## ColdGAN Resolving Cold Start User Recommendation by using Generative Adversarial Networks

ColdGAN is an end-to-end GAN based model with no use of side information . Mitigating the new user cold-start problem has been critical in thereendation system for online service providers to influence user experience in decision making . The main idea of the proposed model is to train a network that learns the rating distributions of experienced users .…

## Combining Semantic Guidance and Deep Reinforcement Learning For Generating Human Level Paintings

Recent research efforts have been focused on teaching machines “howto paint”, in a manner similar to a human painter . Previous methods have been limited to datasets with little variation in position, scale and saliency of the foreground object . Semantic Guidancepipeline proposes bi-level painting procedure for learning the distinction between foreground and background brush strokes at training time .…

## Wasserstein k means with sparse simplex projection

This paper presents a proposal of a faster Wasserstein $k$-means algorithm for histogram data . We shrink data samples, centroids, and groundcost matrix, which leads to considerable reduction of the computations used to solve optimal transport problems without loss of clustering quality .…

## Denmark s Participation in the Search Engine TREC COVID 19 Challenge Lessons Learned about Searching for Precise Biomedical Scientific Information on COVID 19

Two Danish universities participated in the 2020 TREC-COVID Challenge . The aim of the competition was to find the best search engine strategy for retrieving precise biomedical scientificinformation on COVID-19 . CORD-19 was the result of a call to action to the tech community by the U.S.…

## Ranking Deep Learning Generalization using Label Variation in Latent Geometry Graphs

Latent Geometry Graphs (LGGs) are used to represent the latent spaces of trained DNN architectures . Such graphs are obtained by connecting samplesthat yield similar latent representations at a given layer of the consideredDNN . We then obtain a generalization score by looking at how strongly connectedare samples of distinct classes in LGGs .…

## Assessing the Quality of Gridded Population Data for Quantifying the Population Living in Deprived Communities

Over a billion people live in slums in settlements that are often located inecologically sensitive areas and hence highly vulnerable . This is a problem in many parts of the world, but it is more prominent in low-income countries . Building resilient communities requires quantifying the population of slums and improving their living conditions .…

Spatially decompose a scene and dedicate smaller networks for each decomposed part . This allows us near-constantinference time regardless of the number of decomposed parts . Voronoi spatial decomposition is preferable for this purpose, as it is compatible with the Painter’s Algorithm for efficient and GPU-friendlyrendering .…

## DeepKoCo Efficient latent planning with an invariant Koopman representation

DeepKoCo is a novel model-based agent that learns a latentKoopman representation from images . This representation allows the agent to planefficiently using linear control methods, such as linear model predictivecontrol . Compared to traditional agents, the agent is invariant totask-irrelevant dynamics, thanks to the use of a tailored lossy autoencodernetwork .…

## World Model as a Graph Learning Latent Landmarks for Planning

Planning – the ability to analyze the structure of a problem in the large – is a hallmark of humanintelligence . Deep reinforcement learning (RL) has shown great promise for solving relatively straightforward control tasks, but it remains an openproblem how to best incorporate planning into existing deep RL paradigms to handle increasingly complex environments .…

## Distributed Additive Encryption and Quantization for Privacy Preserving Federated Deep Learning

Homomorphic encryption is a very useful gradient protection technique used inprivacy preserving federated learning . But existing encrypted federatedlearning systems need a trusted third party to generate and distribute keypairs to connected participants . The proposed method avoidssencryption and decryption of the entire model .…

## AI virtues The missing link in putting AI ethics into practice

Several seminal ethics initiatives have stipulated sets of principles and standards for good technology development in the AI sector . However, widespreadcriticism has pointed out a lack of practical realization of these principles . Thispaper proposes a different approach to AI ethics .…

## Towards Playing Full MOBA Games with Deep Reinforcement Learning

MOBA games pose grand challenges to AI systems such as multi-agent, enormous state-action space, complex action control, etc. OpenAI’s Dota AI limits theplay to a pool of only 17 heroes. As a result, full MOBA . games without . restrictions are far from being mastered by any existing AI system.…

## Interpreting U Nets via Task Driven Multiscale Dictionary Learning

U-Nets have been tremendously successful in many imaging inverse problems . We show that one can reduce a U-Net to a tractable, well-understood sparsity-driven dictionary model . This model can be trained in a task-drivendictionary learning framework and yield comparable results to standard U-Ns on a number of relevant tasks, including CT and MRI reconstruction .…

## Wedge Lifted Codes

We define wedge-lifted codes, a variant of lifted codes, and study theirlocality properties . They give improved trade-offs between redundancy and locality among binary codes . We show that (taking the trace of) wedge-lift codesyields binary codes with the $t$-disjoint repair property ($t$)…

## Neural Representations for Modeling Variation in English Speech

Transformer-based speech representations lead to significant performance gains over the use of phonetic transcriptions . We use these representations to computeword-based pronunciation differences between non-native and native speakers of English . We alsodemonstrate that these neural speech representations capture segmentaldifferences, but also intonational and durational differences that cannot be represented by a set of discrete symbols used in phonetic transcripts .…

## Improving Augmentation and Evaluation Schemes for Semantic Image Synthesis

Little attention has been paid to augmentation strategies for generative adversarial networks (GANs) We introduce a novel augmentation scheme designed specifically forGAN-based semantic image synthesis models . We propose to randomly warp objectshapes in the semantic label maps used as an input to the generator .…