US 12,462,147 B2
Firing neural network computing system and method for brain-like intelligence and cognitive computing
Hualong Ren, Guangdong (CN)
Assigned to NEUROCEAN TECHNOLOGIES INC., Guangdong (CN)
Appl. No. 17/623,753
Filed by NEUROCEAN TECHNOLOGIES INC., Guangdong (CN)
PCT Filed Jul. 1, 2020, PCT No. PCT/CN2020/099714
§ 371(c)(1), (2) Date Dec. 29, 2021,
PCT Pub. No. WO2021/000890, PCT Pub. Date Jan. 7, 2021.
Claims priority of application No. 201910588964.5 (CN), filed on Jul. 2, 2019.
Prior Publication US 2022/0253675 A1, Aug. 11, 2022
Int. Cl. G06N 3/04 (2023.01); G06N 3/049 (2023.01); G06N 3/08 (2023.01); G06N 3/10 (2006.01)
CPC G06N 3/049 (2013.01) [G06N 3/08 (2013.01); G06N 3/105 (2013.01)] 38 Claims
OG exemplary drawing
 
1. A firing neural network computing system for brain-like intelligence and cognitive computing, wherein the firing neural network computing system comprises a computer processor including:
a model description module configured to: provide an interface for a user to design and describe a network model, and specify an operation to be executed and a rule for an object of the network model including at least one execution rule for automatic pruning and/or genesis of synapses and neurons;
a parameter database configured to store all parameter data of the network model in a form of the parameter database;
a configuration description module configured to: describe configuration parameters of a current network operating environment, and a condition for initiating a synapse and/or neuron pruning and genesis process;
a configuration manager configured to invoke configuration parameters from the configuration description module;
a network builder configured to: read the model description module, parse a topological structure of a network, read a data file through a data manager, and construct the network model object in a memory;
a network manager configured to: construct, traverse, access, and/or update the network model object;
a rule manager configured to: read rules declared by the user in the model description module, interpret the rules declared by the user when a scheduler schedules an operation of the network model object, and arbitrate a conflict between the rules;
the data manager configured to: read and parse the parameter database, convert a data format and serialize data;
the scheduler configured to: allocate hardware resources and schedule operation processes, and schedule and execute a corresponding operation, the scheduler further configured to perform pruning and/or genesis of synapses and neurons in accordance with the at least one execution rule, wherein the scheduler without requiring manual programming intervention by the user automatically performs pruning and/or genesis of synapses and neurons in accordance with the at least one execution rule when the condition for initiating the synapse and/or neuron pruning and genesis process is met;
an operation manager configured to manage running operations;
a log manager configured to: record a log generated when the system is running, record a working status of the system, and prompt an abnormal condition;
an operation monitoring module configured to: receive and respond to an instruction input by the user, and manage an operating status of the system; and,
a graphical display module configured to read and display data of the network.