A method for safe reinforcement learning receives an action and a current state of an environment. The method evaluates, using a Logical Neural Network (LNN) structure, an action safetyness logical inference based on the current state of an environment and a current action candidate from an agent. The method outputs upper and lower bounds on the action, responsive to an evaluation of the action safetyness logical inference. The method calculates a contradiction value for the action by using the upper and lower bounds. The contradiction value indicates a level of contradiction for each of a plurality of logic rules implemented by the LNN structure. The method evaluates the action L with respect to safetyness based on the contradiction value. The method selectively performs the action responsive to an evaluation of the action indicating that the action is safe to perform based on the contradiction value exceeding a safetyness threshold.

In response to receiving observed data of mixed observed variables, a mixed causality objective function, being suitable for continuous observed variables and discrete observed variables is determined, wherein the mixed causality objective function includes a causality objective function for continuous observed variables and a causality objective function for discrete observed variables and the fitting inconsistency is adjusted based on weighted factors of the observed variables. Then, the mixed causality objective function is optimally solved by using a mixed sparse causal inference, suitable for both continuous observed variables and discrete observed variables, using the mixed observed data under a constraint of a directed acyclic graph, to estimate causality among the observed variables.

Techniques and systems for solving a set of constraints are described. Binary decision diagram (BDD) learning can be applied to a proper subset of the first set of constraints to obtain a set of bit-level invariants. The set of bit-level invariants can then be used for solving the set of constraints. The set of bit-level invariants can include (1) forbidden invariants, (2) conditional invariants, and/or (3) bit-level invariants that are determined by applying BDD learning to a conjunction of constraints and range expressions. If multiple implied constraints have a common right-hand-side (RHS) expression, then BDD learning can be applied to the common RHS expression only once.

Generate, from a logical formula, a directed acyclic graph having a plurality of nodes and a plurality of edges. Assign an initial embedding to each mode and edge, to one of a plurality of layers. Compute a plurality of initial node states by using feed-forward networks, and construct cross-dependent embeddings between conjecture node embeddings and premise node embeddings. Topologically sort the DAG with the initial embeddings and node states. Beginning from a lowest rank, compute layer-by-layer embedding updates for each of the plurality of layers until a root is reached. Assign the embedding update for the root node as a final embedding for the DAG. Provide the final embedding for the DAG as input to a machine learning system, and carry out the automatic theorem proving with same.

A method of identifying an incident root cause probability that includes identifying, using a monitoring system, a first incident/incident class and generating, using a change management application, a first change request and change class. The method also includes generating, from the change management application, a second change request from a second incident, and where the first and second incidents are in a set of incidents, and where the first and second change requests are in a set of changes, mapping, by a cause analysis application, the set of incidents to the set of changes to identify a root cause probability, where the probability is formed by from a Galois linkage chain between the two sets, and developing, from the cause analysis application, a root cause probability value of the first incident, and executing, using a parameter management application, a mitigation process.

In some embodiments, a robotic process automation (RPA) design application provides a user-friendly graphical user interface that unifies the design of automation activities performed on desktop computers with the design of automation activities performed on mobile computing devices such as smartphones and wearable computers. Some embodiments connect to a model device acting as a substitute for an actual automation target device (e.g. smartphone of specific make and model) and display a model GUI mirroring the output of the respective model device. Some embodiments further enable the user to design an automation workflow by directly interacting with the model GUI.

Aspects and examples are disclosed for improving multi-factor authentication techniques to control access to secured electronic resources. In one example, a decisioning computer system evaluates, based on a passive-dimension decision process, an access request, received from a user device, for a secured electronic resource. The passive-dimension decision process can evaluate dimensions associated with the access request, such as identity or device characteristics, to determine whether the dimensions of the access request are outside of norms for the user. Based on the passive-dimension decision model, the decisioning computing device may communicate to the user device an access decision, the access decision describing one or more of an access authorization, a denial of access, or a supplemental authentication challenge.

Described herein are systems and methods for extracting requirements from regulatory content data. The method including: receiving the regulatory content data; classifying an associated type for each citation in the regulatory content data using a trained classifier machine learning model, the classifier machine learning model trained using regulatory content data including expert labelled annotations; splitting citations in the regulatory content data, including determining whether each citation includes more than one requirement; merging one or more citations in the regulatory content data, including identifying child-parent relationships for the citations and merging citations based on conjunctive language; and outputting the citations and their associated type. In a particular case, the types of citations for classification include one of a requirement (REQ), an optional or site-specific requirement (OSR), a description (DSC), and part of another requirement.

Aspects of the invention include obtaining a set of data that includes inputs and outputs to be modelled and performing a symbolic regression to find a symbolic model that fits the inputs and the outputs of the set of data. The symbolic model is a symbolic expression discovered by the symbolic regression in a search space. Automated reasoning is performed to affect a final symbolic model that is used to obtain new outputs from new inputs based on the final symbolic model.

Techniques and systems for solving Boolean satisfiability (SAT) problems are described. Some embodiments solve SAT problems using efficient construction of truth tables. Some embodiments can improve performance of SAT solvers by using truth tables instead of incurring the overhead of Conjunctive Normal Form (CNF) conversion.