Search

This paper proposes a novel self-supervised learning method for learning better representations with small batch sizes. Many self-supervised learning methods based on certain forms of the siamese network have emerged and received significant attention. However, these methods need to use large batch sizes to learn good representations and require heavy computational resources. We present a new triplet network combined with a triple-view loss to improve the performance of self-supervised representation learning with small batch sizes. Experimental results show that our method can drastically outperform state-of-the-art self-supervised learning methods on several datasets in small-batch cases. Our method provides a feasible solution for self-supervised learning with real-world high-resolution images that uses small batch sizes., Comment: Published as a conference paper at ICASSP 2022

The global outbreak of the Coronavirus 2019 (COVID-19) has overloaded worldwide healthcare systems. Computer-aided diagnosis for COVID-19 fast detection and patient triage is becoming critical. This paper proposes a novel self-knowledge distillation based self-supervised learning method for COVID-19 detection from chest X-ray images. Our method can use self-knowledge of images based on similarities of their visual features for self-supervised learning. Experimental results show that our method achieved an HM score of 0.988, an AUC of 0.999, and an accuracy of 0.957 on the largest open COVID-19 chest X-ray dataset., Comment: Published as a conference paper at ICASSP 2022

We evaluate the zero-point renormalization (ZPR) due to electron-phonon interactions of 28 solids using the projector-augmented-wave (PAW) method. The calculations cover diamond, many zincblende semiconductors, rock-salt and wurtzite oxides, as well as silicate and titania. Particular care is taken to include long-range electrostatic interactions via a generalized Fr\"ohlich model, as discussed in Phys. Rev. Lett. 115, 176401 (2015) and Phys. Rev. B 92, 054307 (2015). The data are compared to recent calculations, npj Computational Materials 6, 167 (2020), and generally very good agreement is found. We discuss in detail the evaluation of the electron-phonon matrix elements within the PAW method. We show that two distinct versions can be obtained depending on when the atomic derivatives are taken. If the PAW transformation is applied before taking derivatives with respect to the ionic positions, equations similar to the ones conventionally used in pseudopotential codes are obtained. If the PAW transformation is used after taking the derivatives, the full-potential spirit is largely maintained. We show that both variants yield very similar ZPRs for selected materials when the rigid-ion approximation is employed. In practice, we find however that the pseudo version converges more rapidly with respect to the number of included unoccupied states., Comment: Added pseudopotential information; revised Sec. II B to avoid possible misconceptions; revised discussion in Sec. III A; added diamond calculation with harder pseudopotential; corrected ZPR comparisons for AlSb and GaP due to inconsistent band-gap transitions; conclusion unchanged

Longitudinal-optical (LO) mode phonon branches of KCl and NaCl were measured using inelastic X-ray scattering (IXS) at 300 K and calculated by the first-principles phonon calculation with the stochastic self-consistent harmonic approximation. Spectral shapes of the IXS measurements and calculated spectral functions agreed well. We analyzed the calculated spectral functions that provide higher resolutions of the spectra than the IXS measurements. Due to strong anharmonicity, the spectral functions of these phonon branches have several peaks and the LO modes along $\Gamma$--L paths are disconnected.