolving graph query primitives.
Personal information is typically fragmented across multiple, heterogeneous, distributed sources and saved as small, heterogeneous data objects, or traces. The DigitalSelf project at Rutgers University focuses on developing tools and techniques to manage (organize, search, summarize, make inferences on and personalize) such heterogeneous collections of personal digital traces. We propose to demonstrate YourDigitalSelf, a mobile phone-based personal information organization application developed as part of the DigitalSelf project. The demonstration will use a sample user data set to show how several disparate data traces can be integrated and combined to create personal narratives, or coherent episodes, of the user’s activities. Attendees will be given the option to install YourDigitalSelf on their own devices to interact with their own data.
There is excellent software and hardware infrastructure for every part of neural network development except for understanding what the model has learned. Deep Neural Inspection (DNI) is an emerging technique that visually or statistically measures the relationship between hidden unit behaviors (e.g., activations, activation gradients) and data point annotations (e.g., annotated pixels, part of speech tags). However, each analysis requires hundreds or thousands of lines of analysis code to extract hidden unit behaviors, formulate hypotheses to test, and ensure that the analysis completes quickly. DeepBase is a general purpose DNI system that provides a high-level programming and query interface to express DNI analyses. This paper describes the current system and illustrates how it can simplify the workflow for inspecting models in two domains: an agent’s game-playing policy model and a facial recognition model.
To facilitate collaboration over sensitive data, we present DataSynthesizer, a tool that takes a sensitive dataset as input and generates a structurally and statistically similar synthetic dataset with strong privacy guarantees. The data owners need not release their data, while potential collaborators can begin developing models and methods with some con dence that their results will work similarly on the real dataset. The distinguishing feature of DataSynthesizer is its usability — the data owner does not have to specify any parameters to start generating and sharing data safely and e ectively. DataSynthesizer consists of three high-level modules — DataDescriber, DataGenerator and ModelInspector. The rst, DataDescriber, investigates the data types, correlations and distributions of the attributes in the private dataset, and produces a data summary, adding noise to the distributions to preserve privacy. DataGenerator samples from the summary computed by DataDescriber and outputs synthetic data. ModelInspector shows an intuitive description of the data summary that was computed by DataDescriber, allowing the data owner to evaluate the accuracy of the summarization process and adjust any parameters, if desired.
We consider the SQL Selection-GroupBy-Aggregation (SGA) query evaluation on an untrusted MapReduce system in which mappers and reducers may return incorrect results. We design AssureMR, a system that supports efficient verification of result correctness for both intermediate and final results of SGA queries. AssureMR includes the design of Pedersen Merkle R-tree (PMR-tree), a new authenticated data structure (ADS). To enable efficient verification, AssureMR includes a distributed ADS construction mechanism that allows mappers/reducers to construct PMR-trees in parallel without a centralized party.
AssureMR provides the following verification functionality: (1) correctness verification of PMR-trees by replication; (2) correctness verification of intermediate (final, resp.) query results by constructing local (global, resp.) PMR-trees and verification objects. Our experimental results demonstrate the efficiency and effectiveness of AssureMR.”
We designed a crowdsourcing platform that uses human-in-the-loop optimization to understand the behavior of the workers and model the preference of a worker such that it can be used for a variety of different applications like task recommendation for workers or workforce curation for requesters (optimizing latency, quality and etc.).
Registration and Submission deadline: 10/30/2018
Submission Notifications: 11/3/2018
Submission and registration
Please register so we can order an appropriate amount of food. Optionally submit talk and poster abstracts.