Events Search and Views Navigation
October 2016
Seminar: Automatically Comparing Memory Consistency Models
S-REPLS Seminar Slides Speaker name: Dr. John Wickerson Abstract A memory consistency model (MCM) is the part of a programming language or computer architecture specification that defines which values can legally be read when a thread in a concurrent program reads from a shared memory location. Because MCMs have to take into account various optimisations employed by modern architectures (such as store buffering and instruction reordering) and compilers (such as constant propagation), they often end up being complex and counterintuitive, which…
Find out more »December 2016
Seminar: Deep Learning Financial Market Data
Speaker: Steven Hutt Seminar title: Deep Learning Financial Market Data Abstract: An introduction to learning patterns in financial market data Slides
Find out more »February 2017
Seminar – Predicting User Demographics in Social Networks
Speaker name: Nikolaos Aletras Abstract: Automatically inferring user demographics in social networks is useful for both social science research and a range of downstream applications in marketing and politics. Our main hypothesis is that language use in social networks is indicative of user attributes. This talk presents recent work on inferring a new set of socioeconomic attributes, i.e. occupational class, income and socioeconomic class. We define a predictive task for each attribute where user-generated content is utilised to train supervised…
Find out more »Predicting User Demographics in Social Networks
Speaker name: Nikolaos Aletras Automatically inferring user demographics in social networks is useful for both social science research and a range of downstream applications in marketing and politics. Our main hypothesis is that language use in social networks is indicative of user attributes. This talk presents recent work on inferring a new set of socioeconomic attributes, i.e. occupational class, income and socioeconomic class. We define a predictive task for each attribute where user-generated content is utilised to train supervised non-linear…
Find out more »March 2017
Seminar: Challenges in Operational Technology for the Process Industry
Abstract The following main challenges at the interface of automation technology, process technology and information technology will be discussed: Models within the life-cycle Vertical integration from ERP (Enterprise Resource Planning) to actuator and vice versa Horizontal integration requirements for highly reliable solutions Human Machine Integration For all of these challenges the state-of-the-art and current BASF approaches will be presented. The goal of the seminar is to create awareness for practical problems in the process industry and stimulate further discussions, as…
Find out more »Lecture/Workshop: Model Predictive Control from an Application Point of View in Process Industry
The aim of this lecture is to provide an industrial perspective on Model Predictive Control with emphasis on: Important practical requirements Practice-proven solutions and approaches Understanding the required efforts in terms of time and money Differentiation between state of the art, state of science and vision (= inspirations for further developments) Lecture outline Motivation Model Predictive Control for Continuous Plants Nonlinear Model Predictive Control for Batch Plants Real Time Optimization Summary and Outlook Short Bio Dr Birk obtained his PhD…
Find out more »HiPEDS Seminar – Big graphs on big machines
Speaker: Dr. Tim Harris, Oracle Labs Cambridge Seminar Title: HiPEDS Seminar - Big graphs on big machines Abstract: Oracle's largest SPARC M7 system provides 4096 hardware threads spread over 16 sockets in one cache-coherent address space. I will talk about our experience tuning graph analytics workloads to run well on this system, and how we went from an implementation that stopped scaling at around 200 threads to a version that provides super-linear speed-ups on PageRank and SSSP running on 1TB+ inputs over the full machine. I will focus on…
Find out more »May 2017
Data Co-Management with Modern Hardware
Upcoming lecture by Raja Appuswamy, Postdoctoral Researcger in the DIAS lab at EPFL, on data management systems Abstract The design of data management systems has always been driven based on two aspects, namely, underlying hardware and applications requirements. The past few years have, however, witnessed dramatic changes in both these aspects. On the hardware front, computer architecture is in a phase of constant change resulting in growing heterogeneity in the design of processors, memory, interconnects, and storage. With the wide-spread…
Find out more »August 2017
HiPEDS Seminar: Web Data Extraction: A Crash Course
HiPEDS CDT Seminar Series with Giorgio Orsi, Senior Research Scientist at Meltwater. Abstract: Data acquisition plays an important role in modern organisations and is a strategic business process for data-driven companies such as insurers, retailers, and search engines. Data acquisition processes range from manual data collection and purchase, to cheaper but often technically challenging methods such as automated collection and crowdsourcing. The abundance of web data has made web scraping (also known as web data extraction or web wrapping) an…
Find out more »October 2017
HiPEDS Seminar: Database storage tiering, fast and slow
Abstract: In 1987, Jim Gray and Gianfranco Putzolu introduced the five-minute rule for trading memory to reduce disk I/O using the then-current price-performance characteristics of DRAM and Hard Disk Drives (HDD). Since then, the five-minute rule has gained wide-spread acceptance as an important rule-of-thumb in data engineering. In the first part of this talk, we will revisit the five-minute rule three decades since its introduction and use it to identify impending changes in today's multi-tier storage hierarchy given recent trends…
Find out more »