Abstract: In an example, an iterative method for generating designs includes receiving, by a computing system, a plurality of symbolic rules and a plurality of design objectives for a design of a system; generating, by the computing system, a first plurality of designs for the system based on the plurality of the symbolic rules; evaluating performance of the first plurality of designs; training a machine learning model using the first plurality of designs and performance metrics; generating a second plurality of designs; evaluating, by the computing system, using a machine learning model, performance of the second plurality of designs to filter one or more designs that meet one or more of the plurality of the design objectives; evaluating performance of the filtered designs; and updating, by the computing system, the plurality of the design objectives and/or the plurality of the symbolic rules based on the evaluated performance of the filtered designs.

Abstract: In an example, a method of designing a system or architecture includes, receiving a plurality of parameter values and a set of requirements for a plurality of objective functions related to a design problem; compressing the plurality of parameters to generate a latent representation; forward processing, with one or more Invertible Neural Networks (INNs), the latent representation to generate a plurality of objective values corresponding to the plurality of the objective functions; inverse processing the plurality of objective values; and generating, based on the latent representation, a plurality of solutions to the design problem that satisfy the set of requirements for the plurality of objective functions.

Abstract: In general, the disclosure describes techniques for facilitating trust in neural networks using a trusted neural network system. For example, described herein are multi-headed, trusted neural network systems that can be trained to satisfy one or more constraints as part of the training process, where such constraints may take the form of one or more logical rules and cause the objective function of at least one the heads of the trusted neural network system to steer, during machine learning model training, the overall objective function for the system toward an optimal solution that satisfies the constraints. The constraints may be non-temporal, temporal, or a combination of non-temporal and temporal. The constraints may be directly compiled to a neural network or otherwise used to train the machine learning model.

Abstract: Techniques are disclosed for identifying fixed bits of a bitstring format. One or more processors are configured to generate a first bitstring having respective first bit values that have a first satisfiability state and generate a second bitstring having respective second bit values that have a second satisfiability state. The one or more processors are configured to identify first potential free bits having respective first common values and generate a third bitstring having first potential free bits with the respective first common values and third remaining bits. The one or more processors are configured to identify second potential free bits having respective second common values and identify a fixed bit that is not included in the first potential free bits and is not included in the second potential free bits.

Abstract: Processing circuitry is configured to obtain a data structure that defines a plurality of conversions of data between pairs of fields; perform a search to identify a plurality of paths from a source node of the data structure to a destination node of the data structure, wherein the source node corresponds to a first field of the fields and the destination node corresponds to a second field of the fields; convert, for each path of the plurality of paths, transforms represented by corresponding edges of the path to a sequence of transforms that conform to a solver format; process the sequence of transforms for each path to determine whether all paths of the plurality of paths are equivalent up to an equivalence relation; and output an indication of whether all paths of the plurality of paths are equivalent up to an equivalence relation.

Abstract: In general, the disclosure describes techniques for characterizing a dynamical system and a neural ordinary differential equation (NODE)-based controller for the dynamical system. An example analysis system is configured to: obtain a set of parameters of a NODE model used to implement the NODE-based controller, the NODE model trained to control the dynamical system; determine, based on the set of parameters, a system property of a combined system comprising the dynamical system and the NODE-based controller, the system property comprising one or more of an accuracy, safety, reliability, reachability, or controllability of the combined system; and output the system property to modify one or more of the dynamical system or the NODE-based controller to meet a required specification for the combined system.

Abstract: Techniques are disclosed for identifying fixed bits of a bitstring format. One or more processors are configured to generate a first bitstring having respective first bit values that have a first satisfiability state and generate a second bitstring having respective second bit values that have a second satisfiability state. The one or more processors are configured to identify first potential free bits having respective first common values and generate a third bitstring having first potential free bits with the respective first common values and third remaining bits. The one or more processors are configured to identify second potential free bits having respective second common values and identify a fixed bit that is not included in the first potential free bits and is not included in the second potential free bits.

Abstract: In general, the disclosure describes techniques for facilitating trust in neural networks using a trusted neural network system. For example, described herein are multi-headed, trusted neural network systems that can be trained to satisfy one or more constraints as part of the training process, where such constraints may take the form of one or more logical rules and cause the objective function of at least one the heads of the trusted neural network system to steer, during machine learning model training, the overall objective function for the system toward an optimal solution that satisfies the constraints. The constraints may be non-temporal, temporal, or a combination of non-temporal and temporal. The constraints may be directly compiled to a neural network or otherwise used to train the machine learning model.