A deep reinforcement learning (DRL) method is proposed for hybrid beamforming in near-field wideband communication systems, aiming to alleviate the harmful near-field beam splitting issue caused by the spatial-wideband effect. Compared to the far-field beam splitting issue that occurs only in the angle domain, the near-field beam splitting issue extends to the distance domain and is thus more challenging to address. A hybrid beamforming architecture with true-time delayers (TTDs) is exploited to address the near-field beam splitting issue and facilitate the maximization of spectral efficiency (SE) across large bandwidth for downlink multi-user communications. However, the conventional iterative hybrid beamforming optimization methods, such as the weighted minimum mean-square method, often exhibit high computational complexity and are thus difficult to adapt the dynamic channel in a real-time manner. Hence, a DRL-based algorithm is propose to select fully-digital codewords from the designed near-field wideband beamforming codebook. To further approximate the selected codewords in the considered hybrid-beamforming architecture, a three-stage hybrid beamforming approximation (HBA) algorithm is proposed to minimize the fully-digital approximation error by jointly optimizing the TTD-based analog beamformers and baseband digital beamformers. Numerical results show that: (1) the proposed DRL-HBA solution for hybrid beamforming significantly increases the SE; (2) the hybrid beamforming architecture with TTDs can eliminate the near-field beam splitting effect and achieve beam focusing across wide band.

Nowadays, detecting sentiment or emotion from user generated texts has been intensively studied in natural language understanding, especially via neural-based models based on text representation. However, the interpretability on how could the final text sentiment be determined by neural-based text representation has not been thoroughly unfolded yet. Consequently, in this paper, we proposeCogLignwhich injects theneural-cognitionderived from Electroencephalogram (EEG)-signal into theneural-basedtext sentiment analysis model, aimed at learning the activation of brain regions stimulated by different sentiments, so as to guide our proposedCogLignto make proper determination on text sentiment in brain-like way. Specifically, on the one hand, the given videos in different sentiments have been watched bysubjects, during which the EEG-signals are monitored to construct brain connectivity pattern asbrain graph(BG), attaining more obvious sentiment response on brain region activation forneural-cognition. On the other hand, we interpret the video-plots (or video-semantics) along timeline into text, where the entire video-interpreted-text will bestrictly boundwith the wholeEEG-signal-sequencebysegmentvia the fixed size oftime-window. Then, entities and relations are extracted from the video-interpreted-text to constructknowledge graph(KG), depicting text semantics. Next, mapping fromentities(or nodes) inKGtoEEG-Electrodes(or nodes) inBG, further dated back to different brain regions, has been learned viacognition alignmentbetween the EEG-derivedBGand text-derivedKG. In this way, by aligningneural cognitionfrombrain graphwith thesemantic cognitionfromknowledge graph, our proposed frameworkCogLigncan not only achieve the overall best sentiment analysis performance on thevideo-interpreted-text, but can also detect brain connectivity patterns in different sentiments more consistent with the prior conclusion of brain region sentiment preference, revealing competitiveinterpretabilityon text sentiment determination.