Program 2020
Sunday January 19, 2020
Location: University of Tromsø - The Arctic University of Norway (Aud Cerebrum, MH2 building, U.08.316)
12:45 Pre-registration
13:00 - 18:00 Mini Deep Learning School
The Mini Deep Learning School will consist of 2 x 1.5 hour talks. The 1st talk is a brief Introduction to Deep Learning, introducing concepts such as CNN, RNNs as well as unsupervised Deep Learning models. The 2nd talk will describe the development of deep learning methodology mainly within the context of industrial power line inspection using optical images from UAVs, involving the development of novel a multi-stage detection and classification pipeline. Then, a new line-segment detector network will be presented, called LS-Net, to detect the power lines themselves, explaining in detail properties of convolution that enables this type of network. The LS-Net is fully convolutional and is trained solely on synthetic images. Finally, some recent developments within prototypical networks for few-shot learning will be presented.
Monday January 20, 2020
Location: University of Tromsø - The Arctic University of Norway (Aud Cerebrum, MH2 building, U.08.316)
08:30 Registration opens
09:00 Opening
09:15 Invited Talk by Bernhard Schölkopf: Towards causal deep learning (Session chair: Robert Jenssen, UiT)
10:00 Coffee break
10:30-11:30 Oral Session 1 (Session chair: Rudolf Mester, NTNU)
10:30 Talk 1: Pipeline for Artificial Generation of Echocardiography Datasets, Andrew Gilbert
10:50 Talk 2: Joint Attention Neural Model for Demand Prediction in Online Marketplaces, Ashish Gupta
11:10 Talk 3: Efficient Normalizing Flows to Polytopes, Sebastian Mair
11:30 Lunch
12:30 Invited Talk by Julia Schnabel: Deep Learning for Smart Medical imaging (Session chair: Anne Solberg, UiO)
13:15-13:55 Oral Session 2 (Session chair: Narada Warakagoda, FFI)
13:15 Talk 4: An Appropriate Prior Distribution for Interpolating Latent Samples in Flow-based Generative Models, Samuel Gomes Fadel
13:35 Talk 5: Enforcing perceptual consistency on Generative Adversarial Networks by using the Normalised Laplacian Pyramid Distance, Alexander Hepburn
13:55 Posters and Coffee (Session chair: Jonas Myhre and Karl Øyvind Mikalsen, UiT)
- An Automatic sorting of lily bulbs in production using deep learning, Uldanay Bairam
- Controlling Blood Glucose For Patients With Type 1 DiabetesUsing Deep Reinforcement Learning - The Influence Of Changing The Reward Function, Miguel Angel Tejedor Hernandez
- Focused losses for intersection-over-union measures, David Völgyes
- Code-Aligned Autoencoders for Multimodal Change Detection in Remote Sensing Images, Luigi Tommaso Luppino
- Using Deep Learning Methods to Monitor Non-Observable States in a Building, Kristoffer Tangrand
- PointNet and geometric reasoning for detection of grape vines from single frame RGB-D data in outdoor conditions, Polina Kurtser
- Machine listening in spatial acoustic scenes with deep networks in different microphone geometries, Jörn Anemüller
- Multi-output prediction of global vegetation distribution with incomplete data, Rita Beigaite
- Automatic interpretation of salmon scales using deep learning, Ketil Malde
- Deep canonical correlation analysis using robust correlation-objective on small data sets, Kasper Einarson
- Towards detection and classification of microscopic foraminifera using transfer learning, Thomas Haugland Johansen
- Estimation of end-diastole in Spectral Doppler Images Using Deep Learning, Tollef Jahren
- Cross-Encoder for Learning Word Meta Embeddings, Rickard Brannvall
- Short-Term Load Forecasting with Dilated Recurrent Attention Networks in Presence of Missing Data, Chankyu Choi
- Data-Driven Spectrum Cartography via Deep Completion Autoencoders, Yves Teganya
- mJ-Net: CNN based segmentation of infarcted regions in ischemic cerebral stroke from CTP imaging, Luca Tomasetti
- Deep Learning Footprints on Synaptic Vesicle Detection Using Electron Microscopy Images, Mahdieh Khanmohammadi
- Word2Vec: Optimal Hyper-parameters & Their Impact on NLP Downstream Tasks, Oluwatosin Adewumi
- Understanding a neural network decision by visualizing it: application to fish age prediction, Alba Ordonez
- Visual Object Detection For Autonomous UAV Cinematography, Fotini Patrona
- Tumor Detection in Brain MRIs by Computing Dissimilarities in the Latent Space of a Variational AutoEncoder, Alexandra-Ioana Albu, Alina Enescu, and L. Malagò
- A Robustness Analysis of Personalized Propagation of Neural Prediction, Chankyu Choi
- Towards a Framework for Noctilucent Cloud Analysis, Puneet Sharma
- CNNs for sequence-to-structure mapping in proteins, Marloes Arts
- Evaluating the Robustness of Defense Mechanisms based on AutoEncoder Reconstructions against Carlini-Wagner Adversarial Attacks, Petru Hlihor, Riccardo Volpi, and L. Malagò
15:40-16:40 Oral Session 3 (Session chair: Aase Reyes, OsloMet)
15:40 Talk 7: On the Delta Method for Uncertainty Approximation in Deep Learning, Geir Nilsen
16:00 Talk 8: Modelling the Information Plane of Recurrent Neural Networks, Kristoffer Wickstrøm
16:20 Talk 9: Fast reasoning visualization for deep convolutional networks, Marianne Bakken
18:25 Northern light activity (See info below)
Tuesday January 21, 2020
Location: Location: University of Tromsø - The Arctic University of Norway (Aud Cerebrum, MH2 building, U.08.316)
09:00 Invited Talk by Dino Sejdinovic: Kernel Embeddings for Meta Learning (Session chair: Arnt-Børre Salberg, NR)
09:45 Coffee break
10:10-11:30 Oral Session 4 (Session chair: Line Eikvil, NR)
10:10 Talk 10: Generative Adversarial Immitation Learning for Steering an Unmanned Surface Vehicle, Alexandra Vedeler
10:30 Talk 11: Towards automated marine biodiversity monitoring with taxonomic rank classification, Freek Daniëls
10:50 Talk 12: Perceiving Music Quality with GANs, Agrin Hilmkil
11:10 Talk 13: Co-training in deep learning for small non-translational data sets, Line Clemmensen
11:30 Lunch
12:30 Invited Talk by Christoph H. Lampert: Efficient and Adaptive Models for Visual Scene Analysis (Session chair: Michael Kampffmeyer, UiT)
13:15 NORA Panel Discussion (Moderator: Klas Pettersen, NORA)
14:00 Coffee break
14:20 Invited Talk by Xiaoxiang Zhu: AI and Data Science in Earth Observation (Session chair: Stian Anfinsen, UiT)
15:05-16:05 Oral Session 5 (Session chair: Nello Blaser, UiB)
15:05 Talk 14: The problems with using STNs to align CNN feature maps, Ylva Jansson
15:25 Talk 15: Evaluating the Robustness of Defense Mechanisms based on AutoEncoder Reconstructions against Carlini-Wagner Adversarial Attacks, Petru Hlihor
15:45 Talk 16: Notes on the Symmetries of 2-Layer ReLU-Networks, Henning Petzka
16:05 Closing
16:30 Leaving for dinner
17:00 Workshop dinner (until around 19:30)
NORA Panel Discussion
The Norwegian Artificial Intelligence Research Consortium – aims to strengthen Norwegian research and education within artificial intelligence, machine learning and robotics.
NORA is a collaboration between the University of Oslo (UiO), University of Bergen (UiB), University of Stavanger (UiS), The Arctic University of Norway (UiT), Oslo Metropolitan University (OsloMet), University of Agder (UiA), Norwegian University of Life Sciences (NMBU), NORCE Norwegian Research Centre AS (NORCE) and Simula Research Laboratory (Simula).
Discover Northern Lights at Sommarøy 20th January 2020
Sommarøy is a natural gem by the ocean, at the outer coast, a nice one-hour drive from Tromsø. A vibrant and traditional coastal community, where hunting, fishing, and trapping has been a part of daily life.
We invite to dinner in beautiful surroundings. The dinner will be served inside but you will have the opportunity to take a walk outside.
Experience the North-Norwegian winter with Aurora Borealis. (We hope!)
Dress code: Casual (the environment is rough but nice!)
Price: NOK 1090 per person;
Duration: 4 hours
Pickup/Drop off: Scandic Ishavshotell 18:30/23:00
Included: Transport (75 minute each way), dinner buffet and 1 glass of wine/beer/mineral water.