Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
and JavaScript.
A methodology — called auto tiny classifiers — is proposed to directly generate predictor circuits for the classification of tabular data, searching over the space of combinational logic using an evolutionary algorithm to maximize training prediction accuracy. Prediction performance is comparable to typical machine learning methods, but substantially fewer hardware resources and power are required.
An ingestible electronic device can record biopotential electrical signals from the gastric environment—including the gastric slow wave, respiration signal and heart signal—and can monitor slow wave activity in freely moving and feeding animals.
The quantum anomalous Hall effect holds promise for quantum resistance metrology, but has been limited to low operating currents. A measurement scheme that increases the effect’s operational current is now demonstrated — a scheme that could also be used more generally to improve the performance of existing primary quantum standards of resistance based on the conventional quantum Hall effect.
Robust conductive hydrogels made purely from a conducting polymer can be fabricated using a laser-induced phase separation method that also improves adhesion to a polymer substrate and allows high-resolution selective patterning.
As the scale and application of artificial intelligence technologies continues to grow, addressing challenges related to the wider accessibility of the underlying technology becomes increasingly important.
A non-volatile memory device that is based on an aluminium scandium nitride (Al0.68Sc0.32N) ferroelectric diode can operate at temperatures of up to 600 °C.
By integrating a metal‒oxide‒semiconductor capacitor into a two-terminal diode, a multifunctional single device can be created that operates as a tunable light-emitting diode with a built-in bias tee circuit and a detector with a reconfigurable optoelectronic logic function.
A graph-based genetic programming method can be used to automatically generate small and energy-efficient circuits from tabular data for machine learning classification tasks.
A method for integrating polycrystalline molybdenum disulfide using processes in a 200 mm fab facility can create transistors with high robustness and performance comparable with single-crystalline devices.
Nanofluidic memristors that rely on mechanical deformations to modulate ionic conductance can be coupled to form logic circuits, opening a route to ionic machinery that could implement neural networks.
High-density device arrays can be integrated on flexible substrates using a dip-transfer coating method that suppresses adhesive layers from forming between closely spaced devices and uses magnetically self-assembled particles to increase the anisotropic conductivity.
A measurement scheme in which current is injected simultaneously into two disconnected perimeters of a multi-terminal Corbino device can be used to increase the robustness of a zero-magnetic-field quantum anomalous Hall resistor, thus extending its operating range to higher currents.
Using an intrinsically stretchable nanocomposite of quantum dots, an elastomer and a hole transport polymer as an emissive layer, stretchable light-emitting diodes can be fabricated that exhibit high brightness even under 50% strain.