École Normale Supérieure

About: École Normale Supérieure is a other organization based out in Paris, Île-de-France, France. It is known for research contribution in the topics: Population & Catalysis. The organization has 68439 authors who have published 99414 publications receiving 3092008 citations.

Non-local sparse models for image restoration

Abstract: We propose in this paper to unify two different approaches to image restoration: On the one hand, learning a basis set (dictionary) adapted to sparse signal descriptions has proven to be very effective in image reconstruction and classification tasks. On the other hand, explicitly exploiting the self-similarities of natural images has led to the successful non-local means approach to image restoration. We propose simultaneous sparse coding as a framework for combining these two approaches in a natural manner. This is achieved by jointly decomposing groups of similar signals on subsets of the learned dictionary. Experimental results in image denoising and demosaicking tasks with synthetic and real noise show that the proposed method outperforms the state of the art, making it possible to effectively restore raw images from digital cameras at a reasonable speed and memory cost.

Stability of organic carbon in deep soil layers controlled by fresh carbon supply

Abstract: The world's soils store more carbon than is present in biomass and in the atmosphere. New experimental evidence suggests that the delivery of fresh plant-derived carbon to the subsoil stimulates microbial activity and results in mineralization of thousand-year-old carbon. This supports the recent proposal that the conservation of organic carbon at depth results from a lack of energy for decomposers. This large pool of deep carbon is unlikely to respond to future changes in temperature if no fresh carbon is supplied, limiting the predicted positive feedback between global warming and soil organic carbon decomposition. The results imply that management practices that increase the distribution of fresh carbon along the soil profile (such as deep ploughing and the use of drought-resistant crops with extensive root systems) will stimulate loss of this ancient buried carbon. It is shown that the supply of fresh plant-derived carbon to deep soil layers stimulated the microbial mineralization of carbon that is thousands of years old, and is suggested that a lack of supply of fresh-carbon may prevent the decomposition of the organic carbon pool in deep soil layers in response to future changes in temperature. The world’s soils store more carbon than is present in biomass and in the atmosphere1. Little is known, however, about the factors controlling the stability of soil organic carbon stocks2,3,4 and the response of the soil carbon pool to climate change remains uncertain5,6. We investigated the stability of carbon in deep soil layers in one soil profile by combining physical and chemical characterization of organic carbon, soil incubations and radiocarbon dating. Here we show that the supply of fresh plant-derived carbon to the subsoil (0.6–0.8 m depth) stimulated the microbial mineralization of 2,567 ± 226-year-old carbon. Our results support the previously suggested idea7 that in the absence of fresh organic carbon, an essential source of energy for soil microbes, the stability of organic carbon in deep soil layers is maintained. We propose that a lack of supply of fresh carbon may prevent the decomposition of the organic carbon pool in deep soil layers in response to future changes in temperature. Any change in land use and agricultural practice that increases the distribution of fresh carbon along the soil profile1,8,9 could however stimulate the loss of ancient buried carbon.

The axial vector current in beta decay

Abstract: In order to derive in a convincing manner the formula of Goldberger and Treiman for the rate of charged pion decay, we consider the possibility that the divergence of the axial vector current in β-decay may be proportional to the pion field. Three models of the pion-nucleon interaction (and the weak current) are presented that have the required property. The first, using gradient coupling, has the advantage that it is easily generalized to strange particles, but the disadvantages of being unrenormalizable and of bringing in the vector and axial vector currents in an unsymmetrical way. The second model, using a strong interaction proposed bySchwinger and a weak current proposed byPolkinghorne, is renormalizable and symmetrical betweenV andA, but it involves postulating a new particle and is hard to extend to strange particles. The third model resembles the second one except that it is not necessary to introduce a new particle. (Renormalizability in the usual sense is then lost, however). Further research along these lines is suggested, including consideration of the possibility that the pion decay rate may be plausibly obtained under less stringent conditions.

Muscarinic activation of ionic currents measured by a new whole-cell recording method

Abstract: A new method is described as an alternative to whole-cell recording in order to prevent "wash-out" of the muscarinic response to acetylcholine (ACh) in rat lacrimal gland cells. The membrane of a cell-attached patch is permeabilized by nystatin in the patch pipette, thus providing electrical continuity between the pipette and the cytoplasm of the cell without the loss or alteration of cytoplasmic compounds necessary for the maintenance of the response to ACh. With normal whole-cell recording in these cells, the response to ACh, seen as the activation of Ca-activated K and Cl currents, lasts for approximately 5 min. With the nystatin method, the response is not diminished after 1 h. Nystatin, applied extracellularly, is shown to cause a rapid and reversible increase of membrane conductance to cations. In the absence of wash-out, we were able to obtain dose-response curves for the effect of ACh on Ca-activated K currents. An increase of [ACh] caused an increase in the K current, with apparent saturation at concentrations above approximately 1 microM ACh. The delay between ACh application and the activation of K current was inversely related to [ACh] and reached a minimum value of 0.7-1.0 s at high [ACh].

NetVLAD: CNN Architecture for Weakly Supervised Place Recognition

Abstract: We tackle the problem of large scale visual place recognition, where the task is to quickly and accurately recognize the location of a given query photograph. We present the following three principal contributions. First, we develop a convolutional neural network (CNN) architecture that is trainable in an end-to-end manner directly for the place recognition task. The main component of this architecture, NetVLAD, is a new generalized VLAD layer, inspired by the "Vector of Locally Aggregated Descriptors" image representation commonly used in image retrieval. The layer is readily pluggable into any CNN architecture and amenable to training via backpropagation. Second, we develop a training procedure, based on a new weakly supervised ranking loss, to learn parameters of the architecture in an end-to-end manner from images depicting the same places over time downloaded from Google Street View Time Machine. Finally, we show that the proposed architecture significantly outperforms non-learnt image representations and off-the-shelf CNN descriptors on two challenging place recognition benchmarks, and improves over current state of-the-art compact image representations on standard image retrieval benchmarks.