Abstract: A neural network model processing method includes obtaining a first low-bit neural network model through training, where the model includes a first operation layer and a second operation layer. Each operation layer includes at least one operation. Values/a value of a parameter and/or data used for the operation are/is represented by using N bits, and N is a positive integer less than 8. The neural network model processing method further includes compressing the model to obtain a second low-bit neural network model, where the compressed model includes a third operation layer. The third operation layer is equivalent to the first operation layer and the second operation layer, and an operation layer other than the third operation layer in the at least one operation layer is the same as an operation layer other than the first operation layer and the second operation layer in the at least two operation layers.

Abstract: This application provides a method and apparatus for distributed processing of a UX element, and pertains to the field of terminal artificial intelligence. The method includes: obtaining UX capability information of at least two terminal devices, where the UX capability information is used to indicate a UX capability corresponding to each of interaction manners supported by the terminal device; receiving a target task request, where the target task includes UX requirement information of at least two UX elements; determining, based on the UX capability information of the at least two terminal devices and the UX requirement information of the at least two UX elements, a UX element that needs to be processed by each of the at least two terminal devices; and providing, for each of the at least two terminal devices, the UX element that needs to be processed by each of the at least two terminal devices.

Abstract: This application relates to the field of artificial intelligence technologies, and discloses a machine learning model search method, a related apparatus, and a device. In the method, before model search and quantization, a plurality of single bit models are generated based on a to-be-quantized model, and evaluation parameters of layer structures in the plurality of single bit models are obtained. Further, after a candidate model selected from a candidate set is trained and tested, to obtain a target model, a quantization weight of each layer structure in the target model may be determined based on a network structure of the target model and evaluation parameters of all layer structures in the target model, a layer structure with a maximum quantization weight in the target model is quantized, and a model obtained through quantization is added to the candidate set.

Abstract: A neural network model processing method includes obtaining a first low-bit neural network model through training, where the model includes a first operation layer and a second operation layer. Each operation layer includes at least one operation. Values/a value of a parameter and/or data used for the operation are/is represented by using N bits, and N is a positive integer less than 8. The neural network model processing method further includes compressing the model to obtain a second low-bit neural network model, where the compressed model includes a third operation layer. The third operation layer is equivalent to the first operation layer and the second operation layer, and an operation layer other than the third operation layer in the at least one operation layer is the same as an operation layer other than the first operation layer and the second operation layer in the at least two operation layers.

Abstract: A voice information processing method includes: receiving, by a terminal, voice information, and converting the voice information into text information; obtaining a field probability that the text information belongs to each of preset M event fields; obtaining a prior probability that the text information belongs to each of N event fields, where N?M; obtaining confidence that the text information belongs to each of the N event fields; calculating, based on the field probability, the prior probability, and the confidence that the text information belongs to each of the N event fields, values of N probabilities that the text information separately belongs to the N event fields; and outputting a semantic understanding result of performing semantic understanding based on an event field corresponding to a largest probability value in the N probability values.

Abstract: A voice information processing method includes: receiving, by a terminal, voice information, and converting the voice information into text information; obtaining a field probability that the text information belongs to each of preset M event fields; obtaining a prior probability that the text information belongs to each of N event fields, where N?M; obtaining confidence that the text information belongs to each of the N event fields; calculating, based on the field probability, the prior probability, and the confidence that the text information belongs to each of the N event fields, values of N probabilities that the text information separately belongs to the N event fields; and outputting a semantic understanding result of performing semantic understanding based on an event field corresponding to a largest probability value in the N probability values.

Abstract: The present disclosure provides a network service establishment method, an orchestration control center, and a network system. The method includes: obtaining, by means of screening according to QoP information and SLA information of a service, all target service protection types that satisfy the QoP information and/or the SLA information of the service for each layer network in a multi-layer network; determining, according to all the target service protection types, at least one group of multi-layer service protection types that satisfy the QoP information and the SLA information of the service for the multi-layer network; determining a group of target multi-layer service protection types of the service according to network topology information and resource usage of each layer network from the at least one group of the multi-layer service protection types; and sending an instruction message that carries the target multi-layer service protection types to a network controller.

Abstract: The present application provides a method for planning a recovery resource for resisting N-time faults and an optical transmission device, and the method includes: planning, on an optical transmission device according to preset network planning information, a recovery resource for resisting (N?1)-time faults for preset (N?1)-time faults, and the recovery resource for resisting (N?1)-time faults is an optimal recovery resource corresponding to each interrupted service during the preset (N?1)-time faults; and planning, by the optical transmission device according to the network planning information and the recovery resource for resisting (N?1)-time faults, a recovery resource for resisting N-time faults for preset N-time faults, where the recovery resource for resisting N-time faults is a network-wide optimal recovery resource corresponding to interrupted services during the N-time faults.

Abstract: The present disclosure provides a network service establishment method, an orchestration control center, and a network system. The method includes: obtaining, by means of screening according to QoP information and SLA information of a service, all target service protection types that satisfy the QoP information and/or the SLA information of the service for each layer network in a multi-layer network; determining, according to all the target service protection types, at least one group of multi-layer service protection types that satisfy the QoP information and the SLA information of the service for the multi-layer network; determining a group of target multi-layer service protection types of the service according to network topology information and resource usage of each layer network from the at least one group of the multi-layer service protection types; and sending an instruction message that carries the target multi-layer service protection types to a network controller.

Abstract: The present application provides a method for planning a recovery resource for resisting N-time faults and an optical transmission device, and the method includes: planning, on an optical transmission device according to preset network planning information, a recovery resource for resisting (N?1)-time faults for preset (N?1)-time faults, and the recovery resource for resisting (N?1)-time faults is an optimal recovery resource corresponding to each interrupted service during the preset (N?1)-time faults; and planning, by the optical transmission device according to the network planning information and the recovery resource for resisting (N?1)-time faults, a recovery resource for resisting N-time faults for preset N-time faults, where the recovery resource for resisting N-time faults is a network-wide optimal recovery resource corresponding to interrupted services during the N-time faults.

Abstract: A method for protecting a bi-connected network against a dual link failure, the method comprising finding a plurality of cut-pairs for a data flow, finding a work path and a primary backup path for the data flow, determining the cut-pairs that overlap both the work path and the primary backup path, merging the cut-pairs that overlap the work path and the primary backup path into a plurality of cut-groups, wherein each cut-group comprises a first half and a second half, and constructing a plurality of partially disjoint backup paths based on the cut-groups, wherein one of the partially disjoint backup paths protects against the dual link failure that causes a failure on the second half of a first cut-group and a failure on the first half of a second cut-group.

Abstract: A method for protecting a bi-connected network against a dual link failure, the method comprising finding a plurality of cut-pairs for a data flow, finding a work path and a primary backup path for the data flow, determining the cut-pairs that overlap both the work path and the primary backup path, merging the cut-pairs that overlap the work path and the primary backup path into a plurality of cut-groups, wherein each cut-group comprises a first half and a second half, and constructing a plurality of partially disjoint backup paths based on the cut-groups, wherein one of the partially disjoint backup paths protects against the dual link failure that causes a failure on the second half of a first cut-group and a failure on the first half of a second cut-group.