Training deep learning networks involves continuous weight updates across the various layers of the deep network while using a backpropagation algorithm (BP). This results in expensive computation overheads during training. Consequently, most deep learning accelerators today employ pre-trained weights and focus only on improving the design of the inference phase. The recent trend is to build a complete deep learning accelerator by incorporating the training module. Such efforts require an ultra-fast chip architecture for executing the BP algorithm. In this article, we propose a novel photonics-based backpropagation accelerator for high performance deep learning training. We present the design for a convolutional neural network, BPLight-CNN, which incorporates the silicon photonics-based backpropagation accelerator. BPLight-CNN is a first-of-its-kind photonic and memristor-based CNN architecture for end-to-end training and prediction. We evaluate BPLight-CNN using a photonic CAD framework (IPKISS) on deep learning benchmark models including LeNet and VGG-Net. The proposed design achieves (i) at least 34x speedup, 34x improvement in computational efficiency, and 38.5x energy savings, during training; and (ii) 29x speedup, 31x improvement in computational efficiency, and 38.7x improvement in energy savings, during inference compared to the state-of-the-art designs. All these comparisons are done at a 16-bit resolution; and BPLight-CNN achieves these improvements at a cost of approximately 6% lower accuracy compared to the state-of-the-art.

We present measurements of the branching fraction and time-dependent CP asymmetries in B0→J/ψπ0 decays based on (231.8±2.6) ×106Υ(4S)→BB̄ decays collected with the BABAR detector at the SLAC PEP-II asymmetric-energy B factory. We obtain a branching fraction B(B0→J/ψπ0)=(1.94±0.22(stat)±0.17(syst))×10-5. We also measure the CP asymmetry parameters C=-0.21±0.26(stat)±0. 06(syst) and S=-0.68±0.30(stat)±0.04(syst). © 2006 The American Physical Society.

We present a measurement of the time-dependent CP asymmetry for the neutral B-meson decay B0-->phiK0. We use a sample of approximately 114 x 10(6) B-meson pairs taken at the Upsilon(4S) resonance with the BABAR detector at the PEP-II B-meson factory at SLAC. We reconstruct the CP eigenstates phiK0S and phiK0L, where phi-->K+K-, K0S-->pi+pi-, and K0L is observed via its hadronic interactions. The other B meson in the event is tagged as either a B0 or Bbar0 from its decay products. The values of the CP-violation parameters are SphiK=0.47+/-0.34(stat)+0.08-0.06(syst) and CphiK=0.01+/-0.33(stat)+/-0.10(syst).

The Belle II detector will provide a major step forward in precision heavy flavor physics, quarkonium and exotic states, searches for dark sectors, and many other areas. The sensitivity to a large number of key observables can be improved by about an order of magnitude compared to the current measurements, and up to two orders in very clean search measurements. This increase in statistical precision arises not only due to the increased luminosity, but also from improved detector efficiency and precision for many channels. Many of the most interesting observables tend to have very small theoretical uncertainties that will therefore not limit the physics reach. This book has presented many new ideas for measurements, both to elucidate the nature of current anomalies seen in flavor, and to search for new phenomena in a plethora of observables that will become accessible with the Belle II dataset. The simulation used for the studiesinthis book was state ofthe artat the time, though weare learning a lot more about the experiment during the commissioning period. The detector is in operation, and working spectacularly well.