The combination of Monte Carlo tree search and neural networks has revolutionized online planning. As neural network approximations are often imperfect, we ask whether uncertainty estimates about the network outputs could be used to improve planning. We develop a Bayesian planning approach that facilitates such uncertainty quantification, inspired by classical ideas from the meta-reasoning literature. We propose a Thompson sampling based algorithm for searching the tree of possible actions, for which we prove the first (to our knowledge) finite time Bayesian regret bound, and propose an efficient implementation for a restricted family of posterior distributions. In addition we propose a variant of the Bayes-UCB method applied to trees. Empirically, we demonstrate that on the ProcGen Maze and Leaper environments, when the uncertainty estimates are accurate but the neural network output is inaccurate, our Bayesian approach searches the tree much more effectively. In addition, we investigate whether popular uncertainty estimation methods are accurate enough to yield significant gains in planning.

The synchronization phase of a code division multiple access communication system might pose a limitation on the performance of the system, since unlike other communication systems it is required to obtain the state (i.e., sequence phase) of a linear feedback shift register (LFSR). Usually the state is obtained by using a brute force exhaustive search over all possibilities, which might be a computationally demanding task. In this paper, we suggest a new method for LFSR sequence phase acquisition. Our method is suitable for practical systems in which an unknown frequency drift may be present. Simulation results for the IS-95/CDMA-2000 standard show a dramatic reduction in the time required to acquire synchronization when using the new algorithm compared to the standard synchronization method.

The bit error probability (BER) and the generalized cutoff rate (GCR) of optically amplified, direct-detection on-off keying (OOK) signaling are addressed utilizing the classical theory of Fredholm integral equations. The approach accounts for the impact of optical bandlimiting and upper bounds on the BER are found via Chernoff techniques, which address the resultant intersymbol interference (ISI) incurred by the linear pulse distortion. Simple rigorous upper bounds on the BER are derived in terms of basic system parameters relaxing the need for solving Fredholm related equations. Random coding is used to obtain the GCR of this channel. The theory is demonstrated for the Lorentzian and finite-duration bandpass optical filters and the optical bandwidth which optimally balances the ISI versus the amplified spontaneous emission noise is found.>