Search for decay Bs to Eta' Ks in Belle Data

I report the results of the first search for the decay Bs to Eta' Ks. The results are obtained using a 121.4 fb^−1 data sample that contains (16.60 ± 2.68) × 10^6 Bs mesons collected at the Υ(5S) resonance with the Belle detector at the KEKB asymmetric energy e+ e- collider. I use Monte Carlo simulation to optimize the selection criteria and use the package RooKSFW to maximize the discovery potential for the signal process. An unbinned maximum likelihood fit is performed to obtain the number of signal and background events. No signal is observed, and I report a 90% confidence level upper limit on the branching fraction of the decay Bs to Eta' Ks to be 8.16 × 10^−6 . The theoretical predictions for the branching fraction of this decay are in the range between 0.7 × 10^−6 and 4.5 × 10^−6 . I also report a 90% confidence level upper limit on the product of fs, bs bsbar hadronization factor in Υ(5S) decays, and the branching fraction, i.e. fs × B(Bs to eta' Ks), to be 1.64 × 10^−6.  

I also report my studies of a new neural network I developed using the machine learning library PyTorch to improve the sensitivity of rare decay searches at Belle. Compared with machine learning tools developed by the Belle collaboration, such as RooKSFW and NeuroBayes, my neural network rejects about 50% more background for the same signal efficiency, however, it exhibits a bias making backgrounds to partially resemble the signal.