A maximum flow scheduling method and device based on a time-varying synchronous flow neural network

Abstract

The application discloses a maximum flow scheduling method and device based on a time-varying synchronous flow neural network, which comprises the following steps: processing a time-varying synchronous flow neuron to obtain a neuron hierarchical structure; the neuron hierarchical structure comprises a flow input layer, a flow receiving layer, a neuron state layer, a flow generating layer and a flow sending layer; processing the neuron hierarchical structure to obtain a time-varying synchronous flow neural network; initializing the time-varying synchronous flow neuron and the time-varying synchronous flow neural network to obtain initialized time-varying synchronous flow neuron parameters and initialized time-varying synchronous flow neural network parameters; processing the time-varying synchronous flow neural network according to the initialized time-varying synchronous flow neuron parameters and the initialized time-varying synchronous flow neural network parameters to obtain a maximum flow scheduling result. The method is based on the time shortest path principle, and ensures that the globally optimal flow distribution is realized under the condition of meeting the time-varying capacity constraint.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of flow scheduling, in particular to a maximum flow scheduling method and device based on a time-varying synchronous flow neural network. BACKGROUND

[0002] The maximum flow scheduling problem in time-varying networks needs to deal with the complex scenario that the network topology and edge capacity dynamically change over time. The current mainstream solution is still limited to the maximum flow algorithm (such as Ford-Fulkerson algorithm) of static network or the non-synchronous time-varying network flow method, and these methods have two key defects: first, the traditional static algorithm cannot adapt to the time-varying characteristics of edge capacity (such as the bandwidth of communication link fluctuating with channel quality); second, the existing dynamic method ignores the timing dependency of flow transmission, resulting in serious resource coordination problems. In practical applications, time-varying synchronization characteristics are crucial: in cloud computing resource scheduling, the server processing capacity dynamically changes with the load, while the strict timing execution of distributed tasks needs to be guaranteed; in industrial Internet of Things, the sensor data transmission rate is affected by environmental interference, and the data collection rhythm of multiple devices needs to be coordinated; in network slicing, bandwidth allocation needs to adapt to channel changes in real time, while ensuring that data packets arrive in order to avoid recombination delay. Especially in critical infrastructure such as smart grids, the capacity of power transmission lines fluctuates with temperature changes, and more precise synchronous power scheduling is needed to maintain system stability. Artificial intelligence technology, especially recurrent neural networks, provides a new way to solve the time-varying constraints and synchronization problems simultaneously due to its memory characteristics and dynamic system modeling capabilities. Therefore, developing a maximum flow scheduling method based on a time-varying synchronous flow neural network can not only meet the real-time requirements of dynamic networks, but also guarantee the coordination of flow transmission, which has important theoretical value and engineering significance for improving the operation efficiency of critical infrastructure. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a maximum flow scheduling method and device based on a time-varying synchronous flow neural network, which overcomes the technical defects of traditional methods that are difficult to adapt to network dynamic changes and synchronous flow scheduling.

[0004] To solve the above technical problems, a first aspect of an embodiment of the present application discloses a maximum flow scheduling method based on a time-varying synchronous flow neural network, the method comprising:

[0005] S1, processing a time-varying synchronous flow neuron to obtain a neuron hierarchy structure; the neuron hierarchy structure comprises a flow input layer, a flow receiving layer, a neuron state layer, a flow generating layer and a flow sending layer;

[0006] S2, processing the neuron hierarchy structure to obtain a time-varying synchronous flow neural network;

[0007] S3, initializing the time-varying synchronous flow neuron and the time-varying synchronous flow neural network to obtain initialized time-varying synchronous flow neuron parameters and initialized time-varying synchronous flow neural network parameters;

[0008] The initialized time-varying synchronous flow neuron parameters include a predecessor node pre i number, a successor node next i number and a target node number.

[0009] The time-varying synchronous flow neural network parameters include a timer time T, an iteration step Δt and attribute information.

[0010] S4, processing the time-varying synchronous flow neural network according to the initialized time-varying synchronous flow neuron parameters and the initialized time-varying synchronous flow neural network parameters to obtain a maximum flow scheduling result.

[0011] As an optional implementation, in the first aspect of the embodiment of the present application, the processing of the time-varying synchronous flow neuron to obtain a neuron hierarchy structure comprises:

[0012] S11, processing the time-varying synchronous flow neuron to obtain a flow input layer, the flow input layer including a flow identifier h i , when h i is not empty, there is flow arriving at the neuron i.

[0013] S12, constructing a flow receiving layer of the time-varying synchronous flow neuron; the flow receiving layer is used to receive flow information T g,i output by the flow input layer. Wherein, T g,i is used to store flow from neuron g arriving at neuron i sent by the starting node at time T, is used to store the arrival time of the flow.

[0014] S13, constructing a neuron state layer of the time-varying synchronous flow neuron; the neuron state layer includes an activation state target set pre i and flow record R i .

[0015] S14, constructing a flow generating layer of the time-varying synchronous flow neuron; the flow generating layer includes a tuple

[0016] S15, constructing a flow sending layer of the time-varying synchronous flow neuron; the flow sending layer sends flow according to a successor node set next i to send the flow generated by the flow generating layer to the successor node.

[0017] As an optional implementation, in the first aspect of the embodiment of the present application, the processing of the neuron hierarchy structure to obtain the time-varying synchronous flow neural network comprises:

[0018] S21, obtaining node structure information of a target network;

[0019] S22, processing the node structure information of the target network to obtain a neuron cluster of the time-varying synchronous flow neural network;

[0020] The neuron cluster of the time-varying synchronous flow neural network comprises N time-varying synchronous flow neurons;

[0021] S23, processing the neuron cluster of the time-varying synchronous flow neural network to obtain the time-varying synchronous flow neural network.

[0022] As an optional implementation, in the first aspect of the embodiment of the present application, the initialization of the time-varying synchronous flow neuron and the time-varying synchronous flow neural network to obtain the initialized time-varying synchronous flow neuron parameter and the initialized time-varying synchronous flow neural network parameter comprises:

[0023] S31, initializing a timer time T of the time-varying synchronous flow neural network;

[0024] S32, initializing an iteration step Δt of the time-varying synchronous flow neural network;

[0025] S33, initializing attribute information of the time-varying synchronous flow neural network;

[0026] S34, initializing a number of predecessor nodes pre i of each time-varying synchronous flow neuron, a number of successor nodes next i and a number of target nodes.

[0027] As an optional implementation, in the first aspect of the embodiment of the present application, the processing of the time-varying synchronous flow neural network according to the initialized time-varying synchronous flow neuron parameter and the initialized time-varying synchronous flow neural network parameter to obtain the maximum flow scheduling result comprises:

[0028] S41, updating a state of the time-varying synchronous flow neuron of the time-varying synchronous flow neural network to obtain state update information of the time-varying synchronous flow neuron;

[0029] S42, processing the time-varying synchronous flow neural network according to the state update information to obtain a convergent time-varying maximum flow constraint equation;

[0030] S43, solving the convergent time-varying maximum flow constraint equation to obtain the maximum flow scheduling result.

[0031] As an optional implementation, in the first aspect of the embodiment of the present application, the updating of the state of the time-varying synchronous flow neuron of the time-varying synchronous flow neural network comprises:

[0032] S411, for any time-varying synchronous flow neuron, traversing all the predecessor neurons thereof, inquiring whether there is traffic arriving at the current time-varying synchronous flow neuron at the current time, if there is traffic arriving, receiving the traffic through the flow input layer and the flow receiving layer, and recording in the neuron state layer;

[0033] S412, when the neuron state layer of the neuron k detects that the traffic sent at the time t satisfies the activation state , then the flow generation layer of the neuron k will generate new traffic and send the traffic to the successor node next through the flow sending layer of the neuron k i , obtaining the state updating information of the time-varying synchronous flow neuron.

[0034] As an optional implementation, in the first aspect of the embodiment of the present application, the convergence time-varying maximum flow constraint equation expression is:

[0035] max f

[0036] s.t.A T x=d,

[0037] 0≤x≤u T

[0038] wherein x=(x1,x2,…,x m ) T ∈R m×1 is the flow of each arc in the graph, limited by the upper limit of the flow u T =(u1,u2,…,u m ) T ∈R m×1 , A is an activation state matrix, f represents the maximum flow, R m×1 is a real number set of m×1, T represents transposition, and m is the number of elements in x.

[0039] The second aspect of the embodiment of the present application discloses a maximum flow scheduling device based on a time-varying synchronous flow neural network, the device comprises:

[0040] a neuron hierarchical structure construction module, configured to process the time-varying synchronous flow neuron, and obtain a neuron hierarchical structure; the neuron hierarchical structure comprises a flow input layer, a flow receiving layer, a neuron state layer, a flow generation layer and a flow sending layer;

[0041] A time-varying synchronous flow neural network construction module is used to process the neuron hierarchical structure to obtain a time-varying synchronous flow neural network;

[0042] The parameter initialization module is used to initialize the time-varying synchronous flow neuron and the time-varying synchronous flow neural network, and obtain the initialized time-varying synchronous flow neuron parameters and the initialized time-varying synchronous flow neural network parameters;

[0043] The initialization parameters of the time-varying synchronous stream neuron include the predecessor node pre i Quantity, successor node next i Quantity and target node quantity;

[0044] The parameters of the time-varying synchronous flow neural network include timer time T, iteration step size Δt, and attribute information;

[0045] The traffic scheduling module is used to process the time-varying synchronous flow neural network according to the initial time-varying synchronous flow neuron parameters and the initial time-varying synchronous flow neural network parameters to obtain the maximum traffic scheduling result.

[0046] As an optional implementation, in the second aspect of the present invention, the processing of the time-varying synchronous stream neurons to obtain a neuronal hierarchical structure includes:

[0047] S11, Process the time-varying synchronous stream neuron to obtain the stream input layer, which includes a flow identifier h. i When h i If the value is not empty, then there is flow reaching neuron i;

[0048] S12, Construct the flow receiving layer of the time-varying synchronous flow neuron; the flow receiving layer is used to receive the flow information T output by the flow input layer. g,i and Among them, T g,i Used to store the traffic sent from neuron g to neuron i at time T, starting from the node. Used to store the arrival time of traffic;

[0049] S13, Construct the neuronal state layer of the time-varying synchronous flow neuron; the neuronal state layer includes activation states. target set pre i and traffic records R i constitute;

[0050] S14, Construct a flow generation layer for time-varying synchronous flow neurons; the flow generation layer includes tuples.

[0051] S15, construct the streaming layer of the time-varying synchronous stream neuron; the streaming layer is based on the successor node set nexti Send traffic, send the traffic generated by the flow generation layer to the subsequent node.

[0052] As an optional implementation, in the second aspect of the embodiment of the present application, the processing of the neuron hierarchy to obtain the time-varying synchronous flow neural network comprises:

[0053] S21, obtaining node structure information of a target network;

[0054] S22, processing the node structure information of the target network to obtain a neuron cluster of the time-varying synchronous flow neural network;

[0055] The neuron cluster of the time-varying synchronous flow neural network comprises N time-varying synchronous flow neurons;

[0056] S23, processing the neuron cluster of the time-varying synchronous flow neural network to obtain the time-varying synchronous flow neural network.

[0057] As an optional implementation, in the second aspect of the embodiment of the present application, the initialization of the time-varying synchronous flow neuron and the time-varying synchronous flow neural network to obtain the initialized time-varying synchronous flow neuron parameter and the initialized time-varying synchronous flow neural network parameter comprises:

[0058] S31, initializing a timer time T of the time-varying synchronous flow neural network;

[0059] S32, initializing an iteration step Δt of the time-varying synchronous flow neural network;

[0060] S33, initializing attribute information of the time-varying synchronous flow neural network;

[0061] S34, initializing a number of predecessor nodes pre i of each time-varying synchronous flow neuron, a number of successor nodes next i and a number of target nodes.

[0062] As an optional implementation, in the second aspect of the embodiment of the present application, the processing of the time-varying synchronous flow neural network according to the initialized time-varying synchronous flow neuron parameter and the initialized time-varying synchronous flow neural network parameter to obtain the maximum flow scheduling result comprises:

[0063] S41, updating a state of the time-varying synchronous flow neuron of the time-varying synchronous flow neural network to obtain state update information of the time-varying synchronous flow neuron;

[0064] S42, processing the time-varying synchronous flow neural network according to the state update information to obtain a converged time-varying maximum flow constraint equation;

[0065] S43, solving the convergent time-varying maximum flow constraint equation to obtain a maximum flow scheduling result.

[0066] As an optional implementation, in the second aspect of the embodiment of the application, the updating of the state of the time-varying synchronous flow neuron of the time-varying synchronous flow neural network to obtain state update information of the time-varying synchronous flow neuron comprises:

[0067] S411, for any time-varying synchronous flow neuron, traversing all predecessor neurons thereof to query whether there is flow arriving at the current time-varying synchronous flow neuron at the current time, and if there is flow arriving, receiving the flow through the flow input layer and the flow receiving layer and recording in the neuron state layer;

[0068] S412, when the state layer of the neuron k detects that the flow sent at the time t satisfies the activation state , then the flow generation layer of the neuron k will generate new flow and send the flow to the successor node next through the flow sending layer of the neuron k. i , to obtain the state update information of the time-varying synchronous flow neuron.

[0069] As an optional implementation, in the second aspect of the embodiment of the application, the convergent time-varying maximum flow constraint equation is expressed as:

[0070] max f

[0071] s.t.A T x=d,

[0072] 0≤x≤u T

[0073] wherein x=(x1,x2,…,x m ) T ∈R m×1 is the flow of each arc in the graph, limited by the upper limit u T =(u1,u2,…,u m ) T ∈R m×1 , A is an activation state matrix, f represents the maximum flow, R m×1 is an m×1 real number set, T represents transposition, and m is the number of elements in x.

[0074] The third aspect of the application discloses another maximum flow scheduling device based on a time-varying synchronous flow neural network, which comprises:

[0075] a memory storing executable program codes;

[0076] a processor coupled with the memory;

[0077] The processor invokes the executable program code stored in the memory to execute part or all of the steps of the maximum flow scheduling method based on the time-varying synchronous flow neural network according to the first aspect of the present application.

[0078] The fourth aspect of the present application discloses a computer storage medium, which stores computer instructions, and when the computer instructions are invoked, part or all of the steps of the maximum flow scheduling method based on the time-varying synchronous flow neural network according to the first aspect of the present application are executed.

[0079] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0080] The present application discloses a maximum flow scheduling method based on a time-varying synchronous flow neural network, which is based on the shortest path principle, ensures the globally optimal flow distribution under the condition of meeting the time-varying capacity constraint, can effectively handle the flow scheduling problem in a dynamic network environment, and can obtain a theoretically optimal solution. BRIEF DESCRIPTION OF DRAWINGS

[0081] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0082] Figure 1 is a flowchart of a maximum flow scheduling method based on a time-varying synchronous flow neural network disclosed by the embodiments of the present application;

[0083] Figure 2 is another workflow diagram for solving the maximum flow scheduling problem disclosed by the embodiments of the present application;

[0084] Figure 3 is a network structure diagram of a time-varying network disclosed by the embodiments of the present application;

[0085] Figure 4 is a neuron structure diagram of a time-varying synchronous flow neural network disclosed by the embodiments of the present application;

[0086] Figure 5 is a structure diagram of a time-varying synchronous flow neural network disclosed by the embodiments of the present application.

[0087] Figure 6 is a structure diagram of a maximum flow scheduling device based on a time-varying synchronous flow neural network disclosed by the embodiments of the present application;

[0088] Figure 7is a structural schematic view of another maximum flow scheduling device based on time-varying synchronous flow neural network disclosed by the embodiment of the present application. DETAILED DESCRIPTION

[0089] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0090] The terms "first", "second", and the like in the specification and claims of the present application and the above-described drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product, or equipment including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product, or equipment.

[0091] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearance of the phrase in various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily mutually exclusive of other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0092] The present application discloses a maximum flow scheduling method and device based on time-varying synchronous flow neural network. The method comprises processing time-varying synchronous flow neurons to obtain a neuron hierarchical structure; the neuron hierarchical structure comprises a flow input layer, a flow receiving layer, a neuron state layer, a flow generating layer, and a flow sending layer; processing the neuron hierarchical structure to obtain a time-varying synchronous flow neural network; initializing the time-varying synchronous flow neurons and the time-varying synchronous flow neural network to obtain initialized time-varying synchronous flow neuron parameters and initialized time-varying synchronous flow neural network parameters; processing the time-varying synchronous flow neural network according to the initialized time-varying synchronous flow neuron parameters and the initialized time-varying synchronous flow neural network parameters to obtain a maximum flow scheduling result. The method is based on the shortest path principle to ensure global optimal flow distribution under the condition of meeting time-varying capacity constraints, can effectively process flow scheduling problems in dynamic network environments, and obtain a theoretically optimal solution. The following will be described in detail.

[0093] Embodiment one

[0094] Please refer to Figure 1 , Figure 1 is a flowchart of a maximum flow scheduling method based on a time-varying synchronous flow neural network according to an embodiment of the present application. Wherein, Figure 1 The time-varying synchronous flow neural network-based maximum flow scheduling method described is applied to the technical field of flow scheduling, and embodiments of the present application are not limited. As Figure 1 shown, the time-varying synchronous flow neural network-based maximum flow scheduling method can include the following operations:

[0095] S1, processing the time-varying synchronous flow neuron to obtain a neuron hierarchy; the neuron hierarchy includes a flow input layer, a flow receiving layer, a neuron state layer, a flow generation layer and a flow sending layer;

[0096] S2, processing the neuron hierarchy to obtain a time-varying synchronous flow neural network;

[0097] S3, initializing the time-varying synchronous flow neuron and the time-varying synchronous flow neural network to obtain initialized time-varying synchronous flow neuron parameters and initialized time-varying synchronous flow neural network parameters;

[0098] The initialized time-varying synchronous flow neuron parameters include the number of predecessor nodes pre i , the number of successor nodes next i and the number of target nodes;

[0099] The time-varying synchronous flow neural network parameters include the timer time T, the iteration step Δt and the attribute information;

[0100] S4, processing the time-varying synchronous flow neural network according to the initialized time-varying synchronous flow neuron parameters and the initialized time-varying synchronous flow neural network parameters to obtain a maximum flow scheduling result.

[0101] Optionally, the processing of the time-varying synchronous flow neuron to obtain the neuron hierarchy includes:

[0102] S11, processing the time-varying synchronous flow neuron to obtain a flow input layer, the flow input layer including flow identification h i , when h i is not empty, there is flow arriving at neuron i;

[0103] S12, constructing a flow receiving layer of the time-varying synchronous flow neuron; the flow receiving layer is used to receive flow information T g,i and wherein, T g,ia flow record R for storing the arrival time of the flow, a flow record R for storing the arrival time of the flow;

[0104] S13, constructing a neuron state layer of the time-varying synchronous flow neuron; the neuron state layer comprises an activation state a target set pre i and a flow record R i comprises;

[0105] S14, constructing a flow generation layer of the time-varying synchronous flow neuron; the flow generation layer comprises a tuple

[0106] S15, constructing a flow sending layer of the time-varying synchronous flow neuron; the flow sending layer sends the flow generated by the flow generation layer to a successor node set next i according to the successor node set next.

[0107] Optionally, the processing of the neuron hierarchy to obtain the time-varying synchronous flow neural network comprises:

[0108] S21, obtaining node structure information of a target network;

[0109] S22, processing the node structure information of the target network to obtain a neuron cluster of the time-varying synchronous flow neural network;

[0110] The neuron cluster of the time-varying synchronous flow neural network comprises N time-varying synchronous flow neurons;

[0111] S23, processing the neuron cluster of the time-varying synchronous flow neural network to obtain the time-varying synchronous flow neural network.

[0112] Optionally, the initialization of the time-varying synchronous flow neuron and the time-varying synchronous flow neural network to obtain initialized time-varying synchronous flow neuron parameters and initialized time-varying synchronous flow neural network parameters comprises:

[0113] S31, initializing a timer time T of the time-varying synchronous flow neural network;

[0114] S32, initializing an iteration step Δt of the time-varying synchronous flow neural network;

[0115] S33, initializing attribute information of the time-varying synchronous flow neural network;

[0116] S34, initializing a number of predecessor nodes pre i of each time-varying synchronous flow neuron, a number of successor nodes next i and a number of target nodes.

[0117] Optionally, the processing of the time-varying synchronous flow neural network according to the initialized time-varying synchronous flow neuron parameters and the initialized time-varying synchronous flow neural network parameters to obtain the maximum flow scheduling result comprises:

[0118] S41, updating the state of the time-varying synchronous flow neuron of the time-varying synchronous flow neural network to obtain state update information of the time-varying synchronous flow neuron;

[0119] S42, processing the time-varying synchronous flow neural network according to the state update information to obtain a converged time-varying maximum flow constraint equation;

[0120] S43, solving the converged time-varying maximum flow constraint equation to obtain the maximum flow scheduling result.

[0121] Optionally, the updating of the state of the time-varying synchronous flow neuron of the time-varying synchronous flow neural network to obtain the state update information of the time-varying synchronous flow neuron comprises:

[0122] S411, for any time-varying synchronous flow neuron, traversing all predecessor neurons thereof to query whether there is flow arriving at the current time-varying synchronous flow neuron at the current time, and if there is flow arriving, receiving the flow through the flow input layer and the flow receiving layer and recording in the neuron state layer;

[0123] S412, when the state layer of the neuron k detects that the flow sent at the time t satisfies the activation state , then the flow generation layer of the neuron k will generate new flow and send the flow to the successor node next through the flow sending layer of the neuron k i to obtain the state update information of the time-varying synchronous flow neuron.

[0124] Optionally, the converged time-varying maximum flow constraint equation is expressed as:

[0125] max f

[0126] s.t.A T x=d,

[0127] 0≤x≤u T

[0128] wherein x=(x1, x2, …, x m ) T ∈R m×1 is the flow of each arc in the graph, which is subject to the upper limit of the flow u T =(u1, u2, …, u m ) T ∈R m×1subject to A x ≤ u, x ≥ 0, where A is the activation status matrix, f represents the maximum flow, R m×1 is a real m x 1 set, T denotes transpose, and m is the number of elements in x.

[0129] The new convergent time-varying maximum flow constraint equation is constructed as:

[0130] Let e n = (0, 0, 0,..., 1) T , e1= (1, 0, 0,..., 0) T ∈ R m , then (e n - e1) f = -d. The equation is equivalent to:

[0131] max f

[0132] s.t. (e n - e1) f + Ax = 0,

[0133] 0 ≤ x ≤ u T

[0134] where, Let x m+1 = f, then the equation is equivalent to:

[0135] max f

[0136] s.t. (e n - e1) x m+1 + Ax = 0,

[0137]

[0138] Let X = (x1, x2,..., x m , x m+1 ) T ∈ R (m+1)×1 ,

[0139] Let q = (-1, 0, 0,..., 1) T ∈ R n×1 , B = [A | q], then BX = (e n - e1) x m+1 + Ax

[0140] Let c = (0, 0,..., 1) T ∈ R (m+1)×1 , then c T X = f

[0141] Therefore, the equation is equivalent to

[0142] max cT X

[0143] s.t.BX=0,

[0144] 0≤X≤U T

[0145] According to the new time-varying maximum flow constraint equation, the KKT equation is constructed as follows:

[0146] According to the above equation, we get

[0147] L(X,y,h)=c T X+y T (BX)+h T (X-U T )

[0148] Therefore, the following KKT conditions can be obtained:

[0149] (-c T +B T y+h)X=0,-BX=0

[0150] -c T +B T y+h≥0,-(X-U T )≥0

[0151] -h T (X-U T )≥0,X,h≥0,y is free

[0152] Solving the KKT conditions, we get:

[0153] Let

[0154]

[0155] is equivalent to:

[0156]

[0157] When F(w) = 0, X in w is the optimal flow on each edge.

[0158] Introducing the projection where That is, when X in w is the optimal flow on each edge. This is because when F(w) = 0, while ensuring that the first 2(m+1) terms in w are non-negative.

[0159] Introducing a dynamic model to solve

[0160]

[0161] where 0 < λ < 1,

[0162]

[0163] Convergence occurs when the error e(w, β) = 0 or tends to 0 infinitely. At this time, X in w is the optimal flow on each edge, and the maximum flow is x in X m+1 .

[0164] Example Two

[0165] The embodiment is a maximum flow scheduling method based on a time-varying synchronous flow neural network. The overall flow of solving the maximum flow scheduling problem is as shown in Figure 2 . The maximum flow scheduling of a time-varying network of four nodes is as shown in Figure 3 . The travel time function of each edge in the network is as shown in Table 1. The specific implementation scheme includes the following contents and steps:

[0166] Table 1 Travel time function of each edge in the network

[0167]

[0168] Step 1, design a time-varying synchronous flow neuron, wherein each neuron is composed of a flow input layer, a flow receiving layer, a neuron state layer, a flow generating layer, and a flow sending layer. The structure of the neuron is as shown in Figure 4 .

[0169] Step 2, construct a time-varying synchronous flow neural network, and initialize parameters of the time-varying synchronous flow neural network. The specific steps are as follows:

[0170] Step 2.1, initialize the nodes in the time-varying network as time-varying synchronous flow neurons;

[0171] Step 2.2, construct a time-varying synchronous flow neural network model according to the topological structure of the time-varying network. The TVSFNN model is composed of four time-dependent wave neurons and four edge neural network models. Meanwhile, the starting neuron a and the target neuron d are set. The structure of the time-varying synchronous flow neural network is as shown in Figure 5 .

[0172] Step 3, initialize the parameters of the time-varying synchronous flow neural network as described in the initialization of the time-varying synchronous flow neuron parameters. The specific steps are as follows:

[0173] Step 3.1, initialize the timer of the time-varying synchronous flow neural network as T = 0;

[0174] Step 3.2, initialize the iteration step of the time-varying synchronous flow neural network as Δt = 1;

[0175] Step 3.3, initialization of properties of time-varying synchronous-flow neural network and and initialization of predecessor pre and successor next of each neuron i and successor next i target node next i ;

[0176] Step 4, iterative update of states of all time-varying synchronous-flow neurons as described in Step 4.1, with the following steps:

[0177] Step 4.1, at T = 0, the start neuron a sends traffic to its successor neuron next a .

[0178] Step 4.2, for neuron b, calculate for neuron c, calculate for neuron d, calculate

[0179] Step 4.3, target neuron d does not satisfy the condition, set T + Δt = 1

[0180] Step 4.4, for neuron b, calculate neuron b sends traffic to its successor neuron c; for neuron c, calculate neuron c does not send traffic to its successor neuron c; for neuron d, calculate

[0181] Step 4.5, target neuron d does not satisfy the condition, set T + Δt = 2

[0182] Step 4.6, for neuron b, calculate for neuron c, calculate neuron c sends traffic to its successor neuron d; for neuron d, calculate

[0183] Step 4.7, target neuron d does not satisfy the condition, set T + Δt = 3

[0184] Step 4.8, for neuron b, calculate for neuron c, calculate neuron c sends traffic to its successor neuron d; for neuron d, calculate R d = pre d ;

[0185] Step 4.9, determine R d = pre dSatisfy the condition, generate time-varying flow limit u 0 .

[0186] The construction of the time-varying maximum flow constraint equation in step 5 converges, the specific steps are as follows:

[0187] Step 5.1, construct the time-varying maximum flow constraint equation, the matrix A is a node-arc matrix, each row represents a node in the graph, and each column represents an arc in the graph. Then the topological relationship of the time-varying network is converted into a 4x4 node-arc A as follows

[0188]

[0189] Then the time-varying maximum flow constraint equation of the time-varying network can be constructed:

[0190] max f

[0191] s.t.A T x=d,

[0192] 0≤x≤u T

[0193] Where x=(x ab ,x ac ,x bc ,x cd ) T ∈R 4×1 is the flow of each arc in the graph, subject to the upper limit u T =(u ab ,u ac ,u bc ,u cd ) T =(3,1,2,3) T ∈R 4×1 .

[0194] Step 5.2, problem conversion, construct a new time-varying maximum flow constraint equation.

[0195] Let e4=(0,0,0,1) T ,e1=(1,0,0,0) T ∈R 4 , then (e4-e1)f=-d. The equation can be equivalent to:

[0196] max f

[0197] s.t.(e4-e1)f+Ax=0,

[0198] 0≤x≤u 0

[0199] Where, Let x5=f, So the equation is equivalent to:

[0200] max f

[0201] s.t.(e4-e1)x5+Ax=0,

[0202]

[0203] Let X=(x1,x2,x3,x4,x5) T ∈R 5×1 , U 0 =(3,1,2,3,9) T ∈R 5×1 Let q=(-1,0,0,,1) T ∈R 4 ×1 Let B=[A|q],

[0204] So

[0205]

[0206] So BX=(e4-e1)x5+Ax

[0207] Let c=(0,0,0,0,1) T ∈R 5×1 , then c T X=f

[0208] So, the equation can be equivalent to

[0209] max c T X

[0210] s.t.BX=0,

[0211] 0≤X≤U 0

[0212] Step 5.3, construct KKT equation according to time-varying maximum flow constraint equation.

[0213] According to the above equation, we get

[0214]

[0215] So, we can get the following KKT conditions:

[0216] (-c T +B T y+h)X=0,-BX=0

[0217] -c T +B T y+h≥0,-(X-U 0) > 0

[0218] - h T (X-U 0 ) > 0, X, h > 0, y is free

[0219] Step 5.4, solve KKT conditions

[0220] Let

[0221]

[0222] is equivalent to

[0223]

[0224] When F(w) = 0, X in w is the optimal flow on each edge.

[0225] Introduce projection where When F(w) = 0, X in w is the optimal flow on each edge. When F(w) = 0, X in w is the optimal flow on each edge. This is because when F(w) = 0, While ensuring that the first 2(m+1) terms in w are non-negative.

[0226] Introduce a dynamic model to solve

[0227]

[0228] Where 0 < λ < 1,

[0229]

[0230] From the above formula, we can know

[0231]

[0232] When the error e(w, β) tends to 0, the optimal solution of the time-varying maximum flow scheduling problem can be obtained, and X = (x ab , x ac , x bc , x cd , x5) T = (2, 1, 2, 3, 3) T , where the flow of each edge is shown by the first m terms, and the time-varying maximum flow scheduling to the target node d is finally obtained as 3.

[0233] The output time-varying synchronous flow neural network query result is described in step 6, and the specific steps are as follows:

[0234] Query the convergence result of the time-varying maximum flow constraint equation, and output x in Xm+1

[0235] It can be seen that the application discloses a maximum flow scheduling method based on a time-varying synchronous flow neural network, ensures global optimal flow distribution under the condition of meeting time-varying capacity constraints based on a time shortest path principle, can effectively process flow scheduling problems in a dynamic network environment, and obtains a theoretically optimal solution.

[0236] Embodiment three

[0237] Please refer to Figure 6 , Figure 6 is a structural schematic diagram of a maximum flow scheduling device based on a time-varying synchronous flow neural network disclosed by the embodiment of the application. Wherein, Figure 6 The maximum flow scheduling device based on a time-varying synchronous flow neural network described in the application is applied to the technical field of flow scheduling, and the embodiment of the application is not limited. As Figure 6 shown, the maximum flow scheduling device based on a time-varying synchronous flow neural network can include the following operations:

[0238] S301, a neuron hierarchy structure construction module, configured to process time-varying synchronous flow neurons to obtain a neuron hierarchy structure; the neuron hierarchy structure includes a flow input layer, a flow receiving layer, a neuron state layer, a flow generating layer and a flow sending layer;

[0239] S302, a time-varying synchronous flow neural network construction module, configured to process the neuron hierarchy structure to obtain a time-varying synchronous flow neural network;

[0240] S303, a parameter initialization module, configured to initialize the time-varying synchronous flow neurons and the time-varying synchronous flow neural network to obtain initialized time-varying synchronous flow neuron parameters and initialized time-varying synchronous flow neural network parameters;

[0241] The initialized time-varying synchronous flow neuron parameters include a number of predecessor nodes pre i , a number of successor nodes next i and a number of target nodes;

[0242] The time-varying synchronous flow neural network parameters include a timer time T, an iteration step length Δt and attribute information;

[0243] S304, a flow scheduling module, configured to process the time-varying synchronous flow neural network according to the initialized time-varying synchronous flow neuron parameters and the initialized time-varying synchronous flow neural network parameters to obtain a maximum flow scheduling result.

[0244] Embodiment four

[0245] Please refer to Figure 7 , Figure 7is another structure schematic view of the maximum flow scheduling device based on time-varying synchronous flow neural network disclosed by the embodiment of the present application. Wherein, Figure 7 The maximum flow scheduling device based on time-varying synchronous flow neural network described is applied to the technical field of flow scheduling, and the embodiments of the present application are not limited. Figure 7 As shown in the figure, the maximum flow scheduling device based on time-varying synchronous flow neural network can include the following operations:

[0246] The memory 401 stores executable program codes.

[0247] The processor 402 is coupled with the memory 401.

[0248] The processor 402 calls the executable program codes stored in the memory 401, and is used for executing the steps in the maximum flow scheduling method based on time-varying synchronous flow neural network described in the embodiment one and the embodiment two.

[0249] Embodiment five

[0250] The embodiment of the present application discloses a computer readable storage medium which stores a computer program for electronic data exchange, wherein the computer program makes the computer used for executing the steps in the maximum flow scheduling method based on time-varying synchronous flow neural network described in the embodiment one and the embodiment two.

[0251] The device embodiments described above are only schematic, wherein the modules illustrated as separate components can or can not be physically separated, and the components illustrated as modules can or can not be physical modules, that is, they can be located in one place, or can be distributed on multiple network modules. According to actual needs, part or all of the modules can be selected to achieve the purpose of the present embodiment scheme. Those skilled in the art can understand and implement without creative labor.

[0252] Those skilled in the art can clearly understand the implementation of the various embodiments by means of software and the necessary general hardware platform through the above specific description of the embodiments, and of course, the embodiments can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that contributes to the present application can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, including a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disk storage, a magnetic disk storage, a magnetic tape storage, or any other medium that can be used to carry or store data in a computer readable manner.

[0253] Finally, it should be noted that: the maximum flow scheduling method and device based on time-varying synchronous flow neural network disclosed by the embodiments of the present application are only the preferred embodiments of the present application, and are used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that; it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A time-varying synchronous flow neural network-based maximum flow scheduling method, characterized in that, The method comprises: S1, processing time-varying synchronous flow neurons to obtain a neuron hierarchy, comprising: S11, processing the time-varying synchronous stream neuron to obtain a stream input layer, the stream input layer including a flow identifier When If not empty, there is flow to the neuron i ; S12, constructing a flow receiving layer of the time-varying synchronous flow neuron; the flow receiving layer is used for receiving flow information output by the flow input layer and wherein, a flow from neuron is used for storing flow sent by a starting node at a T moment arriving at a neuron i flow, a time of arrival of the flow is stored S13, constructing a neuron state layer of the time-varying synchronous stream neuron; the neuron state layer comprises an activation state , a target set , and a flow record are constituted; S14, constructing a stream generation layer of the time-varying synchronous stream neuron; the stream generation layer comprises a tuple ; S15, constructing a flow sending layer of the time-varying synchronous flow neuron; the flow sending layer is configured to send the flow generated by the flow generation layer to a successor node set according to the successor node set sending traffic, sending the traffic generated by the flow generation layer to a successor node; The neuron hierarchy comprises a flow input layer, a flow receiving layer, a neuron state layer, a flow generating layer, and a flow sending layer; S2, processing the neuron hierarchy to obtain a time-varying synchronous flow neural network, comprising: S21, obtaining node structure information of a target network; S22, processing the node structure information of the target network to obtain a neuron cluster of the time-varying synchronous flow neural network; The neuron cluster of the time-varying synchronous flow neural network comprises N time-varying synchronous flow neurons; S23, processing the neuron cluster of the time-varying synchronous flow neural network to obtain a time-varying synchronous flow neural network; S3, initializing the time-varying synchronous flow neurons and the time-varying synchronous flow neural network to obtain initialized time-varying synchronous flow neuron parameters and initialized time-varying synchronous flow neural network parameters; The initialized time-varying synchronous stream neuron parameters include a predecessor node a number of successors a number of successors and a target node number The time-varying synchronous stream neural network parameters include a timer time T , an iteration step , and attribute information; S4, processing the time-varying synchronous flow neural network according to the initialized time-varying synchronous flow neuron parameters and the initialized time-varying synchronous flow neural network parameters to obtain a maximum flow scheduling result, comprising: S41, updating the state of a time-varying synchronous flow neuron of the time-varying synchronous flow neural network to obtain state update information of the time-varying synchronous flow neuron, comprising: S411, for any time-varying synchronous flow neuron, traversing all predecessor neurons thereof to query whether there is flow arriving at the current time-varying synchronous flow neuron, and if there is flow arriving, receiving the flow through the flow input layer and the flow receiving layer and recording in the neuron state layer; S412, when the state layer of neuron k detects that the traffic sent at time t satisfies the activation state , then the flow generation layer of neuron k will generate new traffic , and send the traffic through the flow sending layer of neuron k to the successor node , obtaining the state update information of the time-varying synchronous flow neuron; S42, processing the time-varying synchronous flow neural network according to the state update information to obtain a convergent time-varying maximum flow constraint equation; S43, solving the convergent time-varying maximum flow constraint equation to obtain a maximum flow scheduling result.

2. The time-varying synchronous flow neural network based maximum flow scheduling method according to claim 1, wherein, The initialization of the time-varying synchronous flow neurons and the time-varying synchronous flow neural network to obtain the initialized time-varying synchronous flow neuron parameters and the initialized time-varying synchronous flow neural network parameters comprises: S31, initialize a timer time of the time-varying synchronization flow neural network T ; S32, initializing an iteration step size of the time-varying synchronization flow neural network ; S33, initializing attribute information of the time-varying synchronous flow neural network; S34, initializing predecessor nodes of each time-varying synapse neuron number, successor nodes number and target node number.

3. A time-varying synchronous flow neural network based maximum flow scheduling apparatus, characterized in that, The device comprises: A neuron hierarchy construction module for processing time-varying synchronous flow neurons to obtain a neuron hierarchy, comprising: S11, processing the time-varying synchronous stream neuron to obtain a stream input layer, the stream input layer including a flow identifier When If not empty, there is flow to the neuron i ; S12, constructing a flow receiving layer of the time-varying synchronous flow neuron; the flow receiving layer is used for receiving flow information output by the flow input layer and wherein, a flow from neuron is used for storing flow sent by a starting node at a T moment arriving at a neuron i flow, a time of arrival of the flow is stored. S13, constructing a neuron state layer of the time-varying synchronous stream neuron; the neuron state layer comprises an activation state , a target set , and a flow record are constituted; S14, constructing a stream generation layer of the time-varying synchronous stream neuron; the stream generation layer comprises a tuple ; S15, constructing a flow sending layer of the time-varying synchronous flow neuron; the flow sending layer is configured to send the flow generated by the flow generation layer to a successor node set according to the successor node set sending traffic, sending the traffic generated by the flow generation layer to a successor node; The neuron hierarchy comprises a flow input layer, a flow receiving layer, a neuron state layer, a flow generating layer, and a flow sending layer; A time-varying synchronous flow neural network construction module for processing the neuron hierarchy to obtain a time-varying synchronous flow neural network, comprising: S21, obtaining node structure information of a target network; S22, processing the node structure information of the target network to obtain a neuron cluster of the time-varying synchronous flow neural network; The neuron cluster of the time-varying synchronous flow neural network comprises N time-varying synchronous flow neurons; S23, processing the neuron cluster of the time-varying synchronous flow neural network to obtain a time-varying synchronous flow neural network; A parameter initialization module is configured to initialize the time-varying synchronous flow neuron and the time-varying synchronous flow neural network to obtain initialized time-varying synchronous flow neuron parameters and initialized time-varying synchronous flow neural network parameters. The initialized time-varying synchronous stream neuron parameters include a predecessor node a number of successors a number of successors and a target node number The time-varying synchronous stream neural network parameters include a timer time T , an iteration step , and attribute information; A flow scheduling module is configured to process the time-varying synchronous flow neural network according to the initialized time-varying synchronous flow neuron parameters and the initialized time-varying synchronous flow neural network parameters to obtain a maximum flow scheduling result, including: S41, updating a state of a time-varying synchronous flow neuron of the time-varying synchronous flow neural network to obtain state update information of the time-varying synchronous flow neuron, including: S411, for any time-varying synchronous flow neuron, traversing all predecessor neurons thereof to query whether there is flow arriving at the time-varying synchronous flow neuron at a current time, and if there is flow arriving, receiving the flow through a flow input layer and a flow receiving layer and recording in a neuron state layer; S412, when the state layer of neuron k detects that the traffic sent at time t satisfies the activation state , then the flow generation layer of neuron k will generate new traffic , and send the traffic to the successor node through the flow sending layer of neuron k , obtaining the state update information of the time-varying synchronous flow neuron S42, processing the time-varying synchronous flow neural network according to the state update information to obtain a converged time-varying maximum flow constraint equation; S43, solving the converged time-varying maximum flow constraint equation to obtain the maximum flow scheduling result.

4. A time-varying synchronous flow neural network based maximum flow scheduling apparatus, characterized in that, The device includes: a memory storing executable program codes; a processor coupled with the memory; the processor invokes the executable program codes stored in the memory to execute the maximum flow scheduling method based on the time-varying synchronous flow neural network according to any one of claims 1-2.

5. A computer storable medium, characterized by The computer storage medium stores computer instructions, which are invoked to execute the maximum flow scheduling method based on the time-varying synchronous flow neural network according to any one of claims 1-2.

Images