A computing method and system based on constrained network minimum action evolution

Abstract

The application relates to a calculation method and system based on constrained network minimum action evolution, belonging to the cross field of computer science and computational physics. The technical problem to be solved by the application is to overcome the limitation that the existing calculation paradigm completely depends on instruction sequence execution. The technical solution of the application is as follows: receiving constraint declaration information input by a user, including calculation element declaration, partial order relation declaration and action contribution item declaration; constructing an initial coupling network and initializing coupling strength according to a chaotic rule; under the constraint of the partial order relation, optimizing the coupling network through an iterative search method, searching for a target network configuration which makes the total action approach minimization; and performing coarse-grained observation on the target network configuration and outputting an emergent result. The application breaks through the existing Turing machine calculation paradigm, for the first time, takes the minimum action principle as the core driving force of calculation, replaces the instruction sequence with the partial order relation as the sequential constraint of a program, makes the program automatically evolve into a macro result from the constraint, realizes the fundamental innovation of the calculation paradigm, and can be applied to the fields of physical simulation, material design, cosmology research, new type of calculation hardware design and the like.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of computer science and computational physics, and specifically relates to a novel computational method and its interpreter system that does not rely on instruction sequence execution, particularly a computational paradigm implementation based on the minimum action evolution principle of constrained networks. Background Technology

[0002] Existing computational models are built upon the theory of Turing machines. The core of the Turing machine model is the sequential execution pattern of "state + instruction → new state". Since Turing proposed this model in 1936, the interpreters and compilers of all mainstream programming languages ​​(such as C, Python, Java, etc.) are essentially concrete implementations of the Turing machine model at different levels of abstraction.

[0003] In this traditional computing paradigm, a program is defined as a sequence of instructions, and the computation process involves executing these instructions one by one in a specific order, modifying the state of memory during execution. The output of the program is entirely determined by the sequence of instructions written by the programmer; the computing system itself does not have the ability to automatically evolve the result from constraints.

[0004] Existing technologies suffer from the following fundamental limitations: the singularity of computational paradigms—for nearly ninety years, all computer programs have followed an instruction sequence execution model, failing to simulate the widespread phenomenon of "order emerging from constraints" in nature; disconnection from underlying physical laws—fundamental laws of nature (such as the principle of least action) lack direct computational primitives in existing computational models; lack of emergent computational capabilities—existing models cannot automatically generate macroscopic structures and observable macroscopic quantities from microscopic rules; and inherent limitations of sequential execution—the execution order of instruction sequences is artificially defined, fundamentally different from the natural order determined by causal partial order relationships in the physical world. Summary of the Invention

[0005] The technical problem to be solved by this invention is to overcome the limitation of existing computing paradigms that rely entirely on the execution of instruction sequences, and to provide a novel computing method and system based on the minimum action evolution of constraint networks, so that the program can automatically evolve macroscopic results from user-declared constraints, realizing the "emergence" of computing rather than "execution".

[0006] This invention provides a computational method based on the minimum action evolution of constrained networks, comprising the following steps: Step S1: Receive constraint declaration information. The user defines all the constraint information required for computation through declarative syntax, including: declarations of multiple computational primitives. In a preferred embodiment, the computational primitives are qubits, which are the only data primitives in the system. Unlike existing programming languages, there are no other data types such as integers, floating-point numbers, or strings in this system; declaration of at least one partial order relation, which defines the unidirectional constraint relationship between computational primitives. In a preferred embodiment, the partial order relation is a causal edge, which is the only order constraint in the system, completely replacing the concepts of statement execution order, assignment order, and thread scheduling in traditional programming languages; declaration of at least one action contribution term, which defines the components of the physical action that need to be minimized during network evolution. The user can declare multiple action contribution terms and assign weights to them to control the system evolution behavior; and declaration of observation constraints, defining the basis vectors and objectives for coarse-grained observation of the final network configuration.

[0007] Step S2: Construct the initial coupling network. Based on the number N declared computational primitives, construct an N×N initial coupling matrix. Randomly initialize the coupling matrix to form the initial coupling network. In a preferred embodiment, the random initialization follows the SYK maximum chaos model, that is, all computational primitives are randomly assigned Gaussian-distributed coupling strengths pairwise to simulate the maximum chaotic initial state of the physical system.

[0008] Step S3: Iterative search for the minimum action configuration. Under the constraints of the partial order relation declaration, the coupled network is optimized using an iterative search method. In a preferred embodiment, the iterative search method is a Markov chain Monte Carlo sampling method, specifically including: constructing a directed acyclic graph that satisfies the partial order relation constraints based on the current coupling strength matrix; calculating the total action corresponding to the directed acyclic graph; randomly perturbing at least one element in the coupling matrix, reconstructing the directed acyclic graph and calculating the new action; determining whether to accept the new configuration based on the change of the new action relative to the old action according to the Metropolis criterion; repeating the above process until the system gradually converges to the target network configuration where the total action approaches the minimum. In another embodiment, when the number of computational primitives does not exceed a preset threshold, a backtracking search method is used instead, enumerating all legal configurations to solve for the exact value.

[0009] Step S4: Coarse-grained observation. Coarse-grained observation is performed on the target network configuration obtained in Step S3. This coarse-grained observation includes: automatically classifying emerging structure types based on the topological characteristics of the target network configuration; and / or calculating physical prediction values ​​based on the topological invariants of the target network configuration. Finally, the system outputs the emergence results, completing the calculation.

[0010] The essential differences between this invention and existing technologies are as follows: a fundamental change in the driving force of computation, where traditional computation is driven by instruction sequences, while this invention is driven by the principle of least action; a physicalization of the sequential mechanism, where traditional computation is determined by the order of instructions specified by the programmer, while this invention is determined by physical causal partial order relationships; a unified data type, where traditional programming languages ​​have multiple data types, while this invention has only one data primitive: computational primitives; and an emergent output mechanism, where the output of traditional programs is the memory content specified by the programmer, while the output of this invention is the emergent result of coarse-grained observations of the network.

[0011] The beneficial effects of this invention include: pioneering a completely new computing paradigm, taking the principle of least action as the core driving force of the computing paradigm for the first time; natural parallelism, computing without causal correlation can be executed in parallel naturally; physical interpretability, the computing output directly corresponds to physical observations; high computing efficiency, through multi-layer optimization such as dual-mode solver, triangle index cache, lazy computing, intelligent matrix switching, and adaptive step count, it can efficiently handle network scales of hundreds of computing primitives. Attached Figure Description

[0012] Figure 1 This is an overall architecture diagram of the computing system in an embodiment of the present invention.

[0013] Figure 2 This is an iterative flowchart of the MCMC solver in an embodiment of the present invention.

[0014] Figure 3 This is a data flow diagram in an embodiment of the present invention, illustrating the complete processing chain from user input to output of emergent results.

[0015] Figure 4 This is a schematic diagram illustrating the observer's classification of the emergent structure of causal networks in an embodiment of the present invention. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0017] Example 1: System Overall Architecture. (Refer to...) Figure 1This embodiment provides a computational system based on minimum action evolution of constraint networks, comprising: a syntax preprocessing module, responsible for standardizing the source code input by the user, including deleting comments, unifying punctuation, and expanding scope declarations; a syntax parsing module, using the syntax rules defined by EBNF, parsing the preprocessed source code into a syntax tree through a parser, and then converting the syntax tree into a structured model dictionary through a Transformer, wherein the model dictionary contains all constraint information declared by the user; a semantic analysis module, performing semantic-level checks on the model dictionary, including detection of duplicate names of computational primitives and detection of partial order relation loops, wherein the partial order relation loop detection adopts a depth-first search three-color labeling method to ensure that the partial order relations declared by the user do not have closed loops; a solver module, including an exact solver and a parallel MCMC solver, wherein the exact solver is activated when the number of computational primitives does not exceed a preset threshold, and the MCMC solver is activated when it exceeds a preset threshold; and an observer module, performing coarse-grained observation of the network graph output by the solver, classifying the emergent structure according to the observation basis vectors specified by the user, and calculating physical prediction values ​​based on topological invariants.

[0018] Example 2: MCMC solution process. (Refer to...) Figure 2 The iterative process for solving MCMC is as follows: (1) Initialize an N×N coupling matrix. When the number of computational primitives does not exceed a preset threshold, a dense matrix is ​​used for storage; otherwise, a sparse matrix is ​​used. The edge of the partial order relation is assigned a fixed value, and other elements are randomly initialized according to the coupling type. (2) Construct a directed acyclic graph based on the current coupling matrix. Extract all candidate edges with weights greater than the threshold, arrange them in descending order of weight, and try to add them to the directed acyclic graph in turn. During the addition process, check whether the acyclic constraint is satisfied. (3) Perform full calculation of triangles on the constructed directed acyclic graph and cache them. All subsequent calculations that require triangle information are directly read from the cache to avoid repeated traversal. (4) Calculate the total action of the current directed acyclic graph. Use a lazy calculation strategy and only calculate the action contribution of the weights declared by the user. (5) Randomly perturb at least one element in the coupling matrix, reconstruct the directed acyclic graph and calculate the new action. Decide whether to accept the new configuration according to the Metropolis criterion. (6) Repeat step (5) until the adaptive step count is reached or the soft timeout protection is triggered, and return to the directed acyclic graph configuration with the minimum total action.

[0019] Example 3: Emergent Structure Classification. (Refer to...) Figure 4 The observer classifies the emergent structure based on the network's topological characteristics: when the network is a single connected component and the boundary topological load is even, it is classified as a string; when the network is a single connected component and the boundary topological load is odd, it is classified as a point; when the network is multiple connected components and the boundary topological load is even, it is classified as a membrane; and in other cases, it is classified as a cosmic bubble.

Claims

1. A computational method based on the minimum action evolution of constrained networks, characterized in that, Includes the following steps: Step S1: Receive constraint declaration information input by the user. The constraint declaration information includes: declarations of multiple computational primitives, declarations of at least one partial order relation, and declarations of at least one action contribution term. Step S2: Based on the number of declared computational primitives, construct an initial coupling network. The coupling strength between each computational primitive in the initial coupling network is initialized according to a preset chaos rule. Step S3: Under the constraints of the partial order relation declaration, the coupled network is optimized by an iterative search method. After each iteration, the network configuration is redefined and the total action corresponding to the configuration is calculated in order to search for the target network configuration that minimizes the total action. The total action is obtained by weighted summation of the at least one action contribution term. Step S4: Perform coarse-grained observation on the target network configuration and output the emergence results based on the observation constraints in the constraint declaration information.

2. A computational system based on minimum action evolution of constrained networks, characterized in that, include: The parsing module is used to receive and parse constraint declaration information input by the user; The solution module is used to construct an initial coupled network based on the constraint declaration information, and optimize the coupled network under the partial order relation constraint through an iterative search method to search for a target network configuration that minimizes the total action. The observation module is used to perform coarse-grained observation of the target network configuration and output emergence results.

3. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method of claim 1.

4. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method of claim 1.

5. The method according to claim 1, characterized in that, The computational primitive is a qubit, the partial order relation is a unidirectional causal edge between qubits, and the computational primitive is the only data primitive in the system.

6. The method according to claim 1, characterized in that, The contribution term of the action includes at least one of the following: discrete gravity term, spin network gauge term, topological entanglement term, information causal entropy term, matter term, and cosmological constant term.

7. The method according to claim 1, characterized in that, The chaotic rule is SYK fully connected random coupling initialization.

8. The method according to claim 1, characterized in that, The iterative search method in step S3 is the Markov chain Monte Carlo sampling method, and the re-determining of the network configuration includes: constructing a directed acyclic graph that satisfies the partial order relation constraint based on the current coupling strength matrix.

9. The method according to claim 8, characterized in that, The calculation of the total action in step S3 adopts an inertial calculation strategy, only calculating the action contribution items for which the user has declared weights.

10. The method according to claim 8, characterized in that, After the directed acyclic graph is constructed as described in step S3, all triangles of the graph are calculated synchronously and stored in the graph attribute cache. The subsequent calculation of total action and topological invariant are directly read from the cache.

11. The method according to claim 1, characterized in that, When the number of declared computational primitives does not exceed a preset threshold, a backtracking search method is used instead of the iterative search method to enumerate all legal configurations that satisfy the partial order relation constraint and accurately solve for the minimum action configuration.

12. The method according to claim 8, characterized in that, The number of iterations in the iterative search method is automatically adjusted according to the network complexity, which is calculated based on the number of computational primitives, the density of partial order relations, and the average degree.

13. The method according to claim 8, characterized in that, The coupling strength matrix automatically switches its storage format based on the number of computational primitives: dense matrix storage is used when the number of computational primitives does not exceed a preset threshold, and sparse matrix storage is used when the number exceeds the threshold.

14. The method according to claim 1, characterized in that, The coarse-grained observation in step S4 includes: automatically classifying the emergent structure type based on the number of connected components and boundary topological load of the target network configuration.

15. The method according to claim 1, characterized in that, The coarse-grained observation in step S4 includes: calculating physical prediction values ​​based on the topological invariants of the target network configuration, wherein the topological invariants include at least one of the first Chern class, Eulerian class, and discrete Ricci scalar.

16. The method according to claim 1, characterized in that, The constraint declaration information adopts a declarative syntax, the computational primitive is a unique data primitive in the system, and the partial order relation is a unique order constraint in the system, replacing the statement execution order in traditional programming languages.

17. The method according to claim 8, characterized in that, The iterative search method employs multiple independent Markov chains executed in parallel, and selects the chain with the smallest total action as the target network configuration.

18. The method according to claim 8, characterized in that, The iterative search method has a time limit, and returns the currently found optimal configuration when the time limit is reached.

Images