Telecom ParisTech Joint Workshop

NUS-SoC / Telecom ParisTech
Joint Workshop on Data Sciences and Artificial Intelligence

Workshop Date: 12 April 2018
Venue: COM1, Level 2, Cerebro@SoC
Registration Period: 2 to 8 April 2018

Registration Link: Closed

Due to limited places, attendance is by registration and first-come-first serve basis only.

Programme Outline

Session 1Artificial Intelligence
9.30amTitlePresentation by AI Singapore
SpeakerKoo Sengmeng, Deputy Director, Strategic Alliances, AI Singapore
10.30amTitleDiversity Constraints in Public Housing Allocation
SpeakerYair Zick, Assistant Professor, NUS-SoC

The state of Singapore operates a national public housing program, accounting for over 80% of its residential real estate. Singapore uses its housing allocation program to ensure ethnic diversity in its neighborhoods; it does so by imposing ethnic quotas: every ethnic group must not own more than a certain percentage in a housing project, thus ensuring that every neighborhood contains members from each ethnic group. However, imposing diversity constraints naturally results in some welfare loss. Our work studies the tradeoff between diversity and social welfare from the perspective of computational economics. We model the problem as an extension of the classic assignment problem, with additional diversity constraints. While the classic assignment program is poly-time computable, we show that adding diversity constraints makes the problem computationally intractable; however, we identify a 1/2-approximation algorithm, as well as reasonable agent utility models which admit poly-time algorithms. In addition, we study the price of diversity: this is the loss in welfare incurred by imposing diversity constraints; we provide upper bounds on the price of diversity as a function of natural problem parameters; next, we analyze public data from Singapore’s Housing and Development Board, and create a simulated framework testing the welfare loss due to diversity constraints in realistic large-scale scenarios.

Joint work with Nawal Benabbou, Mithun Chakraborty, Vinh Ho Xuan, and Jakub Sliwinski. Accepted to AAMAS 2018.

11.15amTitleMassive Online Analytics for the Internet of Things
SpeakerAlbert Bifet, Professor, Telecom ParisTech
AbstractBig Data and the Internet of Things (IoT) have the potential to fundamentally shift the way we interact with our surroundings. The challenge of deriving insights from the Internet of Things (IoT) has been recognized as one of the most exciting and key opportunities for both academia and industry. Advanced analysis of big data streams from sensors and devices is bound to become a key area of data mining research as the number of applications requiring such processing increases. Dealing with the evolution over time of such data streams, i.e., with concepts that drift or change completely, is one of the core issues in stream mining. In this talk, I will present an overview of data stream mining, and I will introduce some popular open source tools for data stream mining.
12.00pmLunch Break
Session 2Data Analytics
1.30pmTitlePresentation by Singapore Data Science Consortium
SpeakerToh Swee Feng Caroline, Business Development Manager, Singapore Data Science Consortium
AbstractShort introduction of Singapore Data Science and activities.
2.00pmTitlePrivacy in Machine Learning as a Service
SpeakerReza Shokri, Assistant Professor, NUS-SoC

In this talk, I will present different threats against data privacy in machine learning systems, with the focus on machine learning as a service platforms. Google, Amazon, Microsoft, BigML, and other service providers enable data holders to benefit from machine learning as a service by simply uploading their data to the service provider and obtaining API access to machine learning models trained on their data.

Obviously, the service providers get direct access to the data, which may be of a serious concern for sensitive data holders. I show how machine learning as a service platforms could be constructed that enable service providers to train models without seeing the data. Even with a blind training the model is not privacy preserving. The threat also extends to the users of the machine learning as a service. I will show that an attacker (who accesses the model through the machine learning API) can build inference algorithms to determine the members of the model’s training ata. Finally, I will discusss how such subtle threats could be mitigated.

3.00pmTitleReinventing the mind’s bicycle
SpeakerJames Eagan, Associate Professor, Telecom ParisTech
AbstractSteve Jobs once described the computer as “a bicycle for the mind”, a tool that amplifies people’s cognitive capabilities. My work focuses on making this bicycle a better, more expressive tool for humans in a data-rich world: by providing richer tools to interact with and understand data, by making interactions more expressive, and by reinventing how we build software so users can appropriate and adapt it to their own idiosyncratic needs. In this talk, I will present an overview of our work in these three areas.
Session 3Graphic and Vision
4.30pmTitleComputer Vision for Intelligent Systems
SpeakerLee Gim Hee, Assistant Professor, NUS-SoC
AbstractI will present three of our recent works that have been accepted for publication in CVPR 2018. In the first work, I will describe the CVM-Net, i.e. a deep learning architecture that we propose for street-view to satellite-view image-based localization. In the second work, I will talk about a novel CNN-based deep network for human motion trajectory generation. Finally, I will discusss our PointNetVLAD, which is a deep network for the order-less point cloud based place recognition task.
5.15pmTitleArtificial Intelligence in the IMAGES team with a special focus in Computational Anatomy
SpeakerPietro Gori, Assistant Professor, Telecom ParisTech
AbstractIn the first part of this talk, I will describe the IMAGES team of the LTCI lab and its main research topics in Artificial Intelligence: machine learning, pattern recognition, knowledge and uncertainty representation, interaction and spatial reasoning. I will present some applications in computer graphics, medical imaging and texture synthesis. The second part of the talk will instead focus on computational anatomy applied to neuro-imaging. I will present a method to statistically describe the anatomical shape and organisation of the components of the brain modelled as geometrical objects (i.e. 3D polylines and surfaces). Being composed of hundreds of millions of points, approximation schemes and multi-resolution strategies will be presented in order to make the proposed method computationally feasible.
Session 4Round Table

Live Better in Smart Cities Empowered with AI

Facilitated by: Dr Ayesha Khanna, CEO at ADDO AI


Albert Bifet, Head of Data, Intelligence and Graphs Group at Telecom ParisTech

Damien Dhellemmes, Schneider Country President, Singapore

David Hsu, Professor of Computer Science, Vice Dean (Research), NUS-SoC

Bernard Leong, Head of Airbus Aerial Asia

Cocktail refreshment