Analytics Network Event

Please register to this event also using the Eventbrite link: http://bit.ly/AnalyticsNetworkLBS2019

The Analytics Network, part of the Operational Research Society, invites you to join to our next event that will take place at London Business School .

In this occasion we will have three different speakers from a wide range of topics and each session will be followed by Q&A. At the end of the event there will be plenty of time for networking.

Timetable:

6:00pm-6:30pm Registration

6:30pm-8:00pm Talks

1st Speaker: John A Clark, Professor of Computer and Information Security at the University of Sheffield

An Optimization Based Approach to Security

A great many problems can be couched as optimization problems. Various aspects of security lend themselves readily to such formulation and others can be advantageously couched in such terms. In this talk I shall show how optimization approaches can be used to derive security related artifacts from cryptographic building blocks,through security protocols, to system wide security probe placements (how best to use limited resources) and the synthesis of intrusion detection rules. Optimization techniques have shown themselves to be capable of challenging (breaking) theoretical conjectures by leading cryptographers. I will outline further possibilities for discovery of high level artifacts such as jamming strategies and new attacks on systems.

2nd Speaker: Colin Gillespie, O'Reilly author, Data camp presenter, R trainer/consultant (Jumping Rivers) and academic

Data Science and Security

Data science is increasing performed in the cloud or over a network. But how secure is this process? In this talk, we won't look at complex hacking, but instead focus on the relatively easy hacks that can be performed to access systems.

3rd Speaker: Andrie de Vries, Solutions Engineer at RStudio

Taking TensorFlow into Production

TensorFlow is the most popular deep learning framework, and R users can use the full power of TensorFlow using the `keras` package. Many resources exist that describe how to train a model and achieve good results. However, production deployment introduces many complications. For example, the architect must design an appropriate architecture (CPU, GPU or TPU) for training as well as scoring. Production deployment assumes that you can readily create a pipeline that includes data pre-processing, model evaluation and re-training, as well as scoring on a production server with a suitable API. I illustrate the concepts by an RStudio internal deep network that classifies support tickets using natural language processing with recurrent layers. The network is deployed on an instance of RStudio Connect, and informs recommendations on Zendesk ticketing system.

4th Speaker: Lukasz Piwek, Lecturer in Data Science at the University of Bath

Analytics with Smartphones, Wearables and Social Networks: Opportunities and Barriers.

Application of machine learning methods with simple and openly available data generated with smartphone apps, wearables and social networks (so called “digital traces”) could be used to rapidly profile segments of populations in order to support healthcare, transportation, understand social behaviour, and socio-economic indicators. However, there are ongoing concerns about data reliability, privacy, security, and users’ attrition rates in using apps & wearables. We discuss those opportunities and barriers related to the use of "digital traces" in the context of recent studies and other evidence.

8:00pm-8:30pm Networking

For more details of the speakers, their talks and to to register for free as a member of the Analytics Network to receive news of events like this one please visit the following website:

http://www.theorsociety.com/Pages/SpecialInterest/AnalyticsNetwork.aspx

We look forward to seeing you.

Analytics Network, The OR Society