Big Data Workshop
12 January, 2017
Room 00.225 of 200A Celestijnenlaan, Heverlee
Program
| Time | Speaker | Title of talk | ||
|---|---|---|---|---|
| 13:45 - 14:00 | Coffee | |||
| 14:00 - 14:30 | Welcome | Bart De Moor | [Download] | |
| 14:30 - 15:00 | Contribution Computer Sciences | Jesse Davis | Big Data: A Computer Science Perspective | [Download] |
| 15:00 - 15:30 | Contribution Humanities | Bart Baesens | Putting Big Data & Analytics to Work! | [Download] |
| 15:30 - 16:00 | Break | |||
| 16:00 - 16:30 | Contribution ESAT | Jan Aerts | From Data to Insights | |
| 16:30 - 17:00 | Contribution Biomedical Sciences | Stein Aerts | Decoding genomic regulatory landscapes in cancer | |
| 17:00 - 17:30 | Contribution Science & Technology | Jan De Spiegeleer | Data-mining : new paradigm on the trading-floor | [Download] |
| 17:30 | Closing words | Geert Molenberghs | ||
| 18:00 | Reception |
Big Data: A Computer Science Perspective
Jesse Davis
Abstract
Big data poses many challenges in terms of collection, storage, retrieval, processing, analysis, and exploitation of the results. While a wide variety of disciplines are playing crucial roles in addressing each of these areas, this talk will address them from a computer science perspective. The first part of the talk will provide a high-level overview that outlines what some of the key computer science challenges are in these areas. I will then provide intuitions as to central ideas people are pursuing to address them. The second part of talk will go more in depth into the analysis aspect and present several illustrative examples.
Putting Big Data & Analytics to Work!
Bart Baesens
Abstract
Companies are being flooded with tsunamis of data collected in a multichannel business environment, leaving an untapped potential for analytics to better understand, manage and strategically exploit the complex dynamics of customer behavior. In this presentation, we will start by providing a bird’s eye overview of the analytics process model and then illustrate how to fully unleash its power in some example business settings. We will zoom into the key requirements of good analytical models (e.g. statistical validity, interpretability, operational efficiency, regulatory compliance etc.) and discuss emerging applications. The presentation will provide a mix of both theoretical and technical insights, as well as practical implementation details. The presenter will also extensively report on his recent research insights about the topic. Various real-life case studies and examples will be used for further clarification.
From Data to Insights
Jan Aerts
Abstract
The age of big data is changing the very core of how research is performed: we are moving from a hypothesis-first to a data-first paradigm. This talk will address this change from a signal processing point of view, highlighting some of the specific approaches investigated by the STADIUS group. These will include some of the particular techniques developed as well as their application in different domains. In addition, as a research community we will have to address the possible unforeseen consequences of using a big data approach, particularly as it concerns privacy and openness of the developed algorithms.
Decoding genomic regulatory landscapes in cancer
Stein Aerts
Abstract
I will discuss several applications of transcriptome and epigenome profiling in cancer and how these data can be combined with regulatory sequence analysis to decipher gene regulatory networks controlling cellular states in cancer.
The first application uses a cancer model in Drosophila where we evaluate different methods for open chromatin profiling and infer functional networks driven by AP-1 and STAT. In the second application we compare different phenotypic states in human melanoma, and show how decoding the regulatory landscape in each state provides novel insight into the gene networks that underlie clinically relevant events in melanoma, such as phenotype switching, invasion and resistance to therapy. In a final case study we explore massively parallel enhancer reporter assays and deep learning methods to understand what distinguishes functional enhancers from other bound genomic regions that have no regulatory activity, using human TP53 as a model transcription factor. I will conclude by giving a perspective on mapping regulatory landscapes at single-cell resolution using single-cell genomics.
Data-mining : new paradigm on the trading-floor
Jan De Spiegeleer
Abstract
Quantitive finance within financial institutions has undergone an important change. The heydays of financial innovation where exotic financial instruments were engineered and sold to investors seem to be over. In the aftermath of the financial crisis of 2008, risk managers and compliance officers got involved at all the hierarchical levels within banks and insurance companies. At the same time, IT departments got to deal with the implementation of the new regulations such as Basel III, CRDIV, Dodd-Frank, Mifid II, etc… Even before Basel III was fully rolled out, Basel IV came luring around the corner in the summer of 2016. Because financial institutions were very occupied, new entrepreneurial companies jumped on the band-wagon. This was the advent of FinTech and RegTech companies. Hedge-funds but also traditional asset managers, who unlike banks had a much lighter legacy, focused their quantitative research in the direction of algorithmic trading, robo advisors, machine learning and all kinds of implementations of data mining tools. Data scientists became the new quants on the trading floor. This lecture illustrates the new opportunities both from a research and business perspective.