Jiacong Li | Searcharxiv

arXiv2026

CAAL: Contextual Bandits based Online Hand-Craft Active Learning Strategy Selection

The challenge with active learning algorithms is the uncertainty of the statistical distribution of unlabeled data, making it difficult to choose the best hand-crafted strategy. To address this, we introduced Contextual Adaptive Active Learning (CAAL). In CAAL, each "arm" represents a hand-crafted strategy. Unlike existing frameworks that select strategies based only on feedback from labeled data, we dynamically choose strategies for labeling batches of data using reward prediction with external context information. This general framework allows for customization with domain knowledge to design more effective rewards and context candidates. In addition, we experimentally show that CAAL outperforms the existing baseline adaptive strategy on public datasets using our reward and context design. Our results are consistent regardless of batch size in each iteration.

cs.LG↗

arXiv2018

Frictionless Zonal Flow Saturation by Vorticity Mixing

Consideration of wave--flow resonance addresses the long-standing problem of how zonal flows (ZF) saturate in the limit of weak or zero frictional drag, and also determines the ZF scale. For relevant magnetic geometries, the frequently quoted tertiary instability requires unphysical enhancement of ZF shear and thus is irrelevant to the near-marginal, frictionless regime. We show that resonant vorticity mixing, which conserves potential enstrophy, enables ZF saturation in the absence of drag, and so is effective in the Dimits up-shift regime. Vorticity mixing is incorporated as a nonlinear, self-regulation effect in an extended 0D predator--prey model of drift--ZF turbulence. This analysis determines the saturated ZF shear and shows that the mesoscopic ZF width scales as $L_{ZF}\sim f^{3/16} (1-f)^{1/8} ρ_s^{5/8} l_0^{3/8}$ in the relevant adiabatic limit (i.e., $τ_{ck} k_\|^2 D_\| \gg 1$). $f$ is the fraction of turbulence energy coupled to ZF and $l_0$ is the mixing length absent ZF shears. We calculate and compare the stationary flow and turbulence level in frictionless, weakly frictional, and strongly frictional regimes. In the frictionless limit, the results differ significantly from conventionally quoted scalings derived for frictional regimes. The flow is independent of turbulence intensity. The turbulence level scales as $E \sim (γ_L/\varepsilon_c)^2$, which defines the extent of the "near-marginal" regime to be $γ_L < \varepsilon_c$, for the case of avalanche-induced profile variability. Here, $\varepsilon_c$ is the rate of dissipation of potential enstrophy and $γ_L$ is the characteristic linear growth rate of fluctuations. The implications for dynamics near marginality of the strong scaling of saturated $E$ with $γ_L$ are discussed.

physics.plasm-ph↗