arXiv subjects

Keita Sakuma

Publications and source records attributed to Keita Sakuma.

5 recordsLinked to original sources

arXiv2026

Agile Online Model Selection: Resolving Adaptation Lag via Safeguarded Large Learning Rates

Maintaining predictive accuracy in non-stationary environments requires online model selection to adapt autonomously to unknown distribution shifts. However, existing tuning-free algorithms face a fundamental trade-off between robustness and agility. Specifically, to ensure dynamic regret bounds, they must restrict learning rates to small constants (e.g., $O(1)$). This restriction inevitably causes significant adaptation lag during abrupt changes. To resolve this, we propose a novel optimistic online mirror descent that utilizes safeguarded large learning rates up to $Θ(T)$, where $T$ is the number of rounds. Our key technical contribution is a post-hoc penalty mechanism that dynamically monitors unstable updates and excludes learning rates incurring excessive regret, eliminating the need for restrictive a priori constraints. We show that the cumulative penalty remains $O(\log T)$, allowing our algorithm to match near-optimal worst-case guarantees while achieving superior rates in benign cases. Empirical evaluations on three synthetic and eleven diverse real-world datasets demonstrate that our approach reduces the adaptation lag from hundreds of rounds to a few rounds, consistently outperforming tuning-free baselines.

cs.LG↗

arXiv2025

Improved Impossible Tuning and Lipschitz-Adaptive Universal Online Learning with Gradient Variations

A central goal in online learning is to achieve adaptivity to unknown problem characteristics, such as environmental changes captured by gradient variation (GV), function curvature (universal online learning, UOL), and gradient scales (Lipschitz adaptivity, LA). Simultaneously achieving these with optimal performance is a major challenge, partly due to limitations in algorithms for prediction with expert advice. These algorithms often serve as meta-algorithms in online ensemble frameworks, and their sub-optimality hinders overall UOL performance. Specifically, existing algorithms addressing the ``impossible tuning'' issue incur an excess $\sqrt{\log T}$ factor in their regret bound compared to the lower bound. To solve this problem, we propose a novel optimistic online mirror descent algorithm with an auxiliary initial round using large learning rates. This design enables a refined analysis where a generated negative term cancels the gap-related factor, resolving the impossible tuning issue up to $\log\log T$ factors. Leveraging our improved algorithm as a meta-algorithm, we develop the first UOL algorithm that simultaneously achieves state-of-the-art GV bounds and LA under standard assumptions. Our UOL result overcomes key limitations of prior works, notably resolving the conflict between LA mechanisms and regret analysis for GV bounds -- an open problem highlighted by Xie et al.

cs.LG↗

arXiv2022

Thermodynamic approach for enhancing superconducting critical current performance

The addition of artificial pinning centers has led to an impressive increase in critical current density ($J_{\rm c}$) in a superconductor, enabling record-breaking all-superconducting magnets and other applications. $J_{\rm c}$ has reached $\sim 0.2$-$0.3$ $J_{\rm d}$, where $J_{\rm d}$ is the depairing current density, and the numerical factor depends on the pinning optimization. By modifying $λ$ and/or $ξ$, the penetration depth and coherence length, respectively, we can increase $J_{\rm d}$. For (Y$_{0.77}$Gd$_{0.23}$)Ba$_2$Cu$_3$O$_y$ ((Y,Gd)123) we achieve this by controlling the carrier density, which is related to $λ$ and $ξ$. We also tune $λ$ and $ξ$ by controlling the chemical pressure in the Fe-based superconductors, BaFe$_2$(As$_{1-x}$P$_x$)$_2$ films. The variation of $λ$ and $ξ$ leads to an intrinsic improvement of $J_{\rm c}$, via $J_{\rm d}$, obtaining extremely high values of $J_{\rm c}$ of $130$ MA/cm$^2$ and $8.0$ MA/cm$^2$ at $4.2$ K, consistent with an enhancement of $J_{\rm d}$ of a factor of $2$ for both incoherent nanoparticle-doped (Y,Gd)123 coated conductors (CCs) and BaFe$_2$(As$_{1-x}$P$_x$)$_2$ films, showing that this new material design is useful to achieving high critical current densities for a wide array of superconductors. The remarkably high vortex-pinning force in combination with this thermodynamic and pinning optimization route for the (Y,Gd)123 CCs reached $\sim 3.17$ TN/m$^3$ at $4.2$ K and 18 T (${\bf H}\parallel c$), the highest values ever reported in any superconductor.

cond-mat.supr-con↗

arXiv2021

Competition and interplay between topology and quasi-periodic disorder in Thouless pumping of ultracold atoms

Robustness against perturbations lies at the heart of topological phenomena. If, however, a perturbation such as disorder becomes dominant, it may cause a topological phase transition between topologically non-trivial and trivial phases. Here we experimentally reveal the competition and interplay between topology and quasi-periodic disorder in a Thouless pump realized with ultracold atoms in an optical lattice, by creating a quasi-periodic potential from weak to strong regimes in a controllable manner. We demonstrate a disorder-induced pumping in which the presence of quasi-periodic disorder can induce a non-trivial pump for a specific pumping sequence, while no pump is observed in the clean limit. Our highly controllable system, which can also straightforwardly incorporate interatomic interaction, could be a unique platform for studying various disorder-related novel effects in a wide range of topological quantum phenomena.

cond-mat.quant-gas↗

arXiv2020

A superconducting praseodymium nickelate with infinite layer structure

A variety of nickel oxide compounds have long been studied for their manifestation of various correlated electron phenomena. Recently, superconductivity was observed in nanoscale infinite layer nickelate thin films of Nd$_{0.8}$Sr$_{0.2}$NiO$_2$, epitaxially stabilized on SrTiO$_3$ substrates via topotactic reduction from the perovskite precursor phase. Here we present the synthesis and properties of PrNiO$_2$ thin films on SrTiO$_3$. Upon doping in Pr$_{0.8}$Sr$_{0.2}$NiO$_2$, we observe superconductivity with a transition temperature of 7-12 K, and robust critical current density at 2 K of 334 kA/cm$^2$. These findings indicate that superconductivity in the infinite layer nickelates is relatively insensitive to the details of the rare earth 4$f$ configuration. Furthermore, they motivate the exploration of a broader family of compounds based on two-dimensional NiO$_2$ planes, which will enable systematic investigation of the superconducting and normal state properties and their underlying mechanisms.

cond-mat.supr-con↗