Abstract: Disclosed are techniques, and related systems, for developing and implementing a unified framework for quantifying risk in machine learning models, such as deep neural networks. The framework can be easy-to-use and flexible to apply to different types of machine learning models. The framework can be used to automatically assess different forms of risk in parallel, including but not limited to aleatoric uncertainty, epistemic uncertainty, and/or vacuitic uncertainty. To that end, the disclosed framework provides wrappers that compose different automatic risk assessment algorithms. The obtained risk estimates can be used, for example, for: providing a deeper insight into decision boundaries of neural networks; performing downstream tasks by integrating the risk estimates back into a learning lifecycle for the model to improve robustness and generalization; and/or improving safety by identifying potential model failures based on the risk values.

Abstract: A substrate comprising a III-N base layer comprising a first portion and a second portion, the first portion of the III-N base layer having a first natural lattice constant and a first dislocation density; and a first III-N layer having a second natural lattice constant and a second dislocation density on the III-N base layer, the first III-N layer having a thickness greater than 10 nm. An indium fractional composition of the first III-N layer is greater than 0.1; the second natural lattice constant is at least 1% greater than the first natural lattice constant; a strain-induced lattice constant of the first III-N layer is greater than 1.0055 times the first natural lattice constant; and the second dislocation density is less than 1.5 times the first dislocation density.

Abstract: Strain is used to enhance the properties of p- and n-materials so as to improve the performance of III-N electronic and optoelectronic devices. In one example, transistor devices include a channel aligned along uniaxially strained or relaxed directions of the III-nitride material in the channel. Strain is introduced using buffer layers or source and drain regions of different composition.

Abstract: A novel design for a nitrogen polar high-electron-mobility transistor (HEMT) structure comprising a GaN/InGaN composite channel. As A novel design for a nitrogen polar high-electron-mobility transistor (HEMT) structure comprising a GaN/InGaN composite channel. As illustrated herein, a thin InGaN layer introduced in the channel increases the carrier density, reduces the electric field in the channel, and increases the carrier mobility. The dependence of p on InGaN thickness (tInGaN) and indium composition (xIn) was investigated for different channel thicknesses. With optimized tInGaN and xIn, significant improvements in electron mobility were observed. For a 6 nm channel HEMT, the electron mobility increased from 606 to 1141 cm2/(V·s) when the 6 nm thick pure GaN channel was replaced by the 4 nm GaN/2 nm In0.1Ga0.9N composite channel.

Abstract: A novel design for a nitrogen polar high-electron-mobility transistor (HEMT) structure comprising a GaN/InGaN composite channel. As A novel design for a nitrogen polar high-electron-mobility transistor (HEMT) structure comprising a GaN/InGaN composite channel. As illustrated herein, a thin InGaN layer introduced in the channel increases the carrier density, reduces the electric field in the channel, and increases the carrier mobility. The dependence of p on InGaN thickness (tInGaN) and indium composition (xIn) was investigated for different channel thicknesses. With optimized tInGaN and xIn, significant improvements in electron mobility were observed. For a 6 nm channel HEMT, the electron mobility increased from 606 to 1141 cm2/(V·s) when the 6 nm thick pure GaN channel was replaced by the 4 nm GaN/2 nm In0.1Ga0.9N composite channel.