Open-set classification is a problem of handling 'unknown' classes that are not contained in the training dataset, whereas traditional classifiers assume that only known classes appear in the test environment. Existing open-set classifiers rely on deep networks trained in a supervised manner on known classes in the training set; this causes specialization of learned representations to known classes and makes it hard to distinguish unknowns from knowns. In contrast, we train networks for joint classification and reconstruction of input data. This enhances the learned representation so as to preserve information useful for separating unknowns from knowns, as well as to discriminate classes of knowns. Our novel Classification-Reconstruction learning for Open-Set Recognition (CROSR) utilizes latent representations for reconstruction and enables robust unknown detection without harming the known-class classification accuracy. Extensive experiments reveal that the proposed method outperforms existing deep open-set classifiers in multiple standard datasets and is robust to diverse outliers. The code is available in https://nae-lab.

No abstract

BackgroundCD3 +  and CD8 + T-cell infiltration were reported as positive predictive markers of survival in colorectal cancer (CRC) patients. Here, we demonstrate the prognostic significance of CD4 + and FOXP3 + T-cell densities in CRC.MethodsWe quantified the intratumoural densities of CD3 + , CD8 + , CD4 +  and FOXP3 + T cells by immunohistochemistry and digital pathology in 342 CRC patients who underwent curative resection. Microsatellite instability was also assessed in 322 specimens. Patient demographics, clinicopathological features and survival rates were analysed.ResultsHigh CD3 + , CD4 +  and FOXP3 + T-cell densities were associated with improved relapse-free survival (RFS); high CD8 + , CD4 +  and FOXP3 + T-cell densities were associated with improved disease-specific survival (DSS). Patients with low CD4 + and low FOXP3 + T-cell densities exhibited extremely poor prognoses. T stage, vascular/lymphatic invasion and CD4 + T-cell density were independent prognostic indicators for DSS. The distributions of CD4 + and FOXP3 + T-cell densities were not significantly different between the high microsatellite instability group and other groups, in contrast to those of CD3 + and CD8 + T-cell densities.ConclusionsIntratumoural CD4 + T-cell density and combined CD4 + and FOXP3 + T-cell densities were stronger prognostic indicators than other clinicopathological features. These results may facilitate the establishment of novel prognostic factors and therapeutic strategies for CRC.

The compression wave generated by a high-speed train entering a tunnel is studied theoretically and experimentally. It is shown that the pressure rise across the wavefront is given approximately byformula herewhere ρo, U, M, [Ascr ]o and [Ascr ] respectively denote the mean air density, train speed, train Mach number, and the cross-sectional areas of the train and the uniform section of the tunnel. A monopole source representing the displacement of air by the train is responsible for the main pressure rise across the wave, but second-order dipole sources must also be invoked to render theoretical predictions compatible with experiment. The principal dipole is produced by the compression wave drag acting on the nose of the train. A second dipole of comparable strength, but probably less significant in practice, is attributed to ‘vortex sound’ sources in the shear layers of the back-flow out of the tunnel of the air displaced by the train.Experiments are performed that confirm the efficacy of an ‘optimally flared’ portal whose cross-sectional area S(x) varies according to the formulaformula herewhere x is distance increasing negatively into the tunnel, [lscr ] is the prescribed length of the flared section, and [Ascr ]E is the tunnel entrance cross-sectional area, given byformula hereThis portal is predicted theoretically to cause the pressure to increase linearly with distance across a compression wavefront of thickness ∼ [lscr ]/M, which is very much larger than in the absence of flaring. The increased wave thickness and linear pressure variation counteract the effect of nonlinear steepening of the wave in a long tunnel, and tend to suppress the environmentally harmful ‘micro-pressure wave’ radiated from the far end of the tunnel when the compression wave arrives. Our experiments are conducted at model scale using axisymmetric ‘trains’ projected at U ∼ 300 k.p.h. (M ≈ 0.25) along the axis of a cylindrical tunnel fitted with a flared portal. The blockage [Ascr ]o/[Ascr ] = 0.2, which is comparable to the larger values encountered in high-speed rail operations.

No abstract