The Earth mover's distance (EMD) is a useful metric for image recognition and classification, but its usual implementations are not differentiable or too slow to be used as a loss function for training other algorithms via gradient descent. In this paper, we train a convolutional neural network (CNN) to learn a differentiable, fast approximation of the EMD and demonstrate that it can be used as a substitute for computing-intensive EMD implementations. We apply this differentiable approximation in the training of an autoencoder-inspired neural network (encoder NN) for data compression at the high-luminosity LHC at CERN. The goal of this encoder NN is to compress the data while preserving the information related to the distribution of energy deposits in particle detectors. We demonstrate that the performance of our encoder NN trained using the differentiable EMD CNN surpasses that of training with loss functions based on mean squared error.

Precision measurements of the semileptonic decays $D_s^+ \to \eta e^+ \nu_e$ and $D_s^+ \to \eta^\prime e^+ \nu_e$ are performed using 7.33\,fb$^{-1}$ of $e^+e^-$ collision data collected at center-of-mass energies between 4.128 and 4.226 GeV with the BESIII detector. The branching fractions obtained are $\mathcal{B}(D_s^+ \to \eta e^{+} \nu_e)$ = $(2.251\pm0.039_{\rm stat.}\pm 0.051_{\rm syst.})\%$ and $\mathcal{B}(D_s^+ \to \eta^{\prime} e^{+} \nu_e)$ = $(0.810\pm0.038_{\rm stat.}\pm 0.024_{\rm syst.})\%$. Combining these results with the $\mathcal{B}(D^+\to\eta e^+ \nu_e)$ and $\mathcal{B}(D^+\to\eta^\prime e^+ \nu_e)$ obtained from previous BESIII measurements, the $\eta-\eta^\prime$ mixing angle in the quark flavor basis is determined to be $\phi_{\rm P} = (40.0\pm2.0_{\rm stat.}\pm0.6_{\rm syst.})^\circ$. Moreover, from the fits to the partial decay rates of $D_s^+ \to \eta e^+ \nu_e$ and $D_s^+ \to \eta^\prime e^+ \nu_e$, the products of the hadronic transition form factors $f_+^{\eta^{(\prime)}}(0)$ and the modulus of the $c\to s$ Cabibbo-Kobayashi-Maskawa matrix element $|V_{cs}|$ are determined by using different hadronic transition form factor parametrizations. Based on the two-parameter series expansion, the products $f^\eta_+(0)|V_{cs}| = 0.4553\pm0.0071_{\rm stat}\pm0.0061_{\rm syst}$ and $f^{\eta^\prime}_+(0)|V_{cs}| = 0.529\pm0.024_{\rm stat}\pm0.008_{\rm syst}$ are extracted. All results determined in this work supersede those measured in the previous BESIII analyses based on the 3.19 fb$^{-1}$ subsample of data at 4.178 GeV.

We describe two different approaches for incorporating systematics into analyses for parameter determination in the physical sciences. We refer to these as the Pragmatic and the Full methods, with the latter coming in two variants: Full Likelihood and Fully Bayesian. By the use of a simple and readily understood example, we point out the advantage of using the Full Likelihood and Fully Bayesian approaches; a more realistic example from Astrophysics is also presented. This could be relevant for data analyses in a wide range of scientific fields, for situations where systematic effects need to be incorporated in the analysis procedure. This note is an extension of part of the talk by van Dyk at the PHYSTAT-Systematics meeting.

We propose novel methods to determine the $\Upsilon(4S)\to B^+B^-$ and $\Upsilon(4S)\to B^0\bar B^0$ decay rates. The precision to which they and their ratio are known yields at present a limiting uncertainty around $2\%$ in measurements of absolute $B$ decay rates, and thus in a variety of applications, such as precision determinations of elements of the Cabibbo-Kobayashi-Maskawa matrix and flavor symmetry relations. The new methods we propose are based in one case on exploiting the $\Upsilon(5S)$ data sets, in the other case on the different average number of charged tracks in $B^\pm$ and $B^0$ decays. We estimate future sensitivities using these methods and discuss possible measurements of $f_d / f_u$ at the (HL-)LHC.

We include strong parity-violating contributions to inclusive deep inelastic scattering (DIS) of longitudinally polarized leptons off an unpolarized target. At variance with standard results, we obtain nonvanishing parity-violating structure functions in the case of pure photon exchange. The addition of these strong parity-violating contributions improves the description of existing experimental data on DIS parity violating asymmetries. Their size is small but incompatible with zero at about 1.5-$\sigma$ level.