Faculty

Kristian Kersting
Computer Science Department
Centre for Cognitive Science
Technische Universität Darmstadt

Our minds make inferences that appear to go far beyond standard machine learning. Whereas people can learn richer representations and use them for a wider range of learning tasks, machine learning algorithms have been mainly employed in a stand-alone context, constructing a single function from a table of training examples. 

Ivan Stelmakh, PhD candidate
Machine Learning Department, School of Computer Science
Carnegie Mellon University

Peer review is the backbone of scholarly research and fairness of this process is crucial for the successful development of academia. In this talk, we will discuss our two recent works on fairness of peer review. In the first part of the talk, we will focus on the automated assignment of papers to reviewers in the conference setup. We will show that the assignment procedure currently employed by NeurIPS and ICML does not guarantee fairness and may discriminate against some submissions. In contrast, we will present the assignment algorithm that simultaneously ensures fairness and accuracy of the resulting allocation. 

Xin Lian, Master’s candidate
David R. Cheriton School of Computer Science

The problem of language alignment has long been an exciting topic for Natural Language Processing researchers. Current methods for learning cross-domain correspondences at the word level rely on distributed representations of words. Therefore, the recent development in the word computational linguistics and neural language modeling has led to the development of the so-called zero-shot learning paradigm.

Edith Elkind, Department of Computer Science
University of Oxford

Suppose that a group of agents want to select k > 1 alternatives from a given set, and each agent indicates which of the alternatives are acceptable to her: the alternatives could be conference submissions, applicants for a scholarship or locations for a fast food chain. In this setting it is natural to require that the set of winners represents the voters fairly, in the sense that large groups of voters with similar preferences have at least some of their approved alternatives in the winning set. 

Nabiha Asghar, PhD candidate
David R. Cheriton School of Computer Science

In human-computer interaction (HCI), one of the technological goals is to build human-like artificial agents that can think, decide and behave like humans during the interaction. A prime example is a dialogue system, where the agent should converse fluently and coherently with a user and connect with them emotionally. Humanness and emotion-awareness of interactive artificial agents have been shown to improve user experience and help attain application-specific goals more quickly. However, achieving human-likeness in HCI systems is contingent on addressing several philosophical and scientific challenges. In this thesis, I address two such challenges: replicating the human ability to 1) correctly perceive and adopt emotions, and 2) communicate effectively through language.

Chris Eliasmith, Director of the Centre for Theoretical Neuroscience
Department of Systems Design Engineering

Spiking neural networks (SNNs) have received little attention from the AI community, although they compute in a fundamentally different — and more biologically inspired — manner than standard artificial neural networks (ANNs). This can be partially explained by the lack of hardware that natively supports SNNs. However, several groups have recently released neuromorphic hardware that supports SNNs. 

Jaejun Lee, Master’s candidate
David R. Cheriton School of Computer Science

Used for simple voice commands and wake-word detection, keyword spotting (KWS) is the task of detecting pre-determined keywords in a stream of utterances. A common implementation of KWS involves transmitting audio samples over the network and detecting target keywords in the cloud with neural networks because on-device application development presents compatibility issues with various edge devices and provides limited supports for deep learning. Unfortunately, such an architecture can lead to unpleasant user experience because network latency is not deterministic. Furthermore, the client-server architecture raises privacy concerns because users lose control over the audio data once it leaves the edge device. 

Priyank Jaini, PhD candidate
David R. Cheriton School of Computer Science

Multivariate density estimation is a central problem in unsupervised machine learning that has been studied immensely in both statistics and machine learning. Several methods have thus been proposed for density estimation including classical techniques like histograms, kernel density estimation methods, mixture models, and more recently neural density estimation that leverages the recent advances in deep learning and neural networks to tractably represent a density function. In today's age when large amounts of data are being generated in almost every field it is of paramount importance to develop density estimation methods that are cheap both computationally and in memory cost. The main contribution of this thesis is in providing a principled study of parametric density estimation methods using mixture models and triangular maps for neural density estimation.