
本文介紹了一種基於通用微分方程（UDE）的神經網絡，用於模擬宇宙再電離過程，並證明了其在構建快速、靈活且具有物理直觀性的模擬器方面的潛力。


coremsg

基於通用微分方程的神經網絡模擬宇宙再電離過程

title_rewrite


在耦合的可興奮單元網絡中，單元瞬態動力學與耦合之間的相互作用通過將單元捕獲在興奮性區域，產生了多重穩定性，並導致週期性、擬週期性甚至混沌振盪的共存。


瞬態與網絡交互作用如何通過捕獲機制產生多重穩定性


與傳統消退療法相比，反制約作用通過激活獎賞神經通路（而非恐懼消退通路），更持久地抑制了恐懼記憶的恢復，並增強了對制約刺激的記憶。


Counterconditioning (CC) aims to enhance extinction of threat memories by establishing new associations of opposite valence. While its underlying neurocognitive mechanisms remain largely unexplored, previous studies suggest qualitatively different mechanisms from regular extinction. In this functional MRI study, participants underwent categorical threat conditioning (CS+/CS-: images of animals/tools), followed by either CC (CS+ images reinforced with monetary rewards, n=24) or regular extinction (n=24). The following day, we assessed spontaneous recovery of threat responses and episodic memory for CS+ and CS- category exemplars. While the ventromedial prefrontal cortex (vmPFC) was activated during regular extinction, participants undergoing CC showed persistent CS+- specific deactivation of the vmPFC and hippocampus, and CS+-specific activation of the nucleus accumbens (NAcc). The following day, physiological threat responses returned in the regular extinction group, but not in the CC group. Counterconditioning furthermore strengthened episodic memory for CS+ exemplars presented during CC, and retroactively also for CS+ exemplars presented during the threat conditioning phase. Our findings confirm that CC leads to more persistent extinction of threat memories, as well as altered consolidation of the threat conditioning episode. Crucially, we show a qualitatively different activation pattern during CC versus regular extinction, with a shift away from the vmPFC and towards the NAcc.

### Competing Interest Statement

The authors have declared no competing interest.

Unravelling the neurocognitive mechanisms underlying counterconditioning in humans

揭示人類反制約作用背後的神經認知機制


本文探討了卷積神經網絡如何僅憑單詞語音數據，就能自發地學習串聯單詞，展現出語法發展的雛形，並提出「人工去抑制」作為可能的解釋機制。


從語音中學習基本語法-無監督深度神經網絡中的自發串聯現象


從語音中學習基本語法：無監督深度神經網絡中的自發串聯現象



基於腦電圖能量特徵和雙向長短期記憶神經網絡，可以有效區分音樂與語音，並進一步辨識音樂類型和聽者喜好。


基於能量特徵和雙向長短期記憶神經網絡的腦電圖音樂與語音分類


小鼠聽覺皮層中的神經元不僅對聲音有反應，還對視覺刺激和運動有反應，並且這些反應會被預測誤差所影響，特別是當聽覺和視覺預測誤差同時發生時，反應會被放大。


Our movements result in predictable sensory feedback that is often multimodal. Based on deviations between predictions and actual sensory input, primary sensory areas of cortex have been shown to compute sensorimotor prediction errors. How prediction errors in one sensory modality influence the computation of prediction errors in another modality is still unclear. To investigate multimodal prediction errors in mouse auditory cortex (ACx), we used a virtual environment to experimentally couple running to both self-generated auditory and visual feedback. Using two-photon microscopy, we first characterized responses of layer 2/3 (L2/3) neurons to sounds, visual stimuli, and running onsets and found responses to all three stimuli. Probing responses evoked by audiomotor mismatches, we found that they closely resemble visuomotor mismatch responses in visual cortex (V1). Finally, testing for cross modal influence on audiomotor mismatch responses by coupling both sound amplitude and visual flow speed to the speed of running, we found that audiomotor mismatch responses were amplified when paired with concurrent visuomotor mismatches. Our results demonstrate that multimodal and non-hierarchical interactions shape prediction error responses in cortical L2/3.

### Competing Interest Statement

The authors have declared no competing interest.

Multimodal mismatch responses in mouse auditory cortex

小鼠聽覺皮層中的多模態失配反應


本文綜述了基於變換器網絡的重味噴流標記方法的最新進展，探討了不同數據結構的優缺點，強調了將物理結構融入網絡設計的重要性，並介紹了幾種解釋網絡決策過程的方法。


基於變換器網絡的重味噴流標記方法綜述


在多尺度信息處理系統中，非線性積分比非線性加總更能有效地增強輸入與輸出之間的互信息，從而實現更精確的信息編碼和輸入判別。


多尺度非線性積分如何驅動輸入信息的精確編碼


本文提出了一種名為 MaGNet 的新型圖神經網絡框架，旨在通過整合不同階鄰居節點信息來提高模型的表示能力和可解釋性，並通過模擬和實際應用案例驗證了其有效性。


一種用於整合局部和全局信息的模型無關圖神經網絡


本文提出了一種名為耦合積分物理信息神經網絡 (CI-PINN) 的新型深度學習方法，用於解決非線性守恆律中存在衝擊波的問題，並通過與傳統數值方法和普通 PINN 的比較，證明了其在求解此類問題上的優越性和準確性。


耦合積分物理信息神經網絡求解守恆律