CS Research Day 2022

Opening of the event

Associate Professor Janne Lindqvist
Head of Department of Computer Science

Modelling Mental Rotation for Deep Learning

Assistant Professor Stephane Deny
Department of Computer Science

Mental rotation is the process by which humans are able to recognize objects in unusual poses. Deep networks lack this ability, and as a consequence are less robust than the human visual system. In this talk I will describe ongoing work to understand how the human brain performs mental rotation in order to implement a similar process in deep learning.

Temporal Networks for Business and Pleasure

Professor Petter Holme
Department of Computer Science

Temporal networks are extensions of graphs to situations where we know not only who is in contact with whom but also when the contacts happen. Data from social media, human mobility, and phone calls can all be represented as temporal networks. I will present how this paradigm is transforming our understanding of disease and information spreading and its promises and challenges for the future.

Geometric Algorithms and Computational Limits

Assistant Professor Sandor Kisfaludi-Bak
Department of Computer Science

Computer graphics, computer aided manufacturing, geographic information systems, integrated circuit design, and many other fields use geometric algorithms to reach their goals. Typically, the problems solved by these algorithms are about manipulating or computing the relationship of certain objects in low-dimensional spaces. The age-old question in theoretical computer science is to find the most efficient algorithm for each task. Can we find the most efficient algorithms for geometric tasks? What type of evidence can we offer that the efficiency of a given geometric algorithm cannot be improved further?

Tight Proof of Sequential Work from Lattices

Assistant Professor Russell Lai
Department of Computer Science

Imagine that you have discovered a proof x of a famous conjecture, e.g. the P v.s. NP problem, at time t and are about to publish the proof. However, academic publication takes time and a competitor who later discovers an alternative proof might get all the credits if they manage to publish their proof first. To prevent this, you decide to timestamp your proof x by performing a moderately hard computation f(x) starting at time t. If the computation f(x) requires time T, by publishing the proof x and the computation result f(x) at time t+T, everyone will be convinced that you have discovered the proof T units time in the past. A proof of sequential work (PoSW) scheme can be thought of as a special moderately hard computation f where verifying that y = f(x) given (x,y) can be done in much less time than computing f(x). In this talk, I will present to you a new construction of post-quantum-secure PoSW whose sequentiality is based on conjectured hard problems over lattices.

Quantum Computers, Or: don’t let your qubits grow up to be cowboys

Assistant Professor Alexandru Paler
Department of Computer Science

Building the software for large-scale quantum computers is a highly exciting and interdisciplinary challenge involving research of computer engineering, machine learning and theoretical computer science. We list some of the challenges tackled by the Quantum Software and Algorithms group.

Closing words

Associate Professor Janne Lindqvist
Head of Department of Computer Science