Abstract: The embodiment of the present disclosure provides a method for crystallizing compound sugar solution of xylose and sucrose, including: introducing a mixed solution of sucrose crystals, xylose crystals and water into a crystallization device, setting a stirring speed in a range of 60 rpm-120 rpm, a temperature in a range of 75° C.-80° C., a vacuum pump pressure in a range of 50 mbar-200 mbar, evaporating the mixed solution until a Brix value of the mixed solution reaches a range of 78 Brix-81 Brix, and stopping the vacuum evaporation, adjusting the temperature to a range of 70° C.-75° C., dropwise adding food-grade isopropanol solution or ethanol solution to the evaporated mixed solution, adding sucrose seed crystals, and continuing stirring to obtain a solution, when small seed crystals grow in the solution, dropping the temperature to a range of 40° C.-60° C. at a rate of 10° C./h, and then stirring for 6 h to obtain the mixed sugar solution; centrifuging and drying at 40° C.-60° C.

Abstract: In a driving policy determining method, for each object, a first target motion trajectory of the object is calculated on a premise that the object does not collide with another object that moves based on an initial motion trajectory. Then, for the ego vehicle, a second target motion trajectory of the ego vehicle is calculated on a premise that the ego vehicle does not collide with another object that moves based on a first target motion trajectory. Then, a driving policy is determined based on the second target motion trajectory of the ego vehicle and a first target motion trajectory of at least one game object. The foregoing operations are repeated until the determined driving policy matches an initial driving policy.

Abstract: This application provides a method for determining an automatic parking strategy. The method includes: determining, a target parking action corresponding to a current parking stage performing the target parking action; obtaining feedback information, where the feedback information is used to indicate whether a result of performing the target parking action reaches a predetermined objective, and the predetermined objective is a predetermined position of the vehicle relative to a target parking spot, and/or the predetermined objective is a status of the vehicle in the parking process; and updating the automatic parking strategy based on the feedback information. In the foregoing method, the entire parking process is divided into several parking stages, and a control strategy is obtained by using a different method at each stage. This can increase a success rate of automatic parking in a complex parking scenario.

Abstract: This application provides a method, a terminal-side device, and a cloud-side device for data processing and a terminal-cloud collaboration system. The method includes: sending, by the terminal-side device, a request message to the cloud-side device; receiving, by the terminal-side device, a second neural network model that is obtained by compressing a first neural network model and that is sent by the cloud-side device, where the first neural network model is a neural network model on the cloud-side device that is used to process the cognitive computing task, and a hardware resource required when the second neural network model runs on the terminal-side device is within an available hardware resource capability range of the terminal-side device; and processing, by the terminal-side device, the cognitive computing task based on the second neural network model.

Abstract: Methods and systems for generating a floorplan for a circuit are disclosed. A netlist graph of the circuit and block features associated with blocks of the circuit are obtained. A reinforcement learning (RL) agent is used to generate a sequence of corner block list (CBL) actions. Each CBL action is generated by: generating a current state embedding representing a current state of the floorplan; and inputting the current state embedding to a policy network of the RL agent to generate a predicted output vector, which is used to generate the CBL action. After each CBL action is generated, the current CBL representation of the floorplan and the block features are updated to reflect the state of the floorplan after applying the CBL action. The CBL representation is outputted as a final floorplan after all blocks have been placed.

Abstract: In the method for optimizing decision-making regulation and control, a first traveling sequence is obtained, where the first traveling sequence includes a first trajectory sequence of the vehicle in information about a first environment and first target driving behavior output by a behavior decision-making layer of a decision-making and control system based on the information about the first environment. A second traveling sequence is obtained, where the second traveling sequence includes a second trajectory sequence output by a motion planning layer of the decision-making and control system based on preset second target driving behavior and the second target driving behavior. The behavior decision-making layer is optimized based on a difference between the first traveling sequence and a preset traveling sequence, and the motion planning layer is optimized based on a difference between the second traveling sequence and the preset traveling sequence.

Abstract: An agent training method includes: obtaining environment information of a first agent and environment information of a second agent; generating first information based on the environment information of the first agent and the environment information of the second agent; and training the first agent by using the first information, so that the first agent outputs individual cognition information and neighborhood cognition information. The neighborhood cognition information of the first agent is consistent with neighborhood cognition information of the second agent.

Abstract: A neural network training method includes inputting neuron input values of a neural network to the RRAM, and performing calculation for the neuron input values based on filters in the RRAM, to obtain neuron output values of the neural network, performing calculation based on kernel values of the RRAM, the neuron input values, the neuron output values, and backpropagation error values of the neural network, to obtain backpropagation update values of the neural network, comparing the backpropagation update values with a preset threshold, and when the backpropagation update values are greater than the preset threshold, updating the filters in the RRAM based on the backpropagation update values.

Abstract: A peripheral circuit includes a data preparation circuit, configured to selectively import, to a row or column of the resistive random access memory (RRAM) crossbar array based on a first control signal, preprocessed data obtained by first preprocessing on first data that is input into the data preparation circuit, a data selection circuit, configured to selectively export second data from the row or column of the RRAM crossbar array based on a second control signal, and perform second preprocessing on the second data to obtain third data, a data reading circuit, configured to: perform a weight update control operation, and perform a max pooling operation on fourth data that is input into the data reading circuit, to obtain fifth data, and a reverse training computation circuit, configured to calculate an error and a derivative of sixth data that is input into the reverse training computation circuit.

Abstract: An autonomous lane change method and apparatus, and a storage medium are provided. The method includes: calculating a local neighbor feature and a global statistical feature of an autonomous vehicle at a current moment based on travel information of the autonomous vehicle at the current moment and motion information of obstacles in lanes within a sensing range of the autonomous vehicle (S1101); obtaining a target action indication based on the local neighbor feature, the global statistical feature, and a current control policy (S1102); and executing the target action according to the target action indication (S1103). It can be learned that, on the basis of the local neighbor feature, the global statistical feature is further introduced into the current control policy to obtain the target action indication. Therefore, the target action obtained by combining local and global road obstacle information is a globally optimal decision action.

Abstract: A neural network obtaining method and a related device are provided. The method may be applied to a scenario in which reinforcement learning is performed on a neural network in the artificial intelligence field. After obtaining a first task, a server obtains a first success rate of completing the first task by using a first neural network. When the first success rate is less than a preset threshold, the server generates a second neural network and a new skill. The server trains, based on a simulated environment corresponding to the first task, the second neural network by using a reinforcement learning algorithm, until a second success rate of completing the first task by using the second neural network is greater than or equal to the preset threshold. The server stores the trained second neural network and the new skill.

Abstract: This application provides a method for determining an automatic parking strategy. The method includes: determining, a target parking action corresponding to a current parking stage performing the target parking action; obtaining feedback information, where the feedback information is used to indicate whether a result of performing the target parking action reaches a predetermined objective, and the predetermined objective is a predetermined position of the vehicle relative to a target parking spot, and/or the predetermined objective is a status of the vehicle in the parking process; and updating the automatic parking strategy based on the feedback information. In the foregoing method, the entire parking process is divided into several parking stages, and a control strategy is obtained by using a different method at each stage. This can increase a success rate of automatic parking in a complex parking scenario.

Abstract: A neural network training method includes inputting neuron input values of a neural network to the RRAM, and performing calculation for the neuron input values based on filters in the RRAM, to obtain neuron output values of the neural network, performing calculation based on kernel values of the RRAM, the neuron input values, the neuron output values, and backpropagation error values of the neural network, to obtain backpropagation update values of the neural network, comparing the backpropagation update values with a preset threshold, and when the backpropagation update values are greater than the preset threshold, updating the filters in the RRAM based on the backpropagation update values.

Abstract: A peripheral circuit includes a data preparation circuit, configured to selectively import, to a row or column of the resistive random access memory (RRAM) crossbar array based on a first control signal, preprocessed data obtained by first preprocessing on first data that is input into the data preparation circuit, a data selection circuit, configured to selectively export second data from the row or column of the RRAM crossbar array based on a second control signal, and perform second preprocessing on the second data to obtain third data, a data reading circuit, configured to: perform a weight update control operation, and perform a max pooling operation on fourth data that is input into the data reading circuit, to obtain fifth data, and a reverse training computation circuit, configured to calculate an error and a derivative of sixth data that is input into the reverse training computation circuit.

Abstract: A computing device includes: a first computing unit configured to perform a first operation on an input first matrix M times, to obtain a second matrix, a second computing unit, configured to perform a second operation on the input second matrix, and a control unit, configured to: control the first computing unit to perform an ith first operation of the M first operations on the first matrix, to obtain an ith data element of the second matrix, store the ith data element of the second matrix into a first storage unit, and control, if data elements currently stored in the first storage unit are sufficient for performing one second operation, the second computing unit to perform a one second operation.

Abstract: This application provides a method, a terminal-side device, and a cloud-side device for data processing and a terminal-cloud collaboration system. The method includes: sending, by the terminal-side device, a request message to the cloud-side device; receiving, by the terminal-side device, a second neural network model that is obtained by compressing a first neural network model and that is sent by the cloud-side device, where the first neural network model is a neural network model on the cloud-side device that is used to process the cognitive computing task, and a hardware resource required when the second neural network model runs on the terminal-side device is within an available hardware resource capability range of the terminal-side device; and processing, by the terminal-side device, the cognitive computing task based on the second neural network model.