arXiv subjects

Wenbin Li

Publications and source records attributed to Wenbin Li.

167 records · Page 10Linked to original sources

arXiv2014

Learning Multi-Scale Representations for Material Classification

The recent progress in sparse coding and deep learning has made unsupervised feature learning methods a strong competitor to hand-crafted descriptors. In computer vision, success stories of learned features have been predominantly reported for object recognition tasks. In this paper, we investigate if and how feature learning can be used for material recognition. We propose two strategies to incorporate scale information into the learning procedure resulting in a novel multi-scale coding procedure. Our results show that our learned features for material recognition outperform hand-crafted descriptors on the FMD and the KTH-TIPS2 material classification benchmarks.

cs.CV↗

arXiv2014

Persistent Dirac Fermion State on Bulk-like Si(111) Surface

The "multilayer silicene" films were grown on Ag(111), with increasing thickness above 30 monolayers (ML). We found that the "multilayer silicene" is indeed a bulk Si(111) film. Such Si film on Ag(111) always exhibits a root(3)xroot(3) honeycomb superstructure on surface. Delocalized surface state as well as linear energy-momentum dispersion was revealed by quasiparticle interference patterns (QPI) on the surface, which proves the existence of Dirac fermions state. Our results indicate that bulk silicon with diamond structure can also host Dirac fermions, which makes the system even more attractive for further applications compared with monolayer silicene.

cond-mat.mes-hall↗

arXiv2009

Bayes estimators for phylogenetic reconstruction

Tree reconstruction methods are often judged by their accuracy, measured by how close they get to the true tree. Yet most reconstruction methods like ML do not explicitly maximize this accuracy. To address this problem, we propose a Bayesian solution. Given tree samples, we propose finding the tree estimate which is closest on average to the samples. This ``median'' tree is known as the Bayes estimator (BE). The BE literally maximizes posterior expected accuracy, measured in terms of closeness (distance) to the true tree. We discuss a unified framework of BE trees, focusing especially on tree distances which are expressible as squared euclidean distances. Notable examples include Robinson--Foulds distance, quartet distance, and squared path difference. Using simulated data, we show Bayes estimators can be efficiently computed in practice by hill climbing. We also show that Bayes estimators achieve higher accuracy, compared to maximum likelihood and neighbor joining.

q-bio.PE↗

arXiv2001

Photon Dominated Regions in Low UV Fields: A Study of The Peripheral Region of L1204/S140

We have carried out an in-depth study of the peripheral region of the molecular cloud L1204/S140, where the far ultraviolet radiation and the density are relatively low. Our observations test theories of photon-dominated regions (PDRs) in a regime that has been little explored. CII 158 micron and OI 63 micron lines are detected by ISO at all 16 positions along a 1-dimensional cut in right ascension. Emission from molecular hydrogen rotational transitions, at 28 and 17 micron, was also detected at several positions. The CII, OI, and molecular hydrogen intensities along the cut show much less spatial variation than do the rotational lines of CO and other CO isotopes. The average CII and OI intensities and their ratio are consistent with models of PDRs with low FUV radiation (Go) and density. The best-fitting model has Go about 15 and density, n about 1000 per cubic cm. Standard PDR models underpredict the intensity in the H2 rotational lines by up to an order of magnitude. This problem has also been seen in bright PDRs and attributed to factors, such as geometry and gas-grain drift, that should be much less important in the regime studied here. The fact that we see the same problem in our data suggests that more fundamental solutions, such as higher H2 formation rates, are needed. Also, in this regime of low density and small line width, the OI line is sensitive to the radiative transfer and geometry. Using the ionization structure of the models, a quantitative analysis of timescales for ambipolar diffusion in the peripheral regions of the S140 cloud is consistent with a theory of photoionization-regulated star formation. Observations of CII in other galaxies differ both from those of high Go PDRs in our galaxy and from the low Go regions we have studied. ~

astro-ph↗

arXiv1997

Looking for Distributed Star Formation in L1630: A Near-infrared (J, H, K) Survey

We have carried out a simultaneous, multi-band (J, H, K) survey over an area of 1320 arcmin^2 in the L1630 region, concentrating on the region away from the dense molecular cores and with modest visual extinctions (\leq 10 mag). Previous studies found that star formation in L1630 occurs mainly in four localized clusters, which in turn are associated with the four most massive molecular cores (Lada et al. 1991; Lada 1992). The goal of this study is to look for a distributed population of pre-main-sequence stars in the outlying areas outside the known star-forming cores. More than 60% of the pre-main-sequence stars in the active star forming regions of NGC 2024 and NGC 2023 show a near-infrared excess in the color-color diagram. In the outlying areas of L1630, excluding the known star forming regions, we found that among 510 infrared sources with the near-infrared colors ((J-H) and (H-K)) determined and photometric uncertainty at K better than 0.10 mag, the fraction of the sources with a near-infrared excess is 3%--8%; the surface density of the sources with a near-infrared excess is less than half of that found in the distributed population in L1641, and 1/20 of that in the young cluster NGC 2023. This extremely low fraction and low surface density of sources with a near-infrared excess strongly indicates that recent star formation activity has been very low in the outlying region of L1630. The sources without a near-infrared excess could be either background/foreground field stars, or associated with the cloud, but formed a long time ago (more than 2 Myrs). Our results are consistent with McKee's model of photoionization-regulated star formation.

astro-ph↗