Search for: All records

Reservoir Computing (RC) models, a subclass of recurrent neural networks, are distinguished by their fixed, non-trainable input layer and dynamically coupled reservoir, with only the static readout layer being trained. This design circumvents the issues associated with backpropagating error signals through time, thereby enhancing both stability and training efficiency. RC models have been successfully applied across a broad range of application domains. Crucially, they have been demonstrated to be universal approximators of time-invariant dynamic filters with fading memory, under various settings of approximation norms and input driving sources.Simple Cycle Reservoirs (SCR) represent a specialized class of RC models with a highly constrained reservoir architecture, characterized by uniform ring connectivity and binary input-to-reservoir weights with an aperiodic sign pattern. For linear reservoirs, given the reservoir size, the reservoir construction has only one degree of freedom -- the reservoir cycle weight. Such architectures are particularly amenable to hardware implementations without significant performance degradation in many practical tasks. In this study we endow these observations with solid theoretical foundations by proving that SCRs operating in real domain are universal approximators of time-invariant dynamic filters with fading memory. Our results supplement recent research showing that SCRs in the complex domain can approximate, to arbitrary precision, any unrestricted linear reservoir with a non-linear readout. We furthermore introduce a novel method to drastically reduce the number of SCR units, making such highly constrained architectures natural candidates for low-complexity hardware implementations. Our findings are supported by empirical studies on real-world time series datasets. 

Abstract Topological spin textures (e.g., skyrmions) can be stabilized by interfacial Dzyaloshinskii‐Moriya interaction (DMI) in the magnetic multilayer, which has been intensively studied. Recently, Bloch‐type magnetic skyrmions stabilized by composition gradient‐induced DMI (g‐DMI) have been observed in 10‐nm thick CoPt single layer. However, magnetic anisotropy in gradient‐composition engineered CoPt (g‐CoPt) films is highly sensitive to both the relative Co/Pt composition and the film thickness, leading to a complex interplay with g‐DMI. The stability of skyrmions under the combined influence of magnetic anisotropy and g‐DMI is crucial yet remains poorly understood. Here, we condcut a systematic study on the characteristics of magnetic skyrmions as a function of gradient polarity and effective gradient (defined as gradient/thickness) in g‐CoPt single layers (thickness of 10–30 nm) using magnetic force microscopy (MFM), bulk magnetometry, and topological Hall effect measurements. Brillouin light scattering spectroscopy confirms that both the sign and magnitude of g‐DMI depend on the polarity and amplitude of the composition gradient in g‐CoPt films. MFM reveals that skyrmion size and density vary with g‐CoPt film thickness, gradient polarity, and applied magnetic field. An increased skyrmion density is observed in samples exhibiting higher magnetic anisotropy, in agreement with micromagnetic simulations and energy barrier calculations.