Search for: All records

Abstract Reinforcement learning (RL) relies on Gaussian and sigmoid functions to balance exploration and exploitation, but implementing these functions in hardware typically requires iterative computations, increasing power and circuit complexity. Here, Gaussian‐sigmoid reinforcement transistors (GS‐RTs) are reported that integrate both activation functions into a single device. The transistors feature a vertical n‐p‐i‐p heterojunction stack composed of a‐IGZO and DNTT, with asymmetric source–drain contacts and a parylene interlayer that enables voltage‐tunable transitions between sigmoid, Gaussian, and mixed responses. This architecture emulates the behavior of three transistors in one, reducing the required circuit complexity from dozens of transistors to fewer than a few. The GS‐RT exhibits a peak current of 5.95 µA at V_G= −17 V and supports nonlinear transfer characteristics suited for neuromorphic computing. In a multi‐armed bandit task, GS‐RT‐based RL policies demonstrate 20% faster convergence and 30% higher final reward compared to conventional sigmoid‐ or Gaussian‐based approaches. Extending this advantage further, GS‐RT‐based activation function in deep RL for cartpole balancing significantly outperforms the traditional ReLU‐based activation function in terms of faster learning and tolerance to input perturbations. 

Animals can adapt to dynamic environmental conditions by modulating their developmental programs. Understanding the genetic architecture and molecular mechanisms underlying developmental plasticity in response to changing environments is an important and emerging area of research. Here, we show a novel role of cAMP response element binding protein (CREB)-encoding crh-1 gene in developmental polyphenism of C . elegans . Under conditions that promote normal development in wild-type animals, crh-1 mutants inappropriately form transient pre-dauer (L2d) larvae and express the L2d marker gene. L2d formation in crh-1 mutants is specifically induced by the ascaroside pheromone ascr#5 (asc-ωC3; C3), and crh-1 functions autonomously in the ascr#5-sensing ASI neurons to inhibit L2d formation. Moreover, we find that CRH-1 directly binds upstream of the daf-7 TGF-β locus and promotes its expression in the ASI neurons. Taken together, these results provide new insight into how animals alter their developmental programs in response to environmental changes. 

Abstract Widespread ice shelf thinning has been recorded in the Amundsen Sea in recent decades, driven by basal melting and intrusions of relatively warm Circumpolar Deep Water (CDW) onto the continental shelf. The Dotson Ice Shelf (DIS) is located to the south of the Amundsen Sea polynya, and has a high basal melting rate because modified CDW (mCDW) fills the Dotson‐Getz Trough (DGT) and reaches the base of the ice shelf. Here, hydrographic data in the DGT obtained during seven oceanographic surveys from 2007 to 2018 were used to study the interannual variation in mCDW volume and properties and their causes. Although mCDW volume showed relatively weak interannual variations at the continental shelf break, these variations intensified southward and reached a maximum in front of the DIS. There, the mCDW volume was ∼8,000 km³in 2007, rapidly decreased to 4,700 km³in 2014 before rebounding to 7,300 km³in 2018. We find that such interannual variability is coherent with local Ekman pumping integrated along the DGT modulated by the presence of sea ice, and complementing earlier theories involving shelf break winds only. The interannual variability in strength of the dominant south‐southeast coastal wind modulates the amplitude of Ekman upwelling along the eastern boundary of the Amundsen Sea polynya during the austral summers of the surveyed years, apparently leading to change in the volume of mCDW along the DGT. We note a strong correlation between the wind variability and the longitudinal location of the Amundsen Sea Low.