CN109063319B - Simulation method of biological ecosystem based on neural network - Google Patents

Simulation method of biological ecosystem based on neural network Download PDF

Info

Publication number
CN109063319B
CN109063319B CN201810843096.6A CN201810843096A CN109063319B CN 109063319 B CN109063319 B CN 109063319B CN 201810843096 A CN201810843096 A CN 201810843096A CN 109063319 B CN109063319 B CN 109063319B
Authority
CN
China
Prior art keywords
block
sub
attention mechanism
attribute
biological
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.)
Active
Application number
CN201810843096.6A
Other languages
Chinese (zh)
Other versions
CN109063319A (en
Inventor
冀中
李晟嘉
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tianjin University
Original Assignee
Tianjin University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tianjin University filed Critical Tianjin University
Priority to CN201810843096.6A priority Critical patent/CN109063319B/en
Publication of CN109063319A publication Critical patent/CN109063319A/en
Application granted granted Critical
Publication of CN109063319B publication Critical patent/CN109063319B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • 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
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • 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
    • G06N3/045—Combinations of networks
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
  • Image Analysis (AREA)

Abstract

A simulation method of a biological ecosystem based on a neural network comprises the following steps: establishing a data set of a simulation method of a biological ecosystem based on a neural network, wherein the data set comprises a tag block, a tag block 1, a tag sub-block 2, an initial block, an initial sub-block 1 and an initial sub-block 2; inputting the initial sub-block 1 into the sub-network 1, and simulating the influence of an ecosystem on a living being; inputting the initial block into a subnetwork 2, and simulating the interspecies influence of the creatures; inputting the initial sub-block 2 into a sub-network 3, and simulating the intraspecies influence of the creatures; and averaging the similarity matrix 1, the similarity matrix 2 and the similarity matrix 3 according to corresponding elements to obtain an average similarity matrix, and averaging all elements of the average similarity matrix to obtain a similarity score. The input data of the invention is the original cells of the organism, the training iterative process of the neural network is the gradual evolution process of the organism, and the trained neural network is the ecosystem of the organism.

Description

Simulation method of biological ecosystem based on neural network
Technical Field
The invention relates to a simulation method of a biological ecosystem. In particular to a simulation method of a biological ecosystem based on a neural network.
Background
Machine learning and deep learning rise and develop rapidly with the progress of computer hardware, internet, big data and neural networks. Machine learning and deep learning have good effects in image classification, video abstraction, image recognition, voice recognition, image retrieval, subtitle generation, personalized search and other directions. In these directions and areas, many well-behaved neural network models have emerged, such as imgtet, vggtet, googleNet, LSTM, and CapsNet, among others.
The neural network continuously adjusts the parameters in the network through continuous iterative learning, thereby continuously optimizing the whole neural network. The internal parameters of a neural network are the various linear or non-linear changes, effects and constraints of the network on the input data.
Disclosure of Invention
The invention aims to provide a simulation method of a biological ecosystem based on a neural network, which is continuously close to the biological condition existing in the real world in the training process.
The technical scheme adopted by the invention is as follows: a simulation method of a biological ecosystem based on a neural network comprises the following steps:
1) A data set for establishing a simulation method of a biological ecosystem based on a neural network comprises the following steps: setting the attribute of each living being, coding and dividing the attribute vector to form a tag block, a tag block 1, a tag sub-block 2, an initial block, an initial sub-block 1 and an initial sub-block 2;
2) Inputting the initial sub-block 1 into a sub-network 1, and simulating the influence of an ecosystem on a living being, wherein the sub-network 1 is composed of a CapsNet;
3) Inputting the initial block into a sub-network 2 to simulate the interspecies influence of organisms, wherein the sub-network 2 consists of an LSTM and an attention mechanism module;
4) Inputting the initial sub-block 2 into a sub-network 3 to simulate the intraspecies influence of the creature, wherein the sub-network 3 is composed of an LSTM module and a self-attention mechanism module;
5) And averaging the similarity matrix 1, the similarity matrix 2 and the similarity matrix 3 according to corresponding elements to obtain an average similarity matrix, and averaging all elements of the average similarity matrix to obtain a similarity score.
The attributes of each organism in step 1) comprise: kingdom a, phylum b, class c, order d, family e, genus f, species g, ecosystem type h, number i of crossing ecosystems, population number j, quality k, number l of natural enemy biological species, number m of natural enemy biological species, number n of predatory biological species and total number o of predatory biological species.
Encoding and dividing the attribute vector in the step 1) to form a tag block, a tag block 1, a tag sub-block 2, an initial block, an initial sub-block 1 and an initial sub-block 2, and the method comprises the following steps:
(1) The attributes of each living being are combined into an attribute vector A with 15 attributes 1 The code of the boundary a, the gate b, the class c, the order d, the family e, the genus f and the species g is coded in a binary tree mode in a mode of = a, b, c, d, e, f, g, h, i, j, k, l, m, n and o; will attribute vector A 1 Divided into attribute subvectors A with 9 attributes respectively 2 And an attribute subvector A having 6 attributes 3 Wherein A is 2 = { a, b, c, d, e, f, g, h, i } and a 3 ＝{j,k,l,m,n,o}；
(2) Arranging attribute vectors of m × n creatures into tensors of m × n × 15 to form a tag block; sub-vectors A of m × n organisms 2 And A 3 Tensors arranged into mxnx9 and mxnxnx6 respectively form a tag sub-block 1 and a tag sub-block 2;
(3) Forming an initial block by arranging m × n × 15 tensors by using m × n same attribute vectors of the unicellular creatures; the m × n × 15 tensors of m × n identical unicellular organisms are divided into m × n × 9 tensors and m × n × 6 tensors, and the initial sub-block 1 and the initial sub-block 2 are respectively configured.
The step 2) comprises the following steps: respectively calculating cosine similarity of output results of the CapsNet and the position attribute vectors corresponding to the label sub-blocks 1 according to the types of organisms to obtain a similarity matrix 1; and circularly inputting the output result of the CapsNet into the CapsNet for training and continuously calling back the parameters of the CapsNet through a loss function 1.
The step 3) comprises the following steps:
(1) Inputting the initial block into the LSTM, and inputting the output result of the LSTM into an attention mechanism module, wherein the attention mechanism module comprises an attention mechanism submodule 1 and an attention mechanism submodule 2, the attention mechanism submodule 1 is composed of m × n full connection layers and a Softmax layer, the attention mechanism submodule 2 is composed of m × n attention weight modules, the attention weight modules are composed of m × n-1 weight vectors of each biological attribute vector, and the internal elements of each weight vector are the same;
(2) In the mutual attention machine system submodule 1, various biological attribute vectors are respectively input into a full connection layer and a Softmax layer for calculation, then the calculation result of the mutual attention machine system submodule 1 is input into a mutual attention machine system submodule 2, each biological attribute vector is allocated with m multiplied by n < -1 > weight vectors, in the mutual attention machine system submodule 2, each biological attribute vector is respectively multiplied by corresponding elements of the m multiplied by n < -1 > weight vectors allocated to the organism, m multiplied by n < -1 > calculation results are respectively multiplied by attribute vectors of other organisms, each organism obtains m multiplied by n < -1 > weighted attribute vectors of all other organisms, and the m multiplied by n < -1 > weighted attribute vectors are added to be used as new attribute vectors of the organism;
(3) Calculating the cosine similarity between the output result of the mutual attention mechanism module and the label block to obtain a similarity matrix 2, circularly inputting the output result of the LSTM into the LSTM, training and continuously calling back the parameters of the LSTM and the mutual attention mechanism module through a loss function 2.
The step 4) comprises the following steps:
(1) Inputting the initial sub-block 2 into the LSTM, and inputting an output result of the LSTM into an attention mechanism module, wherein the attention mechanism module comprises an attention mechanism sub-module 1 and an attention mechanism sub-module 2; the self-attention machine sub-module 1 consists of m multiplied by n full connection layers and a Softmax layer, the self-attention machine sub-module 2 consists of m multiplied by n weight vectors, and elements in each weight vector are the same;
(2) In the self-attention machine submodule 1, the attribute vectors of various biological species are respectively input into a full connection layer and a Softmax layer for calculation, then the calculation result of the self-attention machine submodule 1 is input into a self-attention machine submodule 2, in the self-attention machine submodule 2, each biological attribute vector is multiplied by corresponding elements of 1 weight vector allocated to the organism, and the calculation result is multiplied by the attribute vector of the organism to be used as a new attribute vector of the organism;
(3) Calculating the cosine similarity between the output result of the self-attention mechanism module and the label sub-block 2 to obtain a similarity matrix 3, circularly inputting the output result of the LSTM into the LSTM for training, and continuously calling back the parameters of the LSTM and the self-attention mechanism module through a loss function 3.
The invention relates to a simulation method of a biological ecosystem based on a neural network, which is based on parameter adjustment of the neural network, cyclic training, optimization network and various linear or branching influences and limits on input data, and simulates the evolution and development of the biological ecosystem. The input data is the original cells of the living beings, the training iterative process of the neural network is the gradual evolution process of the living beings, and the trained neural network is the ecosystem of the living beings. The invention has the following characteristics:
(1) The novelty is as follows: the simulation algorithm of the biological ecosystem based on the neural network is put forward for the first time, and the evolution process of the biological ecosystem is explored by using the relevant knowledge of the neural network, machine learning and deep learning.
(2) Effectiveness: with continuous iteration and parameter adjustment of the neural network, the initially input unified unicellular organisms are gradually differentiated into different organisms living in different ecosystems, have different qualities, community numbers and other various biological attributes, and are continuously close to the biological conditions existing in the real world in the training process.
(3) The practicability is as follows: deep learning and neural networks are used to model the evolutionary processes of living beings and the processes of the evolving ecosystem. By observing the adjustment and change of the neural network parameters, the processes of mutual influence, mutual restriction and mutual game of organisms and environment, interspecies and intraspecies in training can be analyzed and obtained, and therefore the biological ecosystem can be further known.
Drawings
FIG. 1 is a schematic diagram of a neural network structure of a simulation method of a neural network-based biological ecosystem according to the present invention;
FIG. 2 is a schematic diagram of the structure of the mutual attention mechanism module;
fig. 3 is a schematic diagram of a self-attention mechanism module.
Detailed Description
The simulation algorithm of a neural network-based biological ecosystem according to the present invention is described in detail below with reference to the embodiments and the accompanying drawings.
The invention discloses a simulation algorithm of a biological ecosystem based on a neural network, which aims to recognize, analyze and simulate the development and formation of the biological ecosystem by using the neural network.
As shown in fig. 1, the simulation algorithm of a biological ecosystem based on a neural network of the present invention comprises the following steps:
1) A data set for establishing a simulation algorithm for a neural network-based biological ecosystem, comprising: setting the attribute of each living being, coding and dividing the attribute vector to form a tag block, a tag block 1, a tag sub-block 2, an initial block, an initial sub-block 1 and an initial sub-block 2;
the attributes of each living being comprise: kingdom a, phylum b, class c, order d, family e, genus f, species g, ecosystem type h, crossing ecosystem quantity i, population quantity j, quality k, natural enemy biological species number l, natural enemy biological quantity m, predatory biological species number n and total predatory biological quantity o; wherein, the kingdom a, the phylum b, the class c, the order d, the family e, the genus f and the species g are the grade division and the name of the living beings; ecosystem class h is one of the major ecosystems in which the organism exists, and the number i of spanning ecosystems is the total number of ecosystems in which the organism can live; the population quantity j, the quality k, the natural enemy organism quantity m and the total quantity o of the predatory organisms are accurate to the magnitude order; the number of natural enemy biological species l and the number of prey biological species n are the numbers of common natural enemy and prey biological species.
The encoding and division of the attribute vector form a label block, a label block 1, a label sub-block 2, an initial block, an initial sub-block 1 and an initial sub-block 2, and the method comprises the following steps:
(1) The attributes of each living being are combined into an attribute vector A with 15 attributes 1 The code of the boundary a, the gate b, the class c, the order d, the family e, the genus f and the species g is coded in a binary tree mode in a mode of = a, b, c, d, e, f, g, h, i, j, k, l, m, n and o; will attribute vector A 1 Divided into attribute subvectors A with 9 attributes respectively 2 And an attribute subvector A having 6 attributes 3 Wherein A is 2 = a, b, c, d, e, f, g, h, i and 3 ＝{j,k,l,m,n,o}；
(2) Arranging attribute vectors of m × n creatures into tensors of m × n × 15 to form a tag block; sub-vectors A of m × n organisms 2 And A 3 Tensors arranged to be mxnx9 and mxnxnx6 respectively constitute a tag subblock 1 and a tag subblock 2 respectively; the method specifically comprises the following steps:
respectively calculating cosine similarity of output results of the CapsNet and position attribute vectors corresponding to the label sub-blocks 1 according to the types of organisms to obtain a similarity matrix 1; and circularly inputting the output result of the CapsNet into the CapsNet for training and continuously calling back the parameters of the CapsNet through a loss function 1.
(3) Arranging m × n attribute vectors of the same unicellular organisms into m × n × 15 tensors to form an initial block; the m × n × 15 tensors of m × n identical unicellular organisms are divided into m × n × 9 tensors and m × n × 6 tensors, and the initial sub-block 1 and the initial sub-block 2 are respectively configured.
2) Inputting the initial sub-block 1 into a sub-network 1, and simulating the influence of an ecosystem on a living being, wherein the sub-network 1 is composed of a CapsNet;
3) Inputting the initial block into a sub-network 2, and simulating the interspecies influence of organisms, wherein the sub-network 2 consists of an LSTM and an attention interaction mechanism module; the method specifically comprises the following steps:
(1) Inputting the initial block into the LSTM, and inputting the output result of the LSTM into an attention mechanism module, where the attention mechanism module is shown in fig. 2 and includes an attention mechanism submodule 1 and an attention mechanism submodule 2, where the attention mechanism submodule 1 is composed of m × n full-link layers and a Softmax layer, the attention mechanism submodule 2 is composed of m × n attention weight modules, the attention weight modules are composed of m × n-1 weight vectors of each biological attribute vector, and the internal elements of each weight vector are the same;
(2) In the mutual attention machine system submodule 1, various biological attribute vectors are respectively input into a full connection layer and a Softmax layer for calculation, then the calculation result of the mutual attention machine system submodule 1 is input into a mutual attention machine system submodule 2, each biological attribute vector is allocated with m multiplied by n < -1 > weight vectors, in the mutual attention machine system submodule 2, each biological attribute vector is respectively multiplied by corresponding elements of the m multiplied by n < -1 > weight vectors allocated to the organism, m multiplied by n < -1 > calculation results are respectively multiplied by attribute vectors of other organisms, each organism obtains m multiplied by n < -1 > weighted attribute vectors of all other organisms, and the m multiplied by n < -1 > weighted attribute vectors are added to be used as new attribute vectors of the organism;
(3) Calculating the cosine similarity between the output result of the mutual attention mechanism module and the label block to obtain a similarity matrix 2, circularly inputting the output result of the LSTM into the LSTM, training and continuously calling back the parameters of the LSTM and the mutual attention mechanism module through a loss function 2.
4) Inputting the initial sub-block 2 into a sub-network 3 to simulate the intraspecific influence of the creature, wherein the sub-network 3 consists of an LSTM and an attention mechanism module; the method specifically comprises the following steps:
(1) Inputting the initial sub-block 2 into the LSTM, and inputting an output result of the LSTM into the self-attention mechanism module, which includes a self-attention mechanism sub-module 1 and a self-attention mechanism sub-module 2, as shown in fig. 3; the self-attention machine sub-module 1 consists of m multiplied by n full connection layers and a Softmax layer, the self-attention machine sub-module 2 consists of m multiplied by n weight vectors, and elements in each weight vector are the same;
(2) In the self-attention machine submodule 1, the attribute vectors of various biological species are respectively input into a full connection layer and a Softmax layer for calculation, then the calculation result of the self-attention machine submodule 1 is input into a self-attention machine submodule 2, in the self-attention machine submodule 2, each biological attribute vector is multiplied by corresponding elements of 1 weight vector allocated to the organism, and the calculation result is multiplied by the attribute vector of the organism to be used as a new attribute vector of the organism;
(3) And calculating the cosine similarity between the output result of the self-attention mechanism module and the label sub-block 2 to obtain a similarity matrix 3, circularly inputting the output result of the LSTM into the LSTM for training, and continuously calling back the LSTM and the parameters of the self-attention mechanism module through a loss function 3.
5) And averaging the similarity matrix 1, the similarity matrix 2 and the similarity matrix 3 according to corresponding elements to obtain an average similarity matrix, and averaging all elements of the average similarity matrix to obtain a similarity score.

Claims (3)

1. A simulation method of a biological ecosystem based on a neural network is characterized by comprising the following steps:
1) A data set for establishing a neural network-based simulation method for a biological ecosystem, comprising: setting the attribute of each living being, coding and dividing the attribute vector to form a label block, a label sub-block 1, a label sub-block 2, an initial block, an initial sub-block 1 and an initial sub-block 2;
2) Inputting the initial sub-block 1 into a sub-network 1, and simulating the influence of an ecosystem on a living being, wherein the sub-network 1 is composed of a CapsNet; the method specifically comprises the following steps:
respectively calculating cosine similarity of output results of the CapsNet and the position attribute vectors corresponding to the label sub-blocks 1 according to the types of organisms to obtain a similarity matrix 1; circularly inputting output results of the CapsNet into the CapsNet for training and continuously calling back parameters of the CapsNet through a loss function 1
3) Inputting the initial block into a sub-network 2 to simulate the interspecies influence of organisms, wherein the sub-network 2 consists of an LSTM and an attention mechanism module; the method specifically comprises the following steps:
(1) Inputting the initial block into the LSTM, and inputting the output result of the LSTM into an attention mechanism module, wherein the attention mechanism module comprises an attention mechanism submodule 1 and an attention mechanism submodule 2, the attention mechanism submodule 1 is composed of m × n full connection layers and a Softmax layer, the attention mechanism submodule 2 is composed of m × n attention weight modules, the attention weight modules are composed of m × n-1 weight vectors of each biological attribute vector, and the internal elements of each weight vector are the same;
(2) In the mutual attention machine system submodule 1, various biological attribute vectors are respectively input into a full connection layer and a Softmax layer for calculation, then the calculation result of the mutual attention machine system submodule 1 is input into a mutual attention machine system submodule 2, each biological attribute vector is allocated with m multiplied by n < -1 > weight vectors, in the mutual attention machine system submodule 2, each biological attribute vector is respectively multiplied by corresponding elements of the m multiplied by n < -1 > weight vectors allocated to the organism, m multiplied by n < -1 > calculation results are respectively multiplied by attribute vectors of other organisms, each organism obtains m multiplied by n < -1 > weighted attribute vectors of all other organisms, and the m multiplied by n < -1 > weighted attribute vectors are added to be used as new attribute vectors of the organism;
(3) Calculating the cosine similarity between the output result of the mutual attention mechanism module and the label block to obtain a similarity matrix 2, circularly inputting the output result of the LSTM into the LSTM, training and continuously calling back the parameters of the LSTM and the mutual attention mechanism module through a loss function 2;
4) Inputting the initial sub-block 2 into a sub-network 3 to simulate the intraspecies influence of the creature, wherein the sub-network 3 is composed of an LSTM module and a self-attention mechanism module; the method specifically comprises the following steps:
(1) Inputting the initial sub-block 2 into the LSTM, and inputting an output result of the LSTM into a self-attention mechanism module, wherein the self-attention mechanism module comprises a self-attention mechanism sub-module 1 and a self-attention mechanism sub-module 2; the self-attention mechanism submodule 1 consists of m multiplied by n full connection layers and a Softmax layer, the self-attention mechanism submodule 2 consists of m multiplied by n weight vectors, and elements in each weight vector are the same;
(2) In the self-attention mechanism submodule 1, the attribute vectors of various biological species are respectively input into a full connection layer and a Softmax layer for calculation, then the calculation result of the self-attention mechanism submodule 1 is input into a self-attention mechanism submodule 2, in the self-attention mechanism submodule 2, each biological attribute vector is multiplied by corresponding elements of 1 weight vector allocated to the organism, and the calculation result is multiplied by the attribute vector of the organism to be used as a new attribute vector of the organism;
(3) Calculating the cosine similarity between the output result of the self-attention mechanism module and the label sub-block 2 to obtain a similarity matrix 3, circularly inputting the output result of the LSTM into the LSTM for training, and continuously calling back the parameters of the LSTM and the self-attention mechanism module through a loss function 3;
5) And averaging the similarity matrix 1, the similarity matrix 2 and the similarity matrix 3 according to corresponding elements to obtain an average similarity matrix, and averaging all elements of the average similarity matrix to obtain a similarity score.
2. The simulation method of a biological ecosystem based on a neural network according to claim 1, wherein the attribute of each living being in step 1) includes: kingdom a, phylum b, class c, order d, family e, genus f, species g, ecosystem type h, number i of crossing ecosystems, population number j, quality k, number l of natural enemy biological species, number m of natural enemy biological species, number n of predatory biological species and total number o of predatory organisms.
3. The simulation method of a biological ecosystem based on a neural network according to claim 1, wherein the encoding and dividing of the attribute vector of step 1) forms a tag block, a tag sub-block 1, a tag sub-block 2, an initial block, an initial sub-block 1 and an initial sub-block 2, and comprises:
(1) The attributes of each living being are combined into an attribute vector A with 15 attributes 1 The code of the boundary a, the gate b, the class c, the order d, the family e, the genus f and the species g is coded in a binary tree mode in a mode of = a, b, c, d, e, f, g, h, i, j, k, l, m, n and o; will attribute vector A 1 Divided into attribute subvectors A with 9 attributes respectively 2 And an attribute subvector A having 6 attributes 3 Wherein A is 2 = { a, b, c, d, e, f, g, h, i } and a 3 ＝{j,k,l,m,n,o}；
(2) Arranging attribute vectors of m × n creatures into tensors of m × n × 15 to form a tag block; sub-vectors A of m × n organisms 2 And A 3 Tensors arranged into mxnx9 and mxnxnx6 respectively form a tag sub-block 1 and a tag sub-block 2;
(3) Forming an initial block by arranging m × n × 15 tensors by using m × n same attribute vectors of the unicellular creatures; the m × n × 15 tensors of m × n identical single-cell organisms are divided into m × n × 9 tensors and m × n × 6 tensors, and the initial sub-block 1 and the initial sub-block 2 are respectively configured.
CN201810843096.6A 2018-07-27 2018-07-27 Simulation method of biological ecosystem based on neural network Active CN109063319B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810843096.6A CN109063319B (en) 2018-07-27 2018-07-27 Simulation method of biological ecosystem based on neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810843096.6A CN109063319B (en) 2018-07-27 2018-07-27 Simulation method of biological ecosystem based on neural network

Publications (2)

Publication Number Publication Date
CN109063319A CN109063319A (en) 2018-12-21
CN109063319B true CN109063319B (en) 2023-04-07

Family

ID=64836826

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810843096.6A Active CN109063319B (en) 2018-07-27 2018-07-27 Simulation method of biological ecosystem based on neural network

Country Status (1)

Country Link
CN (1) CN109063319B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117407751A (en) * 2023-10-20 2024-01-16 深圳市华美绿生态环境集团有限公司 Data acquisition system for forest biodiversity statistics

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102496077A (en) * 2011-12-06 2012-06-13 北京林业大学 System and method for predicting forest pest disaster
CN107392224A (en) * 2017-06-12 2017-11-24 天津科技大学 A kind of crop disease recognizer based on triple channel convolutional neural networks
CN107886073A (en) * 2017-11-10 2018-04-06 重庆邮电大学 A kind of more attribute recognition approaches of fine granularity vehicle based on convolutional neural networks
CN108171322A (en) * 2017-12-25 2018-06-15 青岛理工大学 A kind of Learning Algorithm based on particle group optimizing
CN108319928A (en) * 2018-02-28 2018-07-24 天津大学 A kind of deep learning model and application based on Multi-objective PSO optimization

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140029196A1 (en) * 2011-04-15 2014-01-30 Kevin Smith System for balanced power and thermal management of mission critical environments

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102496077A (en) * 2011-12-06 2012-06-13 北京林业大学 System and method for predicting forest pest disaster
CN107392224A (en) * 2017-06-12 2017-11-24 天津科技大学 A kind of crop disease recognizer based on triple channel convolutional neural networks
CN107886073A (en) * 2017-11-10 2018-04-06 重庆邮电大学 A kind of more attribute recognition approaches of fine granularity vehicle based on convolutional neural networks
CN108171322A (en) * 2017-12-25 2018-06-15 青岛理工大学 A kind of Learning Algorithm based on particle group optimizing
CN108319928A (en) * 2018-02-28 2018-07-24 天津大学 A kind of deep learning model and application based on Multi-objective PSO optimization

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于GA-BP人工神经网络的樟树林土壤呼吸对施氮响应的研究;张力等;《生态学报》;20170831;全文 *

Also Published As

Publication number Publication date
CN109063319A (en) 2018-12-21

Similar Documents

Publication Publication Date Title
CN110070183B (en) Neural network model training method and device for weakly labeled data
Han et al. A survey on metaheuristic optimization for random single-hidden layer feedforward neural network
Sau et al. Deep model compression: Distilling knowledge from noisy teachers
Assunção et al. Evolving the topology of large scale deep neural networks
CN106095872A (en) Answer sort method and device for Intelligent Answer System
CN110263227A (en) Clique based on figure neural network finds method and system
Jiang et al. Design of structural modular neural networks with genetic algorithm
Cheng et al. Sign: Spatial-information incorporated generative network for generalized zero-shot semantic segmentation
CN111681178B (en) Knowledge distillation-based image defogging method
CN111651762A (en) Convolutional neural network-based PE (provider edge) malicious software detection method
Van Hieu et al. Automatic plant image identification of Vietnamese species using deep learning models
CN114817568B (en) Knowledge hypergraph link prediction method combining attention mechanism and convolutional neural network
CN112232511A (en) Automatic compression method and platform for pre-training language model for multiple tasks
Zhang et al. Evolving neural network classifiers and feature subset using artificial fish swarm
CN109086463B (en) Question-answering community label recommendation method based on regional convolutional neural network
CN112578089A (en) Air pollutant concentration prediction method based on improved TCN
CN109063319B (en) Simulation method of biological ecosystem based on neural network
Cho et al. Genetic evolution processing of data structures for image classification
Chang et al. Semi-supervised learning using generative adversarial networks
CN116958712A (en) Image generation method, system, medium and device based on prior probability distribution
Lal et al. The r-matrix net
CN115168864A (en) Intelligent cross contract vulnerability detection method based on feature cross
CN114021721A (en) Space-time attention knowledge tracking method based on learning process segment
CN114139674A (en) Behavior cloning method, electronic device, storage medium, and program product
CN116541593B (en) Course recommendation method based on hypergraph neural network

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
GR01 Patent grant
GR01 Patent grant