Search

Existing models in cognitive science typically assume human categorization as graded generalization behavior in a multidimensional psychological space. However, category representations in these models may suffer from the curse of dimensionality in a natural setting. People generally rely on a tractable yet sufficient set of features to understand the complex environment. We propose a rational model of categorization based on a hierarchical mixture of probabilistic principal components, that simultaneously learn category representations and an economical collection of features. The model captures dimensional biases in human categorization and supports zero-shot learning. We further exploit a generative process within a low-dimensional latent space to provide a better account of categorization with high-dimensional stimuli. We validate the model with simulation and behavioral experiments.

We describe a simple quantum error correcting code built out of a time-dependent transverse field Ising model. The code is similar to a repetition code, but has two advantages: an $N$-qubit code can be implemented with a finite-depth spatially local unitary circuit, and it can subsequently protect against both $X$ and $Z$ errors if $N\ge 10$ is even. We propose an implementation of this code with 10 ultracold Rydberg atoms in optical tweezers, along with further generalizations of the code., Comment: 15 pages, 7 figures

Objective: To investigate the diagnostic efficiency of truncal-type occlusion and branching-site occlusion in determining the etiology of intracranial large artery occlusion related acute ischemic stroke (AIS). Methods: Patients with intracranial large artery occlusion related AIS who received stent retriever (SR) thrombectomy from November 2014 to June 2019 were included in the study. All patients underwent angiography before SR thrombectomy, which was used to evaluate the occlusion type. Differences in the distribution of occlusion types in intracranial atherosclerosis (ICAS) and embolism were assessed, and the diagnostic indicators, including the area under the ROC curve (AUC), sensitivity, and specificity were calculated. Results: Of the 115 AIS patients with intracranial large artery occlusion, 42 were classified as having ICAS, and 73 having an embolism. In the ICAS group, branching-site occlusion was responsible for 3 (7%) cases and truncal-type occlusion for 39 (93%) cases, while in the embolism group, branching-site occlusion was responsible for 66 (90%) cases and truncal-type occlusion for 7 (10%) cases; the difference was statistically significant (all P < 0.01). The AUC for ICAS predicted by truncal-type occlusion was 0.916, with a sensitivity of 92.86%, and specificity of 90.41%. Conclusion: Truncal-type occlusion showed a high predictability of ICAS. Determine the etiology of intracranial large artery occlusion related AIS before SR thrombectomy may be most helpful in setting up optimal endovascular treatment strategies.

In contrast with Phase-Locked Loop (PLL), Frequency Locked Loop (FLL) has better dynamic performance. There were papers dealing with noise performance of FLL. However, under the circumstance of CW input, there were large gap between simulation results and theory results which were obtained from published papers. This paper analysis the reason of the gap and proposes an improved calculation method in the condition of four-quadrant arctangent and cross product discriminator. Results show that the improved calculation method fits the simulation result very well.