CN117688969A - Port model object organization method, system, platform, intelligent device and medium - Google Patents
Port model object organization method, system, platform, intelligent device and medium Download PDFInfo
- Publication number
- CN117688969A CN117688969A CN202211067985.0A CN202211067985A CN117688969A CN 117688969 A CN117688969 A CN 117688969A CN 202211067985 A CN202211067985 A CN 202211067985A CN 117688969 A CN117688969 A CN 117688969A
- Authority
- CN
- China
- Prior art keywords
- port
- model object
- model
- neuron
- input
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000008520 organization Effects 0.000 title claims abstract description 86
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000013528 artificial neural network Methods 0.000 claims abstract description 22
- 230000003993 interaction Effects 0.000 claims abstract description 17
- 210000002569 neuron Anatomy 0.000 claims description 102
- 210000000225 synapse Anatomy 0.000 claims description 48
- 230000007246 mechanism Effects 0.000 claims description 23
- 230000001537 neural effect Effects 0.000 claims description 17
- 230000000946 synaptic effect Effects 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 11
- 230000003956 synaptic plasticity Effects 0.000 claims description 7
- 239000012528 membrane Substances 0.000 claims description 4
- 230000004044 response Effects 0.000 claims description 2
- 238000010276 construction Methods 0.000 abstract description 12
- 238000010586 diagram Methods 0.000 description 33
- 238000003062 neural network model Methods 0.000 description 20
- 238000013334 tissue model Methods 0.000 description 12
- 238000004364 calculation method Methods 0.000 description 9
- 210000001787 dendrite Anatomy 0.000 description 5
- 210000004242 electrical synapse Anatomy 0.000 description 4
- 230000000670 limiting effect Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000001629 suppression Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 210000004556 brain Anatomy 0.000 description 2
- 230000001149 cognitive effect Effects 0.000 description 2
- 230000036279 refractory period Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000010304 firing Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 210000000653 nervous system Anatomy 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001550 time effect Effects 0.000 description 1
- 238000011144 upstream manufacturing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The embodiment of the invention discloses a port model object organization method, a port model object organization system, a port model object organization platform, an intelligent device and a storage medium. The port model object organization method comprises the following steps: acquiring at least two port model objects, wherein each port model object comprises at least one port and a model main body, the model main body is associated with data and/or operations, and the ports are used for carrying out information interaction or association with the outside of the port model objects; and carrying out information interaction or association on at least two port model objects through at least one port of each port model object to form an organization model object. The invention can effectively improve the construction efficiency of the neural network and reduce the construction cost.
Description
Technical Field
The invention relates to the technical field of neural networks, in particular to a port model object organization method, a port model object organization system, a port model object organization platform, an intelligent device and a storage medium.
Background
In building a neural network model, a builder typically builds a system with the aid of the neural network to reduce the cost of building the neural network model and to increase the efficiency of building the neural network model. The existing neural network construction system can provide a multilingual construction environment, so that a constructor can select a more familiar environment to construct a neural network model.
For the neural network model applied to the fields of intelligent robots, automatic driving, automation and the like, the structure is larger, a large number of neurons, nerve layers and the like are required to be deployed, the more complicated the structure of the neural network model is, the more difficult a user manages the neural network model is, and meanwhile, more workers need to participate in the construction work of the neural network model are required.
Disclosure of Invention
In view of the above, it is necessary to provide a port model object organization method, a system, a platform, an intelligent device, and a storage medium.
A port model object organization method, comprising:
acquiring at least two port model objects, wherein each port model object comprises at least one port and a model main body, the model main body is associated with data and/or operations, and the ports are used for carrying out information interaction or association with the outside of the port model objects;
and carrying out information interaction or association on at least two port model objects through at least one port of each port model object to form an organization model object.
A port model object organization system, comprising:
the system comprises an acquisition module, a storage module and a storage module, wherein the acquisition module is used for acquiring at least two port model objects, each port model object comprises at least one port and a model main body, the model main body is associated with data and/or operation, and the ports are used for carrying out information interaction or association with the outside of the port model;
and the organization module is used for carrying out information interaction or association on at least two port model objects through at least one port of each port model object to form an organization model object.
A neural network build platform, comprising: a front end portion, a core portion, and a rear end portion;
the front-end portion is used for acting the outside of the platform to interact with the core portion and/or the back-end portion;
the back end part is used for running a target model;
the core part is used for responding to the request of the front end part and/or the back end part and describing and/or organizing the target model, and comprises a port module for realizing the method.
A smart device comprising a memory having a computer program stored therein and a processor for executing the computer program to implement the method as described above.
A storage medium storing a computer program capable of being loaded by a processor and executing a method as described above.
The embodiment of the invention has the following beneficial effects:
in this embodiment, at least two port model objects are obtained, and at least two port model objects are subjected to information interaction or association through at least one respective port to form an organization model object, so that a user can be helped to conveniently and rapidly construct various structures required by a neural network, when the user manages a plurality of people to construct the neural network model simultaneously according to modules, respective modules can be completed respectively, and a model with larger scale and higher level can be constructed based on the port modules, thereby effectively improving the construction efficiency of the neural network and reducing the construction cost.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Wherein:
FIG. 1 is a flow chart of a first embodiment of a port model object organization method provided by the present invention;
FIG. 2 is a schematic diagram of a first embodiment of a port model object provided by the present invention;
FIG. 3 is a schematic diagram of a second embodiment of a port model object provided by the present invention;
FIG. 4 is a schematic view of a flat structure provided by the present invention;
FIG. 5 is a schematic diagram of a tree structure according to the present invention;
FIG. 6 is a schematic diagram of a cascade structure provided by the present invention;
FIG. 7 is a schematic illustration of a nested configuration provided by the present invention;
FIG. 8 is a schematic diagram of an embodiment of a unidirectional neural link provided by the present invention;
FIG. 9 is a schematic diagram of an embodiment of a neural circuit provided by the present invention;
FIG. 10 is a schematic diagram of an embodiment of an electrical synapse provided in accordance with the present disclosure;
FIG. 11 is a schematic diagram of an embodiment of neuronal refractory period feedback provided by the present invention;
FIG. 12 is a schematic diagram of an embodiment of an associative synaptic plasticity mechanism model provided by the present invention;
FIG. 13 is a schematic diagram of a structure of a neuron group model object according to the present invention;
FIG. 14 is a schematic diagram of a structure of a synapse set model object provided in the present invention;
FIG. 15 is a schematic diagram of a first embodiment of a tissue model object provided by the present invention;
FIG. 16 is a schematic diagram of a second embodiment of a tissue model object provided by the present invention;
FIG. 17 is a schematic diagram of a third embodiment of a tissue model object provided by the present invention;
FIG. 18 is a schematic diagram illustrating the architecture of one embodiment of a port model object organization system provided by the present invention;
FIG. 19 is a schematic diagram illustrating the structure of an embodiment of a neural network building platform according to the present invention;
fig. 20 is an internal structural diagram of the smart device in one embodiment provided by the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In building a neural network model, such as a brain-like neural network model, using a neural network building system, a constructor needs to manually write information flow between neurons and between neurons. When the quantity of the neural network model to be constructed is large, the writing workload is large, the constructor is easy to make mistakes, and the time cost and the labor cost for constructing the neural network model are increased.
In order to solve the problems, the invention discloses a port model object, two port model objects can be organized through ports, such as interconnection, or open authority and the like, the information flow direction of a neural network can be clearer through the port model object, the model organization form is more flexible, the modularization degree is better, and the cooperation development of different cooperators is easier.
In modeling DNN (Deep Neural Networks, deep neural network), more complex models are built primarily in a hierarchical stack of modules, with a single module or model supporting both input and output, one module can receive 0 or more inputs, yielding one or more outputs. However, many defects still exist in the mode of topological connection of DNNs:
first, ring connection is not supported in DNN, which is determined by information transmission, learning mode and calculation mode of DNN, and currently, the mainstream DNN framework adopts this mode. Since ring topologies are not supported, there are great limitations in describing complex topologies because there are a large number of ring connection topologies in the biological nervous system, such as the basic loops of feedback suppression, feedforward suppression, and mutual suppression are all ring topologies.
Secondly, the topological connection mode of DNN only supports simple information transmission, does not support more complex connection description, has relatively weak description capability for connection topology, and also lacks an effective expansion mechanism.
Finally, as the brain-like calculation and the traditional DNN calculation paradigm have large difference, DNN adopts a front-to-back calculation paradigm or a back-to-front calculation paradigm, strict data dependence exists among modules, so that the locality of calculation is low, certain parallelism exists only when a plurality of parallel lines exist in a topological structure, and otherwise, strict calculation sequence exists among operations; the inverse brain-like calculation is performed by taking neurons as cores, has good locality, does not emphasize strict information dependence although having a complex topological structure, and has a good filtering effect on the influence of data dependence due to pulse neurons, and the accumulation of the pulse neurons on time effect can effectively eliminate the delay influence caused by the data dependence, so that the brain-like neural network has high parallelism, and the calculation paradigm of the brain-like neural network and the DNN (digital network) are completely different.
Therefore, the invention provides a port model object organization method, a port expresses the contact information of the module object and other module objects, the contact information comprises connection information or authority information, a port mechanism is introduced, the description capacity and the expansion capacity of a topological structure are increased, the topological relation is subjected to unified modeling through the port, the expansion capacity of the topological structure is ensured by supporting the expansion of a port mode, the expansion can be carried out according to the requirement, the modeling requirement which is changed continuously is met, the efficiency of constructing the neural network by a user is effectively improved, and the construction cost is reduced.
Referring to fig. 1 in combination, fig. 1 is a flowchart illustrating a first embodiment of a port model object organization method according to the present invention.
S101: at least two port model objects are obtained, each port model object comprises at least one port and a model main body, the model main body is associated with data and/or operations, and the ports are used for carrying out information interaction or association with the outside of the port model objects.
In a specific implementation scenario, at least two port model objects are obtained, where the port model objects may be port model objects stored in advance in the system, or may be port model objects customized by a user. For example, a user may add at least one port to a specified model object to form a port model object.
Each port model object comprises at least one port and a model body, the model body is associated with data and/or operation, the operation is an actual dynamic process represented by the model body, the output is generated to the downstream after the input is accepted, and the data expresses the characteristics of the port model object and can be in specific forms such as variable values, character strings, tensors and the like. Referring to fig. 2 in combination, fig. 2 is a schematic structural diagram of a first embodiment of a port model object according to the present invention. The at least one port of the port model object includes at least one of an input port, an output port, a reference port, and a connection port. The input port input1 is used for receiving input information provided by an external module outside the port model object, the output port output1 is used for outputting output information of the model object to the outside of the model object, the reference port reference is used for referencing a mechanism of a variable of other model objects to the model object, and the connection port matrix is used for connecting and binding the model object with the other model objects.
Referring to fig. 3 in combination, fig. 3 is a schematic structural diagram of a second embodiment of a port model object according to the present invention. There is a complex correspondence between the input information received at the input port of the port model object of fig. 3 and the output information output at the output port, and there is a large amount of randomness and uncertainty, such as an uncertain synaptic connection between the upstream and downstream neuron group model objects, a connection of the synaptic model object to the dendritic model object, etc. In other implementation scenarios, the relationship between the input information and the output information is generally a fixed correspondence relationship, that is, after the construction parameters of the port model object are given, the relationship between the input information and the output information is already determined, and no additional specified operation is required by the user.
In one implementation, the model body includes at least one of a neuron, a synapse, a group of neurons, a group of synapses, a dendrite, a group of dendrites. When the model body is a neuron, the port model object is a neuron model object, when the model body is a neuron group, the port model object is a neuron group model object, when the model body is a synapse, the port model object is a synapse group model object, when the model body is a synapse group, the port model object is a dendrite model object, when the model body is a dendrite, the port model object is a dendrite group model object.
In one implementation, the input port may receive only one connection, accepting a new connection will cancel the old connection, but the input information received may be multiple. The input ports in this application can be further extended to specific port forms. In one implementation scenario, the input ports include a variable input port and a proxy input port, and the content actually received by the variable input port may be a numerical value, a character string or a tensor, and the content is determined by a variable bound by the variable input port. The variable input port reloads the receive operation, handling connections from other ports or outside the port model object. The proxy input port, which is a proxy port, may be connected to an input port of an internal module (e.g., at least one port object model inside) inside the port model object, or may accept connection of ports of other model objects outside the port model object, for receiving external input.
The output port can be bound to other input ports to set a topology structure, can be connected to an input port of an internal module (such as at least one port object model) of the port model object, can also be an input port of other model objects outside the port model object, and can be bound to a plurality of input ports. In one implementation, the output ports include a variable output port and a proxy output port. The content actually output by the variable output port is determined by the bound variable, and the content actually output can be a numerical value, a character string or a tensor. The proxy output port may be connected as an output port to an input port of another port model object external to the own port model object, or may accept a connection from an output port of an internal module (for example, at least one port object model inside) of the own port model object, and proxy the output of the internal module. The proxy output port is connected with the output port of at least one port model object in the port model object, and is used for receiving the output information of the at least one port model object and outputting the output information to the outside.
The reference port provides the port model object with a mechanism for referencing the variables of other model objects, and the two parties have strict binding relation. The specific data type referenced is determined by the variable it binds to and may be a numerical value, a string, or a tensor. The reference is a strong association operation, and has strict limitation on both bound model objects. References may be further distinguished as read-only references and general references. The reference port often has a certain limiting effect on the execution scheduling of the port model object, and the scheduling flexibility is reduced to a certain extent.
The connection port is a model management interface used for being bound to a connection model (ConnectionModel), and the extremely strong description capability and expansion capability of the port mechanism provided by the invention are embodied. The connection port takes the connection model as a binding object, and is used for binding the port model object and a specific connection model. The connection port is also a strongly-associated port mechanism, and a strong logic binding relationship exists between the associator and the associator, and the logic relationship is not only reflected on the organization of the port model object, but also can be deeply associated on the realization of the final organization model object, for example, the synaptic plasticity and the binding of the synaptic group, and the mapping conversion and the binding of the dendritic group are all carried out through the connection port.
S102: and carrying out information interaction or association on at least two port model objects through at least one port of each port model object to form an organization model object.
In a specific implementation scenario, at least two port model objects interact or correlate information through at least one port of each to form an organization model object. For example, the output port a of one port model object a is connected to the input port b of another port model object b. The output port C1 and the input port C2 of one port model object C may be connected to form an organization model object of a self-circulation structure.
In one implementation, the structure of the organizational model object includes a flat structure, a tree structure, a cascade structure, and a nested structure. Referring to fig. 4-7 in combination, fig. 4 is a schematic structural view of a flat structure provided by the present invention, fig. 5 is a schematic structural view of a tree structure provided by the present invention, fig. 6 is a schematic structural view of a cascade structure provided by the present invention, and fig. 7 is a schematic structural view of a nested structure provided by the present invention. A port model object is shown in rectangular blocks in fig. 4-7, and there is no limitation as to which type of port is specifically associated between each port model object.
In one implementation scenario, when the port model object is a neuron model object, an input port of the neuron model object is capable of receiving an input of at least one neuron, and an output port of the neuron model object is capable of connecting with the at least one neuron to provide output information to the at least one neuron. And sequentially connecting the input ports of the multiple neuron model objects and the input ports in one way according to the connection direction designated by the calling instruction to form a one-way neural link. Referring to fig. 8 in combination, fig. 8 is a schematic structural diagram of an embodiment of a unidirectional neural link according to the present invention. When the output port of the latter one of the plurality of neuron model objects and the input port of the former one of the plurality of neuron model objects are connected, a neural loop is formed. Referring to fig. 9 in combination, fig. 9 is a schematic structural diagram of a neural loop according to an embodiment of the invention. At least two neuron port model objects are associated through respective reference ports to share membrane potential information to achieve electrical synapses. Referring to fig. 10 in combination, fig. 10 is a schematic structural diagram of an electrical synapse according to an embodiment of the present disclosure. And associating the neuron port model object with the non-stress mechanism model through a connection port so as to realize feedback of the non-stress period of the neuron. Referring to fig. 11 in combination, fig. 11 is a schematic structural diagram of an embodiment of feedback of the refractory period of neurons according to the present invention.
When the port model object is a synaptic model object, the input port of the synaptic model object is capable of receiving an input of at least one neuron and the output port of the synaptic model object is capable of being connected to at least one neuron. The connection port of the synaptic model object is associated with a synaptic plasticity mechanism model for accessing and altering synaptic weights. Referring to fig. 12 in combination, fig. 12 is a schematic structural diagram of an embodiment of an associative synaptic plasticity mechanism model according to the present invention.
When the port model object is a neuron group model object, the input port of the neuron group model object can be connected to the input of at least one synapse group, and the output port of the neuron group model object can be connected to the at least one synapse group. Referring to fig. 13 in combination, fig. 13 is a schematic structural diagram of a neuronal group model object according to the present invention.
When the port model object is a synapse group model object, an input port of the synapse group model object can be connected to an input of at least one neuron group, and an output port of the synapse group model object can be connected to the at least one neuron group. Referring to fig. 14 in combination, fig. 14 is a schematic structural diagram of a synapse set model object provided in the invention.
In another implementation scenario, the model objects further comprise at least one tissue model object or at least one port model object. Referring to fig. 15 in combination, fig. 15 is a schematic structural diagram of a first embodiment of an object of an organization model according to the present invention. In the implementation scenario shown in fig. 15, the model object is the port model object shown in fig. 2. The proxy input port is connected with the input port of the internal port model object, and the proxy output port is connected with the output port of the internal port model object.
Referring to fig. 16 in combination, fig. 16 is a schematic structural diagram of a second embodiment of an object of an organization model according to the present invention. In the implementation scenario shown in fig. 16, the first port model object, the second port model object, and the first organization model object are sequentially connected in a preset order to form a new port model object, and a proxy output port is added to the port model object. The first tissue model object is the tissue model object shown in fig. 15. The output port of the first port model object (sequenceinput model object) is connected to the input port of the second port model object (simpleSynapse model object), the output port of the second port model object (simpleSynapse model object) is connected to the proxy input port of the first organization model object (NeuronModel Instance1 model object), and the proxy output port of the first organization model object (NeuronModel Instance model object) of the proxy output port of the new port model object is connected.
Referring to fig. 17 in combination, fig. 17 is a schematic structural diagram of a third embodiment of an object of an organization model according to the present invention. In the implementation scenario shown in fig. 17, the third port model object, the fourth port model object, and the second tissue model object are sequentially connected in a preset order to form the tissue model object. The second tissue model object is the tissue model object shown in fig. 16. The output port of the third port model object (ExternalModule Instance model object) is connected to the input port of the fourth port model object (simplesynappase model object), and the output port of the second port model object (simplesynappase model object) is connected to the proxy input port of the second organization model object (NeuronModel Instance model object).
As can be seen from the above description, in this embodiment, by acquiring at least two port model objects, and performing information interaction or association on at least two port model objects through at least one respective port to form an organization model object, it is possible to help a user to conveniently and quickly construct various structures required by a neural network, so that when a user manages a neural network model according to a module, conveniently queries parameters and updates parameters, and performs construction of the neural network model simultaneously, respective modules can be completed, and a larger-scale and higher-level model is constructed based on the port modules, thereby reducing construction cost of the neural network and improving construction efficiency.
Referring to fig. 18, fig. 18 is a schematic structural diagram of an embodiment of a port model object organization system according to the present invention. The port model object 10 includes: an acquisition module 11 and an organization module 12.
The obtaining module 11 is configured to obtain at least two port model objects, where each port model object includes at least one port and a model body, where the model body is associated with data and/or operations, and the port is configured to interact or associate information with an outside of the port model; the organization module 12 is configured to interact or associate information of at least two port model objects through at least one respective port to form an organization model object.
Wherein the ports include at least one of an input port, an output port, a reference port, and a connection port.
The input port is used for receiving input information outside the port model object, the output port is used for outputting output information of the model object, the reference port is used for referencing a mechanism of variables of other model objects to the model object, and the connection port is used for connecting and binding the model object with the other model objects.
The input information received by the input port and the output information output by the output port of the same port model object have complex corresponding relation or fixed corresponding relation output information.
Wherein the model body comprises at least one tissue model object and/or at least one port model object.
The input ports comprise variable input ports and proxy input ports, and the output ports comprise variable output ports and proxy output ports.
Wherein when the model body comprises at least one other tissue model object and/or at least one other port model object, the input port comprises a proxy input port and the output port comprises a proxy output port. The agent input port is connected with the input port of at least one port model object in the port model object and is used for receiving input information required by the model main body; the agent output port is connected with the output port of at least one port model object in the port model object, and is used for acquiring the output information of the model main body and outputting the output information to the outside.
The structure of the organization model object comprises a flat structure, a tree structure, a cascade structure and a nested structure.
Wherein the model body further comprises at least one of a neuron, a synapse, a group of neurons, a group of synapses.
When the model body comprises neurons, the port model object is a neuron model object, an input port of the neuron model object can receive input of at least one neuron, and an output port of the neuron model object can be connected to the at least one neuron.
The organization module 12 is further configured to sequentially connect the input ports of the plurality of neuron model objects and the input ports in a unidirectional manner according to a connection direction specified by the call instruction, so as to form a unidirectional neural link; or when the output port of the later one of the plurality of neuron model objects and the input port of the earlier one of the plurality of neuron model objects in the unidirectional neural link are connected according to the calling instruction, the neural loop is formed.
The organization module 12 is further configured to associate at least two neuron port model objects through respective reference ports to share membrane potential information to achieve electrical synapses.
The organization module 12 is further configured to associate the neuron port model object with the model of the refractory mechanism via the connection port to implement feedback of the period of the neuron refractory.
When the model body comprises synapses, the port model object is a synapse model object, an input port of the synapse model object can receive input of at least one neuron, and an output port of the synapse model object can be connected to the at least one neuron.
The organization module 12 is also used to associate connection ports of the synaptic model objects with a synaptic plasticity mechanism model that is used to access and alter synaptic weights.
When the model body comprises a neuron group, the port model object is a neuron group model object, an input port of the neuron group model object can be connected with an input of at least one synapse group, and an output port of the neuron group model object can be connected with at least one synapse group.
When the model body comprises a synapse group, the port model object is a synapse group model object, an input port of the synapse group model object can be connected with an input of at least one neuron group, and an output port of the synapse group model object can be connected with the at least one neuron group.
Fig. 19 is a schematic structural diagram of an embodiment of a neural network building platform according to the present invention, and fig. 1 is applied to the neural network building platform shown in fig. 19. The neural network building platform 100 includes a front-end portion 200, a core portion 300, and a back-end portion 400. The front-end portion 200 is used to interact with the core portion 300 and/or the back-end portion 400 outside of the proxy neural network build platform 100; the back end section 400 is used to run the target model; the core portion 300 is configured to perform description and/or organization of the object model in response to a request from the front end portion 200 and/or the back end portion 300, and the core portion 300 includes a port module 301, where the port module 301 is capable of implementing the port model object organization method as described above.
In one implementation scenario, the target model refers to a neural network model, which may be a deep neural network model, a firing rate neural network model, a impulse neural network model, and mixtures thereof. In one possible implementation, the impulse neural network model includes, but is not limited to, an upper level brain-like model, an upper level cognitive model. In one implementation scenario, the upper level brain-like model includes, but is not limited to, a whole brain model, a brain region level model, a loop model, and the like. In one possible implementation, the upper-level cognitive model includes, but is not limited to, language mechanisms, memory mechanisms, motor mechanisms, decision mechanisms, and the like.
Fig. 20 shows an internal structural diagram of the smart device in one embodiment. The intelligent device can be a terminal or a server. As shown in fig. 20, the smart device includes a processor, a memory, and a network interface connected by a system bus. The memory includes a nonvolatile storage medium and an internal memory. The non-volatile storage medium of the smart device stores an operating system and may also store a computer program that, when executed by a processor, causes the processor to implement an age identification method. The internal memory may also have stored therein a computer program which, when executed by the processor, causes the processor to perform the age identification method. It will be appreciated by those skilled in the art that the structure shown in fig. 20 is merely a block diagram of a portion of the structure associated with the present application and is not limiting of the smart device to which the present application is applied, and that a particular smart device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a smart device is provided that includes a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the method steps as described above.
In an embodiment, a computer-readable storage medium is proposed, storing a computer program which, when executed by a processor, causes the processor to perform the method steps as above.
Those skilled in the art will appreciate that the processes implementing all or part of the methods of the above embodiments may be implemented by a computer program for instructing relevant hardware, and the program may be stored in a non-volatile computer readable storage medium, and the program may include the processes of the embodiments of the methods as above when executed. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples represent only a few embodiments of the present application, which are described in more detail and are not thereby to be construed as limiting the scope of the present application. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.
Claims (37)
1. A method for port model object organization, comprising:
acquiring at least two port model objects, wherein each port model object comprises at least one port and a model main body, the model main body is associated with data and/or operations, and the ports are used for carrying out information interaction or association with the outside of the port model objects;
and carrying out information interaction or association on at least two port model objects through at least one port of each port model object to form an organization model object.
2. The port model object organization method of claim 1, wherein the ports comprise at least one of an input port, an output port, a reference port, and a connection port.
3. The port model object organization method according to claim 2, wherein the input port is used for receiving input information outside the port model object, the output port is used for outputting output information of the model object, the reference port is used for referencing a mechanism of a variable of another model object to the model object, and the connection port is used for connecting and binding the model object with the other model object.
4. The port model object organizing method according to claim 2, wherein there is a complex correspondence or a fixed correspondence between input information received by the input port and output information output by the output port of the same port model object.
5. The port model object organization method according to claim 1, characterized in that the model body comprises at least one of the organization model objects and/or at least one of the port model objects.
6. The port model object organization method of claim 1, wherein the input ports comprise variable input ports and proxy input ports, and the output ports comprise variable output ports and proxy output ports.
7. The port model object organization method of claim 6, wherein when the model body comprises at least one other organization model object and/or at least one other port model object, the input port comprises a proxy input port and the output port comprises the proxy output port;
the proxy input port is connected with the input port of at least one port model object in the port model object and is used for receiving input information required by the model main body;
the proxy output port is connected with the output port of at least one port model object in the port model object, and is used for acquiring the output information of the model main body and outputting the output information to the outside.
8. The port model object organization method of claim 1, wherein the structure of the organization model object comprises a flat structure, a tree structure, a cascade structure, and a nested structure.
9. The port model object organization method of claim 1, wherein the model body further comprises at least one of a neuron, a synapse, a group of neurons, a group of synapses.
10. The port model object organization method of claim 9, wherein when the model body includes neurons, the port model object is a neuron model object, an input port of the neuron model object is capable of receiving an input of at least one neuron, and an output port of the neuron model object is capable of being connected to at least one neuron.
11. The method of port model object organization according to claim 10, characterized in that said step of information interaction or association of at least two of said port model objects through at least one respective port comprises:
sequentially and unidirectionally connecting the input ports of the neuron model objects and the input ports according to the connection direction designated by the calling instruction to form a unidirectional neural link; or (b)
And connecting an output port of a later one of any two neuron model objects in the plurality of neuron model objects in the unidirectional neural link with an input port of a former one of the two neuron model objects in the order according to the calling instruction to form a neural loop.
12. The method of port model object organization according to claim 10, characterized in that said step of information interaction or association of at least two of said port model objects through at least one respective port comprises:
and correlating at least two neuron port model objects through respective reference ports to share membrane potential information so as to realize electric synapses.
13. The method of port model object organization according to claim 10, characterized in that said step of information interaction or association of at least two of said port model objects through at least one respective port comprises:
And associating the neuron port model object with a non-stress mechanism model through the connection port so as to realize feedback of the neuron non-stress period.
14. The port model object organization method of claim 9, wherein when the model body comprises a synapse, the port model object is a synapse model object, an input port of the synapse model object is capable of receiving an input of at least one neuron, and an output port of the synapse model object is capable of being connected to at least one neuron.
15. The method of organizing port model objects according to claim 14, wherein said step of information interacting or associating at least two of said port model objects through at least one respective port comprises:
the connection port of the synaptic model object is associated with a synaptic plasticity mechanism model for accessing and altering synaptic weights.
16. The port model object organization method of claim 9, wherein when the model body includes a neuron group, the port model object is a neuron group model object, an input port of the neuron group model object is connectable to an input of at least one synapse group, and an output port of the neuron group model object is connectable to the at least one synapse group.
17. The port model object organization method of claim 9, wherein when the model body includes a synapse group, the port model object is a synapse group model object, an input port of the synapse group model object is capable of connecting an input of at least one neuron group, and an output port of the synapse group model object is capable of connecting at least one neuron group.
18. A port model object organization system, comprising:
the system comprises an acquisition module, a storage module and a storage module, wherein the acquisition module is used for acquiring at least two port model objects, each port model object comprises at least one port and a model main body, the model main body is associated with data and/or operation, and the ports are used for carrying out information interaction or association with the outside of the port model;
and the organization module is used for carrying out information interaction or association on at least two port model objects through at least one port of each port model object to form an organization model object.
19. The port model object organization system of claim 18, wherein the ports comprise at least one of an input port, an output port, a reference port, and a connection port.
20. The port model object organizing system of claim 19, wherein said input port is for receiving input information outside of said port model object, said output port is for outputting output information of said model object, said reference port is for referencing a mechanism of a variable of another model object to said model object, and said connection port is for connecting and binding said model object with another model object.
21. The port model object organizing system of claim 19, wherein there is a complex correspondence or a fixed correspondence between input information received by the input port and output information output by the output port of the same port model object.
22. The port model object organization system of claim 18, wherein the model body comprises at least one of the organization model objects and/or at least one of the port model objects.
23. The port model object organization system of claim 18, wherein the input ports comprise variable input ports and proxy input ports, and the output ports comprise variable output ports and proxy output ports.
24. The port model object organization system of claim 23, wherein when the model body comprises at least one other organization model object and/or at least one other port model object, the input port comprises a proxy input port and the output port comprises the proxy output port;
the proxy input port is connected with the input port of at least one port model object in the port model object and is used for receiving input information required by the model main body;
The proxy output port is connected with the output port of at least one port model object in the port model object, and is used for acquiring the output information of the model main body and outputting the output information to the outside.
25. The port model object organization system of claim 18, wherein the structure of the organization model object comprises a flat structure, a tree structure, a cascade structure, and a nested structure.
26. The port model object organization system of claim 18, wherein the model body further comprises at least one of a neuron, a synapse, a group of neurons, a group of synapses.
27. The port model object organizing system of claim 26, wherein when said model body includes neurons, said port model object is a neuron model object, an input port of said neuron model object is capable of receiving input of at least one neuron, and an output port of said neuron model object is capable of being connected to at least one neuron.
28. The port model object organization system of claim 27, wherein the organization module is further to:
sequentially and unidirectionally connecting the input ports of the neuron model objects and the input ports according to the connection direction designated by the calling instruction to form a unidirectional neural link; or (b)
And connecting an output port of a later one of any two neuron model objects in the plurality of neuron model objects in the unidirectional neural link with an input port of a former one of the two neuron model objects in the order according to the calling instruction to form a neural loop.
29. The port model object organization system of claim 27, wherein the organization module is further to:
and correlating at least two neuron port model objects through respective reference ports to share membrane potential information so as to realize electric synapses.
30. The port model object organization system of claim 27, wherein the organization module is further to:
and associating the neuron port model object with a non-stress mechanism model through the connection port so as to realize feedback of the neuron non-stress period.
31. The port model object organization system of claim 26, wherein when the model body comprises a synapse, the port model object is a synapse model object, an input port of the synapse model object is capable of receiving an input of at least one neuron, and an output port of the synapse model object is capable of connecting to at least one neuron.
32. The port model object organization system of claim 31, wherein the organization module is further to:
the connection port of the synaptic model object is associated with a synaptic plasticity mechanism model for accessing and altering synaptic weights.
33. The port model object organizing system of claim 26, wherein when said model body comprises a neuron group, said port model object is a neuron group model object, an input port of said neuron group model object is capable of connecting an input of at least one synapse group, and an output port of said neuron group model object is capable of connecting at least one synapse group.
34. The port model object organizing system of claim 26, wherein when said model body includes a synapse group, said port model object is a synapse group model object having an input port connectable to an input of at least one neuron group and an output port connectable to at least one neuron group.
35. A neural network build platform, comprising: a front end portion, a core portion, and a rear end portion;
The front-end portion is used for acting the outside of the platform to interact with the core portion and/or the back-end portion;
the back end part is used for running a target model;
the core part for describing and/or organizing the object model in response to a request of the front-end part and/or the back-end part, the core part comprising a port module for implementing the method of any of claims 1-17.
36. A smart device comprising a memory and a processor, the memory having a computer program stored therein, the processor being configured to execute the computer program to implement the method of any of claims 1-17.
37. A storage medium storing a computer program capable of being loaded by a processor and executing the method according to any one of claims 1-17.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211067985.0A CN117688969A (en) | 2022-09-02 | 2022-09-02 | Port model object organization method, system, platform, intelligent device and medium |
| PCT/CN2023/116448 WO2024046461A1 (en) | 2022-09-02 | 2023-09-01 | Port model object organization method and system, and platform, intelligent apparatus and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211067985.0A CN117688969A (en) | 2022-09-02 | 2022-09-02 | Port model object organization method, system, platform, intelligent device and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117688969A true CN117688969A (en) | 2024-03-12 |
Family
ID=90100448
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211067985.0A Pending CN117688969A (en) | 2022-09-02 | 2022-09-02 | Port model object organization method, system, platform, intelligent device and medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN117688969A (en) |
| WO (1) | WO2024046461A1 (en) |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| MY138544A (en) * | 2003-06-26 | 2009-06-30 | Neuramatix Sdn Bhd | Neural networks with learning and expression capability |
| US8843425B2 (en) * | 2011-07-29 | 2014-09-23 | International Business Machines Corporation | Hierarchical routing for two-way information flow and structural plasticity in neural networks |
| US8738554B2 (en) * | 2011-09-16 | 2014-05-27 | International Business Machines Corporation | Event-driven universal neural network circuit |
| US8874498B2 (en) * | 2011-09-16 | 2014-10-28 | International Business Machines Corporation | Unsupervised, supervised, and reinforced learning via spiking computation |
| US9020867B2 (en) * | 2012-11-06 | 2015-04-28 | International Business Machines Corporation | Cortical simulator for object-oriented simulation of a neural network |
-
2022
- 2022-09-02 CN CN202211067985.0A patent/CN117688969A/en active Pending
-
2023
- 2023-09-01 WO PCT/CN2023/116448 patent/WO2024046461A1/en unknown
Also Published As
| Publication number | Publication date |
|---|---|
| WO2024046461A1 (en) | 2024-03-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108122032B (en) | Neural network model training method, device, chip and system | |
| US8712939B2 (en) | Tag-based apparatus and methods for neural networks | |
| US9117176B2 (en) | Round-trip engineering apparatus and methods for neural networks | |
| CN107578095B (en) | Neural computing device and processor comprising the computing device | |
| US10210452B2 (en) | High level neuromorphic network description apparatus and methods | |
| Zhang et al. | Poseidon: A system architecture for efficient gpu-based deep learning on multiple machines | |
| CN105956021A (en) | Automated task parallel method suitable for distributed machine learning and system thereof | |
| Aarts et al. | Boltzmann machines and their applications | |
| EP2825974A1 (en) | Tag-based apparatus and methods for neural networks | |
| CN106201651A (en) | The simulator of neuromorphic chip | |
| Gu et al. | A parallel computing platform for training large scale neural networks | |
| JPH087683B2 (en) | How to translate knowledge base and reasoning into procedural source program code | |
| CN111708641B (en) | Memory management method, device, equipment and computer readable storage medium | |
| CN114202068B (en) | Self-learning implementation system for brain-like computing chip | |
| CN111368981B (en) | Method, apparatus, device and storage medium for reducing storage area of synaptic connections | |
| CN113821983A (en) | Engineering design optimization method and device based on proxy model and electronic equipment | |
| CN111831354A (en) | Data precision configuration method, device, chip array, equipment and medium | |
| CN111831359A (en) | Weight precision configuration method, device, equipment and storage medium | |
| Saidi et al. | Automatic parallelization of multi-rate fmi-based co-simulation on multi-core | |
| CN111831356B (en) | Weight precision configuration method, device, equipment and storage medium | |
| CN117688969A (en) | Port model object organization method, system, platform, intelligent device and medium | |
| CN109685203A (en) | Data processing method, device, computer system and storage medium | |
| WO2022057459A1 (en) | Tensorcore-based int4 data type processing method and system, device, and medium | |
| Voloshko et al. | Extended Petri Nets Based Approach for Simulation of Distributed Manufacturing Processes | |
| Egele et al. | Asynchronous distributed bayesian optimization at hpc scale |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |