Filters








1,170 Hits in 5.4 sec

Analysis of Explainers of Black Box Deep Neural Networks for Computer Vision: A Survey [article]

Vanessa Buhrmester, David Münch, Michael Arens
2019 arXiv   pre-print
In this survey we differ the mechanisms and properties of explaining systems for Deep Neural Networks for Computer Vision tasks.  ...  As a black box model due to their multilayer nonlinear structure, Deep Neural Networks are often criticized to be non-transparent and their predictions not traceable by humans.  ...  Most used DNNs for image or video processing [39] , [40] , [41] are Convolutional Neural Networks (CNNs) [42] , for videos [41] or sequences of text Recurrent Neural Networks (RNNs) [43] , and  ... 
arXiv:1911.12116v1 fatcat:qgeg6rz6qzgrfikhsgah77yz2a

Deep Learning of Representations: Looking Forward [article]

Yoshua Bengio
2013 arXiv   pre-print
Deep learning research aims at discovering learning algorithms that discover multiple levels of distributed representations, with higher levels representing more abstract concepts.  ...  to much larger models and datasets, reducing optimization difficulties due to ill-conditioning or local minima, designing more efficient and powerful inference and sampling procedures, and learning to disentangle  ...  He is also grateful for the funding support from NSERC, CIFAR, the Canada Research Chairs, and Compute Canada.  ... 
arXiv:1305.0445v2 fatcat:cyfgf5trljfopcjsapicf4ay3q

Enforcing and Discovering Structure in Machine Learning [article]

Francesco Locatello
2021 arXiv   pre-print
Stopping criterion We propose a stopping criterion for boosting VI, which allows us to identify when a reasonably good approximation is reached and save computational effort.  ...  This indicates that the specific representation used matters and that predictions with lower unfairness can be achieved.  ...  The best weakly-supervised disentanglement methods thus learn representations that are useful for training accurate classifiers downstream.  ... 
arXiv:2111.13693v1 fatcat:2urmfjeh6nhvjeiv3qoiy5hrum

RGB-D-based Human Motion Recognition with Deep Learning: A Survey [article]

Pichao Wang and Wanqing Li and Philip Ogunbona and Jun Wan and Sergio Escalera
2018 arXiv   pre-print
Specifically, deep learning methods based on the CNN and RNN architectures have been adopted for motion recognition using RGB-D data.  ...  Particularly, we highlighted the methods of encoding spatial-temporal-structural information inherent in video sequence, and discuss potential directions for future research.  ...  [131] proposed a Disentangled Representation learning Generative Adversarial Networks (DR-GAN) for pose-invariant face recognition.  ... 
arXiv:1711.08362v2 fatcat:cugugpqeffcshnwwto4z2aw4ti

Interpretable Machine Learning: Fundamental Principles and 10 Grand Challenges [article]

Cynthia Rudin, Chaofan Chen, Zhi Chen, Haiyang Huang, Lesia Semenova, Chudi Zhong
2021 arXiv   pre-print
causal inference; (5) Complete supervised disentanglement of neural networks; (6) Complete or even partial unsupervised disentanglement of neural networks; (7) Dimensionality reduction for data visualization  ...  Interpretability in machine learning (ML) is crucial for high stakes decisions and troubleshooting.  ...  Acknowledgments We thank Leonardo Lucio Custode for pointing out several useful references to Challenge 10. Thank you to David Page for providing useful references on early explainable ML.  ... 
arXiv:2103.11251v2 fatcat:52llnswt3ze5rl3zhbai5bscce

Facial Keypoint Sequence Generation from Audio [article]

Prateek Manocha, Prithwijit Guha
2020 arXiv   pre-print
This dataset is then further used to train the model, Audio2Keypoint, a novel approach for synthesizing facial keypoint movement to go with the audio.  ...  To address this, a unique audio-keypoint dataset of over 150,000 videos at 224p and 25fps is introduced that relates the facial keypoint movement for the given audio.  ...  Later, these can be used for the generation of photo-realistic videos.  ... 
arXiv:2011.01114v1 fatcat:ru3q4xmqgrgoph3tcqlgah26bq

VirtualCube: An Immersive 3D Video Communication System [article]

Yizhong Zhang, Jiaolong Yang, Zhen Liu, Ruicheng Wang, Guojun Chen, Xin Tong, Baining Guo
2021 arXiv   pre-print
The key ingredient is VirtualCube, an abstract representation of a real-world cubicle instrumented with RGBD cameras for capturing the 3D geometry and texture of a user.  ...  To achieve real-time rendering of remote participants, we develop the V-Cube View algorithm, which uses multi-view stereo for more accurate depth estimation and Lumi-Net rendering for better rendering  ...  adversarially disentangled audio-visual representation. In AAAI, vol. 33, Kang, P. Kohli, Y. Lutchyn, C. Keskin, and S. Izadi.  ... 
arXiv:2112.06730v2 fatcat:ce7dls4fvbf7zbyqdn7mx4hvpe

Towards Causal Representation Learning [article]

Bernhard Schölkopf, Francesco Locatello, Stefan Bauer, Nan Rosemary Ke, Nal Kalchbrenner, Anirudh Goyal, Yoshua Bengio
2021 arXiv   pre-print
A central problem for AI and causality is, thus, causal representation learning, the discovery of high-level causal variables from low-level observations.  ...  Finally, we delineate some implications of causality for machine learning and propose key research areas at the intersection of both communities.  ...  Thanks to Wouter van Amsterdam for pointing out typos in the first version. We also thank Thomas Kipf, Klaus Greff, and Alexander d'Amour for the useful discussions.  ... 
arXiv:2102.11107v1 fatcat:n25xwac72nfulgl3gvvs4kerca

Toward Causal Representation Learning

Bernhard Scholkopf, Francesco Locatello, Stefan Bauer, Nan Rosemary Ke, Nal Kalchbrenner, Anirudh Goyal, Yoshua Bengio
2021 Proceedings of the IEEE  
A central problem for AI and causality is, thus, causal representation learning, that is, the discovery of highlevel causal variables from low-level observations.  ...  Finally, we delineate some implications of causality for machine learning and propose key research areas at the intersection of both communities.  ...  It has recently been used for learning causal models [131] , modular architectures [29] , [85] , and disentangled representations [159] .  ... 
doi:10.1109/jproc.2021.3058954 fatcat:jqg6jm2f35aynlszy6w5nsyem4

SAS: Self-Augmentation Strategy for Language Model Pre-training [article]

Yifei Xu, Jingqiao Zhang, Ru He, Liangzhu Ge, Chao Yang, Cheng Yang, Ying Nian Wu
2021 arXiv   pre-print
In this paper, we propose a self-augmentation strategy (SAS) where a single network is utilized for both regular pre-training and contextualized data augmentation for the training in later epochs.  ...  In addition, SAS is a general strategy that can be seamlessly combined with many new techniques emerging recently or in the future, such as the disentangled attention mechanism from DeBERTa.  ...  The Electric model is later proposed by Clark et al. (2020a) as an energy-based model to perform the cloze task (Taylor 1953 ) using noise-contrastive estimation (Gutmann and Hyvärinen 2012) .  ... 
arXiv:2106.07176v3 fatcat:6nat7wiebzgtvciy7qo7aoydmu

Deep Learning of Representations: Looking Forward [chapter]

Yoshua Bengio
2013 Lecture Notes in Computer Science  
This has led to various proposals for sampling from this implicitly learned density function, using Langevin and Metropolis-Hastings MCMC.  ...  Whereas previous ways of dealing with the output bottleneck involved non-distributed word classes, we introduce a distributed representation for discrete word classes along with techniques to turn an apparently  ...  He is also grateful for the funding support from NSERC, CIFAR, the Canada Research Chairs, and Compute Canada.  ... 
doi:10.1007/978-3-642-39593-2_1 fatcat:xad2okhdkrfbrhe4ilujsnoqlu

Representation Learning: A Review and New Perspectives

Y. Bengio, A. Courville, P. Vincent
2013 IEEE Transactions on Pattern Analysis and Machine Intelligence  
Although specific domain knowledge can be used to help design representations, learning with generic priors can also be used, and the quest for AI is motivating the design of more powerful representation-learning  ...  This motivates longer-term unanswered questions about the appropriate objectives for learning good representations, for computing representations (i.e., inference), and the geometrical connections between  ...  Acknowledgments The author would like to thank David Warde-Farley, Razvan Pascanu and Ian Goodfellow for useful feedback, as well as NSERC, CIFAR and the Canada Research Chairs for funding.  ... 
doi:10.1109/tpami.2013.50 pmid:23787338 fatcat:2ozfdsn2bjaa5jwtdovszpr5xa

Representation Learning: A Review and New Perspectives [article]

Yoshua Bengio and Aaron Courville and Pascal Vincent
2014 arXiv   pre-print
Although specific domain knowledge can be used to help design representations, learning with generic priors can also be used, and the quest for AI is motivating the design of more powerful representation-learning  ...  This motivates longer-term unanswered questions about the appropriate objectives for learning good representations, for computing representations (i.e., inference), and the geometrical connections between  ...  Acknowledgements The author would like to thank David Warde-Farley and Razvan Pascanu for useful feedback, as well as NSERC, CIFAR and the Canada Research Chairs for funding.  ... 
arXiv:1206.5538v3 fatcat:axiuzjr77zesvkaiewoh7cwmbq

"Here's to a Night of Drunken Mistakes": Exploring Experiences, Regrets, and Optimism in Young Adult Drinkers

Emma L. Davies, Mary Sissons Joshi
2018 Substance Use & Misuse  
Laterality 2018; ePub(ePub): ePub.  ...  Laterality 2018; ePub(ePub): ePub.  ...  (Copyright 2018, Informa -Taylor and Francis Group) Hospital to school transition following traumatic brain injury: a qualitative longitudinal study -Todis B, McCart M, Glang A.  ... 
doi:10.1080/10826084.2018.1461227 pmid:29676652 fatcat:5oedsemlf5ex5pqdvaifhzpssq

The influence of action perception on object recognition: a developmental study

Pierre Mounoud, Katia Duscherer, Guénaël Moy, Sandrine Perraudin
2007 Developmental Science  
Two experiments explored the existence and the development of relations between action representations and object representations.  ...  While substantial priming effects were obtained for all age groups, they were especially important for the youngest participants.  ...  We also want to thank Frédéric Michiels for miming the actions, Alain Perruchoud for creating the corresponding videos, Mathias Durrenberger for getting the program running and Aris Khan for his assistance  ... 
doi:10.1111/j.1467-7687.2007.00624.x pmid:17973800 fatcat:conve3npyjhfhgk2sny7xzzlfe
« Previous Showing results 1 — 15 out of 1,170 results