• ### Zero-Shot Kernel Learning(1802.01279)

Feb. 5, 2018 cs.CV
In this paper, we address an open problem of zero-shot learning. Its principle is based on learning a mapping that associates feature vectors extracted from i.e. images and attribute vectors that describe objects and/or scenes of interest. In turns, this allows classifying unseen object classes and/or scenes by matching feature vectors via mapping to a newly defined attribute vector describing a new class. Due to importance of such a learning task, there exist many methods that learn semantic, probabilistic, linear or piece-wise linear mappings. In contrast, we apply well-established kernel methods to learn a non-linear mapping between the feature and attribute spaces. We propose an easy learning objective inspired by the Linear Discriminant Analysis, Kernel-Target Alignment and Kernel Polarization methods that promotes incoherence. We evaluate performance of our algorithm on the Polynomial as well as shift-invariant Gaussian and Cauchy kernels. Despite simplicity of our approach, we obtain state-of-the-art results on several zero-shot learning datasets and benchmarks including a recent AWA2 dataset.
• ### Museum Exhibit Identification Challenge for Domain Adaptation and Beyond(1802.01093)

Feb. 4, 2018 cs.CV
In this paper, we approach an open problem of artwork identification and propose a new dataset dubbed Open Museum Identification Challenge (Open MIC). It contains photos of exhibits captured in 10 distinct exhibition spaces of several museums which showcase paintings, timepieces, sculptures, glassware, relics, science exhibits, natural history pieces, ceramics, pottery, tools and indigenous crafts. The goal of Open MIC is to stimulate research in domain adaptation, egocentric recognition and few-shot learning by providing a testbed complementary to the famous Office dataset which reaches 90% accuracy. To form our dataset, we captured a number of images per art piece with a mobile phone and wearable cameras to form the source and target data splits, respectively. To achieve robust baselines, we build on a recent approach that aligns per-class scatter matrices of the source and target CNN streams [15]. Moreover, we exploit the positive definite nature of such representations by using end-to-end Bregman divergences and the Riemannian metric. We present baselines such as training/evaluation per exhibition and training/evaluation on the combined set covering 866 exhibit identities. As each exhibition poses distinct challenges e.g., quality of lighting, motion blur, occlusions, clutter, viewpoint and scale variations, rotations, glares, transparency, non-planarity, clipping, we break down results w.r.t. these factors.
• ### Exploring the dark matter inelastic frontier with 79.6 days of PandaX-II data(1708.05825)

Oct. 3, 2017 hep-ph, hep-ex, astro-ph.CO
We report here the results of searching for inelastic scattering of dark matter (initial and final state dark matter particles differ by a small mass splitting) with nucleon with the first 79.6-day of PandaX-II data (Run 9). We set the upper limits for the spin independent WIMP-nucleon scattering cross section up to a mass splitting of 300 keV/c$^2$ at two benchmark dark matter masses of 1 and 10 TeV/c$^2$.
• ### Limits on Axion Couplings from the first 80-day data of PandaX-II Experiment(1707.07921)

Oct. 3, 2017 hep-ph, hep-ex, astro-ph.SR
We report new searches for the solar axions and galactic axion-like dark matter particles, using the first low-background data from PandaX-II experiment at China Jinping Underground Laboratory, corresponding to a total exposure of about $2.7\times 10^4$ kg$\cdot$day. No solar axion or galactic axion-like dark matter particle candidate has been identified. The upper limit on the axion-electron coupling ($g_{Ae}$) from the solar flux is found to be about $4.35 \times 10^{-12}$ in mass range from $10^{-5}$ to 1 keV/$c^2$ with 90\% confidence level, similar to the recent LUX result. We also report a new best limit from the $^{57}$Fe de-excitation. On the other hand, the upper limit from the galactic axions is on the order of $10^{-13}$ in the mass range from 1 keV/$c^2$ to 10 keV/$c^2$ with 90\% confidence level, slightly improved compared with the LUX.
• ### Dark Matter Results From 54-Ton-Day Exposure of PandaX-II Experiment(1708.06917)

We report a new search of weakly interacting massive particles (WIMPs) using the combined low background data sets in 2016 and 2017 from the PandaX-II experiment in China. The latest data set contains a new exposure of 77.1 live day, with the background reduced to a level of 0.8$\times10^{-3}$ evt/kg/day, improved by a factor of 2.5 in comparison to the previous run in 2016. No excess events were found above the expected background. With a total exposure of 5.4$\times10^4$ kg day, the most stringent upper limit on spin-independent WIMP-nucleon cross section was set for a WIMP with mass larger than 100 GeV/c$^2$, with the lowest exclusion at 8.6$\times10^{-47}$ cm$^2$ at 40 GeV/c$^2$.
• ### Spin-Dependent Weakly-Interacting-Massive-Particle--Nucleon Cross Section Limits from First Data of PandaX-II Experiment(1611.06553)

Feb. 15, 2017 hep-ph, hep-ex, astro-ph.CO
New constraints are presented on the spin-dependent WIMP-nucleon interaction from the PandaX-II experiment, using a data set corresponding to a total exposure of 3.3$\times10^4$ kg-days. Assuming a standard axial-vector spin-dependent WIMP interaction with $^{129}$Xe and $^{131}$Xe nuclei, the most stringent upper limits on WIMP-neutron cross sections for WIMPs with masses above 10 GeV/c$^{2}$ are set in all dark matter direct detection experiments. The minimum upper limit of $4.1\times 10^{-41}$ cm$^2$ at 90\% confidence level is obtained for a WIMP mass of 40 GeV/c$^{2}$. This represents more than a factor of two improvement on the best available limits at this and higher masses. These improved cross-section limits provide more stringent constraints on the effective WIMP-proton and WIMP-neutron couplings.
• ### Dark Matter Results from First 98.7-day Data of PandaX-II Experiment(1607.07400)

Aug. 23, 2016 hep-ex, physics.ins-det
We report the WIMP dark matter search results using the first physics-run data of the PandaX-II 500 kg liquid xenon dual-phase time-projection chamber, operating at the China JinPing Underground Laboratory. No dark matter candidate is identified above background. In combination with the data set during the commissioning run, with a total exposure of 3.3$\times10^4$ kg-day,the most stringent limit to the spin-independent interaction between the ordinary and WIMP dark matter is set for a range of dark matter mass between 3.5 and 1000 GeV/c$^2$. The best upper limit on the scattering cross section is found $2.5\times 10^{-46}$ cm$^2$ for the WIMP mass 40 GeV/c$^2$ at 90% confidence level.
• ### Dark Matter Search Results from the Commissioning Run of PandaX-II(1602.06563)

June 5, 2016 hep-ex, physics.ins-det
We present the results of a search for WIMPs from the commissioning run of the PandaX-II experiment located at the China Jinping underground Laboratory. A WIMP search data set with an exposure of 306$\times$19.1 kg-day was taken, while its dominant $^{85}$Kr background was used as the electron recoil calibration. No WIMP candidates are identified, and a 90\% upper limit is set on the spin-independent elastic WIMP-nucleon cross section with a lowest excluded cross section of 2.97$\times$10$^{-45}$~cm$^2$ at a WIMP mass of 44.7~GeV/c$^2$.
• ### A c=1 phase transition in two-dimensional CDT/Horava-Lifshitz gravity?(1412.3873)

Dec. 12, 2014 hep-th, gr-qc, hep-lat
We study matter with central charge $c >1$ coupled to two-dimensional (2d) quantum gravity, here represented as causal dynamical triangulations (CDT). 2d CDT is known to provide a regularization of (Euclidean) 2d Ho\v{r}ava-Lifshitz quantum gravity. The matter fields are massive Gaussian fields, where the mass is used to monitor the central charge $c$. Decreasing the mass we observe a higher order phase transition between an effective $c=0$ theory and a theory where $c>1$. In this sense the situation is somewhat similar to that observed for "standard" dynamical triangulations (DT) which provide a regularization of 2d quantum Liouville gravity. However, the geometric phase observed for $c >1$ in CDT is very different from the corresponding phase observed for DT.
• ### The spectral dimension in 2D CDT gravity coupled to scalar fields(1412.3434)

Dec. 10, 2014 hep-th, gr-qc, hep-lat
Causal Dynamical Triangulations (CDT) provide a non-perturbative formulation of Quantum Gravity assuming the existence of a global time foliation. In our earlier study we analyzed the effect of including $d$ copies of a massless scalar field in the two-dimensional CDT model with imaginary time. For $d > 1$ we observed the formation of a "blob", somewhat similar to that observed in four-dimensional CDT without matter. In the two-dimensional case the "blob" has a Hausdorff dimension $D_H=3$. In this paper we study the spectral dimension $D_S$ of the two-dimensional CDT-universe, both for $d = 0$ (pure gravity) and $d = 4$. We show that in both cases the spectral dimension is consistent with $D_S = 2$.