Chinese Academy of Sciences

About: Chinese Academy of Sciences is a government organization based out in Beijing, Beijing, China. It is known for research contribution in the topics: Catalysis & Population. The organization has 421602 authors who have published 634849 publications receiving 14894293 citations. The organization is also known as: CAS.

A Comprehensive Survey on Transfer Learning

Abstract: Transfer learning aims at improving the performance of target learners on target domains by transferring the knowledge contained in different but related source domains. In this way, the dependence on a large number of target-domain data can be reduced for constructing target learners. Due to the wide application prospects, transfer learning has become a popular and promising area in machine learning. Although there are already some valuable and impressive surveys on transfer learning, these surveys introduce approaches in a relatively isolated way and lack the recent advances in transfer learning. Due to the rapid expansion of the transfer learning area, it is both necessary and challenging to comprehensively review the relevant studies. This survey attempts to connect and systematize the existing transfer learning research studies, as well as to summarize and interpret the mechanisms and the strategies of transfer learning in a comprehensive way, which may help readers have a better understanding of the current research status and ideas. Unlike previous surveys, this survey article reviews more than 40 representative transfer learning approaches, especially homogeneous transfer learning approaches, from the perspectives of data and model. The applications of transfer learning are also briefly introduced. In order to show the performance of different transfer learning models, over 20 representative transfer learning models are used for experiments. The models are performed on three different data sets, that is, Amazon Reviews, Reuters-21578, and Office-31, and the experimental results demonstrate the importance of selecting appropriate transfer learning models for different applications in practice.

The Herschel-SPIRE instrument and its in-flight performance

Abstract: The Spectral and Photometric Imaging REceiver (SPIRE), is the Herschel Space Observatory`s submillimetre camera and spectrometer It contains a three-band imaging photometer operating at 250, 350 and 500 mu m, and an imaging Fourier-transform spectrometer (FTS) which covers simultaneously its whole operating range of 194-671 mu m (447-1550 GHz) The SPIRE detectors are arrays of feedhorn-coupled bolometers cooled to 03 K The photometer has a field of view of 4' x 8', observed simultaneously in the three spectral bands Its main operating mode is scan-mapping, whereby the field of view is scanned across the sky to achieve full spatial sampling and to cover large areas if desired The spectrometer has an approximately circular field of view with a diameter of 26' The spectral resolution can be adjusted between 12 and 25 GHz by changing the stroke length of the FTS scan mirror Its main operating mode involves a fixed telescope pointing with multiple scans of the FTS mirror to acquire spectral data For extended source measurements, multiple position offsets are implemented by means of an internal beam steering mirror to achieve the desired spatial sampling and by rastering of the telescope pointing to map areas larger than the field of view The SPIRE instrument consists of a cold focal plane unit located inside the Herschel cryostat and warm electronics units, located on the spacecraft Service Module, for instrument control and data handling Science data are transmitted to Earth with no on-board data compression, and processed by automatic pipelines to produce calibrated science products The in-flight performance of the instrument matches or exceeds predictions based on pre-launch testing and modelling: the photometer sensitivity is comparable to or slightly better than estimated pre-launch, and the spectrometer sensitivity is also better by a factor of 15-2

The clustering of galaxies in the completed SDSS-III Baryon Oscillation Spectroscopic Survey: cosmological analysis of the DR12 galaxy sample

Abstract: We present cosmological results from the final galaxy clustering data set of the Baryon Oscillation Spectroscopic Survey, part of the Sloan Digital Sky Survey III. Our combined galaxy sample comprises 1.2 million massive galaxies over an effective area of 9329 deg^2 and volume of 18.7 Gpc^3, divided into three partially overlapping redshift slices centred at effective redshifts 0.38, 0.51 and 0.61. We measure the angular diameter distance and Hubble parameter H from the baryon acoustic oscillation (BAO) method, in combination with a cosmic microwave background prior on the sound horizon scale, after applying reconstruction to reduce non-linear effects on the BAO feature. Using the anisotropic clustering of the pre-reconstruction density field, we measure the product D_MH from the Alcock–Paczynski (AP) effect and the growth of structure, quantified by fσ_8(z), from redshift-space distortions (RSD). We combine individual measurements presented in seven companion papers into a set of consensus values and likelihoods, obtaining constraints that are tighter and more robust than those from any one method; in particular, the AP measurement from sub-BAO scales sharpens constraints from post-reconstruction BAOs by breaking degeneracy between D_M and H. Combined with Planck 2016 cosmic microwave background measurements, our distance scale measurements simultaneously imply curvature Ω_K = 0.0003 ± 0.0026 and a dark energy equation-of-state parameter w = −1.01 ± 0.06, in strong affirmation of the spatially flat cold dark matter (CDM) model with a cosmological constant (ΛCDM). Our RSD measurements of fσ_8, at 6 per cent precision, are similarly consistent with this model. When combined with supernova Ia data, we find H_0 = 67.3 ± 1.0 km s^−1 Mpc^−1 even for our most general dark energy model, in tension with some direct measurements. Adding extra relativistic species as a degree of freedom loosens the constraint only slightly, to H_0 = 67.8 ± 1.2 km s^−1 Mpc^−1. Assuming flat ΛCDM, we find Ω_m = 0.310 ± 0.005 and H_0 = 67.6 ± 0.5 km s^−1 Mpc^−1, and we find a 95 per cent upper limit of 0.16 eV c^−2 on the neutrino mass sum.

Tet-Mediated Formation of 5-Carboxylcytosine and Its Excision by TDG in Mammalian DNA

Abstract: The prevalent DNA modification in higher organisms is the methylation of cytosine to 5-methylcytosine (5mC), which is partially converted to 5-hydroxymethylcytosine (5hmC) by the Tet (ten eleven translocation) family of dioxygenases. Despite their importance in epigenetic regulation, it is unclear how these cytosine modifications are reversed. Here, we demonstrate that 5mC and 5hmC in DNA are oxidized to 5-carboxylcytosine (5caC) by Tet dioxygenases in vitro and in cultured cells. 5caC is specifically recognized and excised by thymine-DNA glycosylase (TDG). Depletion of TDG in mouse embyronic stem cells leads to accumulation of 5caC to a readily detectable level. These data suggest that oxidation of 5mC by Tet proteins followed by TDG-mediated base excision of 5caC constitutes a pathway for active DNA demethylation.

Action recognition by dense trajectories

Abstract: Feature trajectories have shown to be efficient for representing videos. Typically, they are extracted using the KLT tracker or matching SIFT descriptors between frames. However, the quality as well as quantity of these trajectories is often not sufficient. Inspired by the recent success of dense sampling in image classification, we propose an approach to describe videos by dense trajectories. We sample dense points from each frame and track them based on displacement information from a dense optical flow field. Given a state-of-the-art optical flow algorithm, our trajectories are robust to fast irregular motions as well as shot boundaries. Additionally, dense trajectories cover the motion information in videos well. We, also, investigate how to design descriptors to encode the trajectory information. We introduce a novel descriptor based on motion boundary histograms, which is robust to camera motion. This descriptor consistently outperforms other state-of-the-art descriptors, in particular in uncontrolled realistic videos. We evaluate our video description in the context of action classification with a bag-of-features approach. Experimental results show a significant improvement over the state of the art on four datasets of varying difficulty, i.e. KTH, YouTube, Hollywood2 and UCF sports.