Výsledky vyhledávání

Nested Variational Inference

Autor: Esmaeili, B., Wu, H., Zimmermann, H., Van De Meent, J.-W., Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Wortman Vaughan, J.

Publikováno v: 35th Conference on Neural Information Processing Systems (NeurIPS 2021): online, 6-14 December 2021, 25, 20423-20435

We develop nested variational inference (NVI), a family of methods that learn proposals for nested importance samplers by minimizing an forward or reverse KL divergence at each level of nesting. NVI is applicable to many commonly-used importance samp

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=narcis______::51785d16c0042dd1c49bf765a64d644a
https://dare.uva.nl/personal/pure/en/publications/nested-variational-inference(5f85ab15-e452-4026-88d8-c8a924df6392).html

Zobrazit plný text záznamu

Topographic VAEs learn Equivariant Capsules

Autor: Keller, T.A., Welling, M., Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Wortman Vaughan, J.

Publikováno v: 35th Conference on Neural Information Processing Systems (NeurIPS 2021): online, 6-14 December 2021, 34, 28585-28597

In this work we seek to bridge the concepts of topographic organization and equivariance in neural networks. To accomplish this, we introduce the Topographic VAE: a novel method for efficiently training deep generative models with topographically org

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::227cfa09cd39a2bdfdc98cdd5d2268c4
https://dare.uva.nl/personal/pure/en/publications/topographic-vaes-learn-equivariant-capsules(2f1c05d1-e81c-4b96-bf4f-aea3c8016fe4).html

Zobrazit plný text záznamu

Shape from Blur: Recovering Textured 3D Shape and Motion of Fast Moving Objects

Autor: Ferrari, V., Oswald, M.R., Pollefeys, M., Rozumnyi, D., Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Wortman Vaughan, J.

Publikováno v: 35th Conference on Neural Information Processing Systems (NeurIPS 2021): online, 6-14 December 2021, 36, 29972-29983

We address the novel task of jointly reconstructing the 3D shape, texture, andmotion of an object from a single motion-blurred image. While previous approachesaddress the deblurring problem only in the 2D image domain, our proposed rigorousmodeling o

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=narcis______::c66a903e85ceef8f49ce83f409deed86
https://dare.uva.nl/personal/pure/en/publications/shape-from-blur-recovering-textured-3d-shape-and-motion-of-fast-moving-objects(402ee23b-e119-4857-824f-b3604fe253e0).html

Zobrazit plný text záznamu

Argmax Flows and Multinomial Diffusion: Learning Categorical Distributions

Autor: Forre, P., Hoogeboom, E., Jaini, P., Nielsen, D., Welling, M., Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Wortman Vaughan, J.

Publikováno v: 35th Conference on Neural Information Processing Systems (NeurIPS 2021): online, 6-14 December 2021, 15, 12454-12465

Generative flows and diffusion models have been predominantly trained on ordinal data, for example natural images. This paper introduces two extensions of flows and diffusion for categorical data such as language or image segmentation: Argmax Flows a

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=narcis______::390b4179bf480870dd633eeb55ea60a9
https://dare.uva.nl/personal/pure/en/publications/argmax-flows-and-multinomial-diffusion-learning-categorical-distributions(4f332e79-de31-4e79-a875-76ce9385af54).html

Zobrazit plný text záznamu

Keeping Your Eye on the Ball: Trajectory Attention in Video Transformers

Autor: Asano, Y., Campbell, D., Feichtenhofer, C., Henriques, J., Metze, F., Misra, I., Patrick, M., Vedaldi, A., Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Wortman Vaughan, J.

Publikováno v: 35th Conference on Neural Information Processing Systems (NeurIPS 2021): online, 6-14 December 2021, 15, 12493-12506

In video transformers, the time dimension is often treated in the same way as the two spatial dimensions. However, in a scene where objects or the camera may move, a physical point imaged at one location in frame t may be entirely unrelated to what i

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=narcis______::84e87b61d3f2e1f7121b1a78b820aac2
https://dare.uva.nl/personal/pure/en/publications/keeping-your-eye-on-the-ball-trajectory-attention-in-video-transformers(65f2b469-7675-4555-ab2b-a191eb413e2b).html

Zobrazit plný text záznamu

Bias Out-of-the-Box: An Empirical Analysis of Intersectional Occupational Biases in Popular Generative Language Models

Autor: Asano, Y., Benussi, E., Dreyer, F., Iqbal, H., Kirk, H.R., Shtedritski, A., Volpin, F., Jun, Y., Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Wortman Vaughan, J.

Publikováno v: 35th Conference on Neural Information Processing Systems (NeurIPS 2021): online, 6-14 December 2021, 4, 2611-2624

The capabilities of natural language models trained on large-scale data have increased immensely over the past few years. Open source libraries such as HuggingFace have made these models easily available and accessible. While prior research has ident

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=narcis______::9e7e388c40bda154e333215beb221b4a
https://dare.uva.nl/personal/pure/en/publications/bias-outofthebox-an-empirical-analysis-of-intersectional-occupational-biases-in-popular-generative-language-models(00579d88-a26f-4a67-be26-b7c84f75c5b8).html

Zobrazit plný text záznamu

Roto-translated Local Coordinate Frames For Interacting Dynamical Systems

Autor: Gavves, E., Kofinas, M., Nagaraja, N., Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Wortman Vaughan, J.

Publikováno v: 35th Conference on Neural Information Processing Systems (NeurIPS 2021): online, 6-14 December 2021, 8, 6417-6429

Modelling interactions is critical in learning complex dynamical systems, namely systems of interacting objects with highly non-linear and time-dependent behaviour. A large class of such systems can be formalized as geometric graphs, i.e. graphs with

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=narcis______::2ecd09a02bdf3f61a92f09206ff007ff
https://dare.uva.nl/personal/pure/en/publications/rototranslated-local-coordinate-frames-for-interacting-dynamical-systems(5c8a8182-8dce-4fdb-bacf-98d987af0bfd).html

Zobrazit plný text záznamu

Hyperbolic Busemann Learning with Ideal Prototypes

Autor: Ghadimi Atigh, M., Keller-Ressel, M., Mettes, P., Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Wortman Vaughan, J.

Publikováno v: 35th Conference on Neural Information Processing Systems (NeurIPS 2021): online, 6-14 December 2021, 1, 103-115

Hyperbolic space has become a popular choice of manifold for representation learning of various datatypes from tree-like structures and text to graphs. Building on the success of deep learning with prototypes in Euclidean and hyperspherical spaces, a

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=narcis______::b641a1c65e04c07653844aaa020440bc
https://dare.uva.nl/personal/pure/en/publications/hyperbolic-busemann-learning-with-ideal-prototypes(8ea6d6ee-5303-4843-8e20-504c6fdf4d97).html

Zobrazit plný text záznamu

Truncated Marginal Neural Ratio Estimation

Autor: Cole, A., Forre, P., Louppe, G., Miller, B.K., Weniger, C., Ranzato, M., Beygelzimer, A., Dauphin, Y., Liang, P.S., Wortman Vaughan, J.

Publikováno v: 35th Conference on Neural Information Processing Systems (NeurIPS 2021): online, 6-14 December 2021, 1, 129-143

Parametric stochastic simulators are ubiquitous in science, often featuring high-dimensional input parameters and/or an intractable likelihood. Performing Bayesian parameter inference in this context can be challenging. We present a neural simulation

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=narcis______::952aededcc57e16d949a557c3e575561
https://dare.uva.nl/personal/pure/en/publications/truncated-marginal-neural-ratio-estimation(e6f23068-5d5b-492f-8318-25195a90abdb).html

Zobrazit plný text záznamu

Vyhledávací nástroje:

Upřesnit hledání