On the Role of Weight Sharing During Deep Option Learning

The options framework is a popular approach for building temporally extended actions in reinforcement learning. In particular, the option-critic architecture provides general purpose policy gradient theorems for learning actions from scratch that are extended in time. However, past work makes the key assumption that each of the components of option-critic has independent parameters. In this work we note that while this key assumption of the policy gradient theorems of option-critic holds in the tabular case, it is always violated in practice for the deep function approximation setting. We thus reconsider this assumption and consider more general extensions of option-critic and hierarchical option-critic training that optimize for the full architecture with each update. It turns out that not assuming parameter independence challenges a belief in prior work that training the policy over options can be disentangled from the dynamics of the underlying options. In fact, learning can be sped up by focusing the policy over options on states where options are actually likely to terminate. We put our new algorithms to the test in application to sample efficient learning of Atari games, and demonstrate significantly improved stability and faster convergence when learning long options. 1

Full . Full O(α) electroweak radiative corrections to e^+ e^-→ZH with beam polarizations at the ILCO(α) electroweak radiative corrections to e^+ e^-→ZH with beam polarizations at the ILC

We present full electroweak radiative corrections to with the initial beam polarizations at the International Linear Collider (ILC). The calculation is checked numerically by using three consistency tests that are ultraviolet finiteness, infrared finiteness, and gauge parameter independence. In phenomenological results, we study the impact of the electroweak corrections to total cross section as well as its distributions. In addition, we discuss the possibility of searching for an additional Higgs in arbitrary beyond the Standard Model (BSM) through ZH production at the ILC.

O pewnym założeniu twierdzenia Bella

Dowód twierdzenia Bella sprowadza się do wyprowadzenia którejś z nierówności Bella. W ich standardowych wyprowadzeniach jednak kluczową rolę odgrywa warunek faktoryzowalności łącznego prawdopodobieństwa warunkowego, który można uzyskać jako konsekwencję dwóch innych warunków, znanych jako parameter independence i outcome independence. Pierwszy z nich jest dość oczywistym wyrazem warunku lokalności, natomiast drugi budzi wątpliwości. Ponieważ jednak jest on uszczegółowieniem warunku screening off zasady wspólnej przyczyny, jego podważenie wymagałoby zakwestionowania również tego warunku. Gdyby się to powiodło, efektywny dowód twierdzenia Bella wymagałby wyprowadzenia nierówności Bella nie wykorzystującego żadnego uszczegółowienia warunku screening off. Przestawiona zostanie sugestia kierunku, w jakim powinny iść poszukiwania modelu naruszającego ten warunek.

THE HESS-PHILIPP MODEL IS NON-LOCAL

Hess and Philipp have constructed what, they claim, is a local hidden variables model reproducing the empirical predictions of quantum mechanics. In this paper explicit expressions for the conditional probabilities for the outcomes of the measurements at the two detectors are calculated. These expressions provide a conclusive demonstration of the falsity of the authors' claim. The authors give two different accounts of their model. The published version omits a crucial detail. As a result it disagrees with quantum mechanics. It also violates signal locality. The unpublished version agrees with quantum mechanics. However, it violates the condition of parameter independence, as Myrvold has previously shown.