Institution
Ghent University
Education•Ghent, Belgium•
About: Ghent University is a education organization based out in Ghent, Belgium. It is known for research contribution in the topics: Population & Context (language use). The organization has 36170 authors who have published 111042 publications receiving 3774501 citations. The organization is also known as: UGent & University of Ghent.
Papers published on a yearly basis
Papers
More filters
••
TL;DR: This research attacked the mode of reinforcement learning in mice by developing a probabilistic approach to assess the importance of social reinforcement in the development of anxiety and depression in mice.
Abstract: Department of Experimental-Clinical and Health Psychology, Ghent University, Ghent, Belgium Research Institute for Psychology & Health, Utrecht, The Netherlands Department of Psychology, University of British Columbia, Vancouver, BC, Canada Academic Unit of Psychiatry & Behavioural Sciences, University of Leeds, Leeds, UK Department of Psychology, University of Montreal, Quebec, Canada Sub-Department of Clinical Health Psychology, University College London, London, UK Department of Psychology, Wayne State University, Detroit, MI, USA
520 citations
••
26 Jun 2018TL;DR: This system can learn quadruped locomotion from scratch using simple reward signals and users can provide an open loop reference to guide the learning process when more control over the learned gait is needed.
Abstract: Designing agile locomotion for quadruped robots often requires extensive expertise and tedious manual tuning. In this paper, we present a system to automate this process by leveraging deep reinforcement learning techniques. Our system can learn quadruped locomotion from scratch using simple reward signals. In addition, users can provide an open loop reference to guide the learning process when more control over the learned gait is needed. The control policies are learned in a physics simulator and then deployed on real robots. In robotics, policies trained in simulation often do not transfer to the real world. We narrow this reality gap by improving the physics simulator and learning robust policies. We improve the simulation using system identification, developing an accurate actuator model and simulating latency. We learn robust controllers by randomizing the physical environments, adding perturbations and designing a compact observation space. We evaluate our system on two agile locomotion gaits: trotting and galloping. After learning in simulation, a quadruped robot can successfully perform both gaits in the real world.
520 citations
••
TL;DR: This work replaced the thymidylate synthase gene thyA of L. lactis with a synthetic human IL10 gene, which produced human IL-10 (hIL-10), and when deprived of thymidine or thymine, its viability dropped by several orders of magnitude, essentially preventing its accumulation in the environment.
Abstract: Genetically modified Lactococcus lactis secreting interleukin 10 provides a therapeutic approach for inflammatory bowel disease. However, the release of such genetically modified organisms through clinical use raises safety concerns. In an effort to address this problem, we replaced the thymidylate synthase gene thyA of L. lactis with a synthetic human IL10 gene. This thyA−hIL10+ L. lactis strain produced human IL-10 (hIL-10), and when deprived of thymidine or thymine, its viability dropped by several orders of magnitude, essentially preventing its accumulation in the environment. The biological containment system and the bacterium's capacity to secrete hIL-10 were validated in vivo in pigs. Our approach is a promising one for transgene containment because, in the unlikely event that the engineered L. lactis strain acquired an intact thyA gene from a donor such as L. lactis subsp. cremoris, the transgene would be eliminated from the genome.
520 citations
•
TL;DR: A single-layer recurrent neural network with a dual softmax layer that matches the quality of the state-of-the-art WaveNet model, the WaveRNN, and a new generation scheme based on subscaling that folds a long sequence into a batch of shorter sequences and allows one to generate multiple samples at once.
Abstract: Sequential models achieve state-of-the-art results in audio, visual and textual domains with respect to both estimating the data distribution and generating high-quality samples. Efficient sampling for this class of models has however remained an elusive problem. With a focus on text-to-speech synthesis, we describe a set of general techniques for reducing sampling time while maintaining high output quality. We first describe a single-layer recurrent neural network, the WaveRNN, with a dual softmax layer that matches the quality of the state-of-the-art WaveNet model. The compact form of the network makes it possible to generate 24kHz 16-bit audio 4x faster than real time on a GPU. Second, we apply a weight pruning technique to reduce the number of weights in the WaveRNN. We find that, for a constant number of parameters, large sparse networks perform better than small dense networks and this relationship holds for sparsity levels beyond 96%. The small number of weights in a Sparse WaveRNN makes it possible to sample high-fidelity audio on a mobile CPU in real time. Finally, we propose a new generation scheme based on subscaling that folds a long sequence into a batch of shorter sequences and allows one to generate multiple samples at once. The Subscale WaveRNN produces 16 samples per step without loss of quality and offers an orthogonal method for increasing sampling efficiency.
520 citations
••
TL;DR: It is shown that heavily degenerated block duplications that can no longer be recognized by directly comparing two segments because of differential gene loss, can still be detected through indirect comparison with other segments and strongly implies that Arabidopsis has undergone three, but probably no more, rounds of genome duplications.
Abstract: Analysis of the genome sequence of Arabidopsis thaliana shows that this genome, like that of many other eukaryotic organisms, has undergone large-scale gene duplications or even duplications of the entire genome. However, the high frequency of gene loss after duplication events reduces colinearity and therefore the chance of finding duplicated regions that, at the extreme, no longer share homologous genes. In this study we show that heavily degenerated block duplications that can no longer be recognized by directly comparing two segments because of differential gene loss, can still be detected through indirect comparison with other segments. When these so-called hidden duplications in Arabidopsis are taken into account, many homologous genomic regions can be found in five to eight copies. This finding strongly implies that Arabidopsis has undergone three, but probably no more, rounds of genome duplications. Therefore, adding such hidden blocks to the duplication landscape of Arabidopsis sheds light on the number of polyploidy events that this model plant genome has undergone in its evolutionary past.
520 citations
Authors
Showing all 36585 results
Name | H-index | Papers | Citations |
---|---|---|---|
Stephen V. Faraone | 188 | 1427 | 140298 |
Peter Carmeliet | 164 | 844 | 122918 |
Monique M.B. Breteler | 159 | 546 | 93762 |
Dirk Inzé | 149 | 647 | 74468 |
Rajesh Kumar | 149 | 4439 | 140830 |
Vishva M. Dixit | 145 | 355 | 96471 |
Ruth J. F. Loos | 142 | 647 | 92485 |
Martin Grunewald | 140 | 1575 | 126911 |
Willy Verstraete | 139 | 920 | 76659 |
Barbara Clerbaux | 138 | 1394 | 96447 |
Peter Vandenabeele | 135 | 729 | 81692 |
Michael Tytgat | 134 | 1449 | 94133 |
Pascal Vanlaer | 133 | 1270 | 91850 |
Filip Moortgat | 132 | 1118 | 97714 |
Emelia J. Benjamin | 131 | 640 | 99972 |