WO2019074504A1 - Procédés de génération automatique de modèles précis en temps réduit - Google Patents

Procédés de génération automatique de modèles précis en temps réduit Download PDF

Info

Publication number
WO2019074504A1
WO2019074504A1 PCT/US2017/056218 US2017056218W WO2019074504A1 WO 2019074504 A1 WO2019074504 A1 WO 2019074504A1 US 2017056218 W US2017056218 W US 2017056218W WO 2019074504 A1 WO2019074504 A1 WO 2019074504A1
Authority
WO
WIPO (PCT)
Prior art keywords
agent
algorithm
data
attribute
outcome
Prior art date
Application number
PCT/US2017/056218
Other languages
English (en)
Inventor
Beau WALKER
Michael COLBUS
Reece COLBUS
Hunter COLBUS
Original Assignee
Liquid Biosciences, Inc.
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 Liquid Biosciences, Inc. filed Critical Liquid Biosciences, Inc.
Priority to PCT/US2017/056218 priority Critical patent/WO2019074504A1/fr
Publication of WO2019074504A1 publication Critical patent/WO2019074504A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/12—Computing arrangements based on biological models using genetic models
    • G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • 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
    • G06N3/08—Learning methods
    • G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming

Definitions

  • the field of the invention is methods for automatically generating accurate models in reduced time.
  • simulations may also be appropriate to model processes that "are inescapably complex.”
  • project Mesa which is licensed under the Apache License, Version 2.0 (available at http://www.apache.Org/licenses/LICENSE-2.0).
  • Neither Peck nor Project Mesa teaches or suggests that simulations can be used as a tool to simulate the process of model generation itself, in which digital agents in a simulation comprise algorithms capable of predicting outcomes based on data, and where agents can mate with similar digital agents to generate improved algorithms for predicting outcomes based on data.
  • the present invention provides a method of agent-based modeling as a framework for evolutionary development of genetic algorithms.
  • the methods and apparatuses described herein have many desirable qualities, including: (1 ) they allow emergent behavior of a simulation to identify "bad" data— a task that computers generally perform poorly if at all; (2) they generate new and useful algorithms at a faster rate than conventional algorithmic generation methods; (3) they generate new and useful algorithms that conventional algorithmic generation methods would be incapable of generating without human intervention; (4) they generate new and useful algorithms that neither conventional algorithmic generation methods nor human invention could generate.
  • a method of generating an algorithm to process data comprises creating a first agent and a second agent within a digital environment, wherein the first agent comprises a first algorithm capable of processing data and an agent attribute, and the second agent comprises a second algorithm capable of processing data; generating a predicted outcome based on applying the first algorithm to a feature of a set of data, wherein the set of data comprises the feature and an outcome; comparing the predicted outcome to the outcome of the set of data; modifying an attribute of the first agent based on the comparison; and combining, based on the modified attribute of the first agent, the first and the second algorithms to create a third algorithm.
  • an algorithmic generator comprises a computing device configured to create a first and second agent within a digital environment, wherein the first agent comprises a first algorithm capable of processing data and an agent attribute and the second agent comprises a second algorithm capable of processing data; process a set of data comprising a feature and an outcome with the first algorithm, thereby generating a predicted outcome; compare the predicted outcome to the outcome of the set of data; determine, based on the comparison result, whether to modify an attribute of the first agent; and determine, based on factors comprising the attribute of the first agent, whether to generate a third algorithm based on the first algorithm.
  • a method to simulate algorithmic generation comprising the steps of: creating an agent within a digital environment, wherein the agent comprises an algorithm capable of processing data, and wherein the algorithm comprises algorithmic components; processing a set of data comprising a feature and an outcome with the algorithm, thereby generating a predicted outcome; comparing the predicted outcome to the outcome of the set of data; and determining, based on the comparison result, whether to generate, based on the algorithmic components and a mutation parameter, a second algorithm capable of processing data.
  • the disclosed subject matter provides advantageous technical effects including improved operation of a computer by dramatically decreasing computational cycles required to perform certain tasks (e.g., genetic programming).
  • genetic programming is not a tenable solution in many situations due in large part to its steep computational requirements that would necessitate sometimes months and years of computing time to, for example, develop models as in embodiments of the inventive subject matter.
  • Figure 1 shows a scheduler and associated process flow.
  • Figure 2 shows process flow and associations for a reproduction event.
  • Figure 2a shows process flow and associations for an alternative reproduction event.
  • Figure 3 shows process flow and associations for a data processing event.
  • Figures 4, 5, and 6 depict exemplary processes according to the invention.
  • Figure 7 depicts a software specification of a virtual space as used in aspects of the invention.
  • Figure 8 depicts a computer listing defining a data collection tool as used in aspects of the invention.
  • Figure 9 depicts a software specification defining a software method for running a batch of simulations.
  • Figure 10 depicts a computer listing implementing a software method for running a batch of simulations. Detailed Description [0022] DEFINITIONS
  • inventive subject matter provides example embodiments of the inventive subject matter. Although each embodiment represents a single combination of inventive elements, the inventive subject matter is considered to include all possible combinations of the disclosed elements. Thus, if one embodiment comprises elements A, B, and C, and a second embodiment comprises elements B and D, then the inventive subject matter is also considered to include other remaining elements.
  • Coupled to is intended to include both direct coupling (in which two elements that are coupled to each other contact each other) and indirect coupling (in which at least one additional element is located between the two elements).
  • Coupled to and “coupled with” are used synonymously.
  • the numbers expressing quantities of ingredients, properties such as concentration, reaction conditions, and so forth, used to describe and claim certain embodiments of the invention are to be understood as being modified in some instances by the term "about.” Accordingly, in some embodiments, the numerical parameters set forth in the written description and attached claims are approximations that can vary depending upon the desired properties sought to be obtained by a particular embodiment. In some embodiments, the numerical parameters should be construed in light of the number of reported significant digits and by applying ordinary rounding techniques. Notwithstanding that the numerical ranges and parameters setting forth the broad scope of some embodiments of the invention are approximations, the numerical values set forth in the specific examples are reported as precisely as practicable.
  • any language directed to a computer should be read to include any suitable combination of computing devices, including servers, interfaces, systems, databases, agents, peers, Engines, controllers, or other types of computing devices operating individually or collectively.
  • the computing devices comprise a processor configured to execute software instructions stored on a tangible, non-transitory computer readable storage medium (e.g., hard drive, solid state drive, RAM, flash, ROM, etc.).
  • the software instructions preferably configure the computing device to provide the roles, responsibilities, or other functionality as discussed below with respect to the disclosed apparatus.
  • the various servers, systems, databases, or interfaces exchange data using standardized protocols or algorithms, possibly based on HTTP, HTTPS, AES, public-private key exchanges, web service APIs, known financial transaction protocols, or other electronic information exchanging methods.
  • Data exchanges preferably are conducted over a packet-switched network, the Internet, LAN, WAN, VPN, or other type of packet switched network.
  • the invention comprises a digital environment in which digital agents may process data (represented as another type of digital agent) and reproduce, where reproduction occurs based on success processing data.
  • the invention provides technological advantages over prior art systems.
  • the invention described herein (1 ) allows emergent behavior of a simulation to identify "bad" data— a task that computers generally perform poorly, if at all; (2) generates new and useful algorithms at a faster rate than conventional algorithmic generation methods; (3) generates new and useful algorithms that conventional algorithmic generation methods would be incapable of generating without human intervention; and (4) generates new and useful algorithms that neither conventional algorithmic generation methods nor human invention would generate.
  • successful digital agents are more likely to reproduce, and the combination of successful agents yields algorithms capable of greater explanatory power.
  • datasets that are more difficult to process can become apparent because, in some embodiments, successful digital agents will tend to avoid those datasets.
  • the digital environment may comprise any computer hardware or software capable of operating the invention described in this application.
  • the digital environment comprises all digital processing aspects of the invention. It is alternately referred to as a digital ecology.
  • One component of the invention is a scheduler, which is capable of activating agents.
  • the scheduler described in more detail below, is a module that controls what agents are activated and "when" within the digital ecology.
  • a simulation may advance time in discrete "ticks.” These ticks may optionally be associated with a continuous clock, such that a defined number of ticks occurs per second/minute/hour/etc.
  • Scheduler 101 may thus activate agents based on the number of elapsed ticks.
  • scheduler comprises a variable for tracking the number of "ticks" which have elapsed in the simulation.
  • Other embodiments may also comprise a variable for tracking the time.
  • Figure 1 depicts one embodiment of a scheduler and associated process.
  • Scheduler 101 further comprises a list of agents capable of activation.
  • Figure 1 depicts agent 104, which is described in greater detail below.
  • Scheduler 101 may further comprise a method to add an agent and a method to remove an agent from the list of agents.
  • Scheduler 101 method to add an agent comprises software code to add to the list of agents to activate, while the method to remove an agent comprises software code to remove an agent from the list of agents to activate.
  • these methods may be implemented through permitting other modules to access the list of agents and modify the list.
  • Scheduler 101 further comprises a method, which comprises software code to activate agents in scheduler 101 's list of agents.
  • Agent 104's behavior on activation is described in greater detail below.
  • changes may be "staged," meaning that changes to the agent's properties are calculated, but not yet applied.
  • commit stage 103 depicts applying changes in a behavior activation stage.
  • Commit stage 103 permits agents to be activated in a simultaneous simulation, rather than sequentially— applying changes after they are staged.
  • agents may be activated in phases.
  • Figure 1 depicts one phase, behavior 102.
  • scheduler 101 would first activate behavior 102 for scheduler agents, then activate a second behavior (not depicted in Figure 1 ) for scheduler agents after processing of behavior 102.
  • an agent may comprise a first activation phase, which determines a first property based on a first parameter, and a second activation phase, which
  • Scheduler 101 thereby determines simulation behavior when the simulation advances one or more "ticks.” It may be advantageous, in some embodiments, to activate all agents by sequential iteration through the list of agents. In other embodiments, random iteration through the list of agents may be
  • the invention comprises a space module, which comprises software code for simulating the space through which agents move.
  • the space module comprises width, height, torus, and/or grid properties.
  • the width and height properties determine a width and height for the simulated grid.
  • the torus property determines whether the grid will be simulated as a torus.
  • the grid property comprises a list comprising the contents of the cells in the grid.
  • the space module further comprises methods for determining aspects of the simulated space. In some embodiments, these methods may be implemented outside the space module through access to the variables described in the space module. These methods comprise methods for getting neighbors, getting neighborhood, getting cell contents, iterating over cell neighbors, determining coordinates, placing agents, moving agents, determining torus wrapping,
  • Example software code illustrates one implementation of the space module. Other implementations, of course, are possible.
  • Figure 7 shows an example specification for a space module, from Mesa, according to the invention.
  • Grid class 701 depicts properties and methods of a programming object that can be used according to an aspect of the invention.
  • digital spatial coordinates may be associated with agents in some embodiments of the invention.
  • Some embodiments of the invention may further comprise a data collector module which comprises software code for collecting data generated by the simulation.
  • a data collection module like the one in Mesa is suitable for use with the present invention.
  • One embodiment of the data collection module is provided below to illustrate the model by example. It will be appreciated by one of skill in the art that other embodiments are possible.
  • Figure 8 shows an example of software code implementing one example of a data collection module.
  • Data collector class 801 provides a programming object suitable for performing several useful functions for data collection, including iteration, data collection, and helper functions such as add_table_row.
  • Some embodiments of the invention may further comprise a batch runner module, which comprises software code for initializing at least two runs of a simulation with a first set of parameters in a first run and a second set of parameters in a second run.
  • a batch runner module which comprises software code for initializing at least two runs of a simulation with a first set of parameters in a first run and a second set of parameters in a second run.
  • a batch runner provided by Mesa may be suitable for use with the present invention.
  • Figure 9 depicts an example software code specification, such as that provided by Mesa, for a batch runner for performing runs of a simulation as described above.
  • Argument specification for BatchRunner 901 illustrates possibilities for batch running parameters.
  • Figure 10 depicts an example implementation of software code, such as that provided by Mesa, for performing simulations as described above, as well as for recording data provided by the simulation.
  • Run_all function 1001 provides a software method for running a batch of simulations given simulation start parameters and storing results.
  • _prepare_report_table function 1002 provides a helper function for storing data provided by a data collection module.
  • the invention further comprises an agent, which comprises software code for storing an agent attribute and determining agent behavior.
  • Figure 1 depicts one such embodiment comprising agent 104, which comprises an algorithm 105 capable of processing data.
  • Agent 104 also comprises an attribute 106, which may be termed "agent energy.”
  • Agent 104 may also comprise additional attributes, e.g., a mutation parameter. Agent attributes need not be identical for all agents in a simulation. For example, some agents may have an attribute for data comprising a feature and an outcome, while other agents may lack an attribute for data comprising a feature and an outcome.
  • the software code for determining agent behavior implements a "step" function, which is invoked when an agent is activated.
  • the "step" function is behavior 102, which may access agent attributes and determine agent behavior based on agent attributes.
  • the step function at either the agent level or above, is capable of causing a data processing event or a
  • behavior 102 determines whether to invoke a data processing event or a reproduction event.
  • behavior 102 determines whether to invoke a data processing event or a reproduction event.
  • many different determinations are possible in many embodiments.
  • one method of determination is based on coordinates associated with an agent in digital simulated space, in which a data processing event is invoked when an agent comprising an algorithm, e.g. agent 104, is sufficiently close in digital simulated space to an agent comprising data, e.g. agent 107.
  • Agent 104 further comprises software code for a reproduction event, and may also comprise software code for determining whether to cause a reproduction event within behavior 102. The determination is made based on at least one agent attribute, e.g. agent energy. For example, if agent energy is compared to a threshold energy determined to be required to reproduce, and the comparison result is that the agent energy is below the threshold, no reproduction event is caused.
  • agent energy e.g. agent energy
  • Figure 2 depicts reproduction event determination and reproduction event process flow.
  • reproduction determination 201 is based on an attribute 202 of agent 203, wherein agent 203 also comprises first algorithm 204 (comprising algorithmic components).
  • Agent 203 may also comprise a mutation parameter 205, but mutation parameter 205 may also be a property of another object. If an affirmative determination is made, reproduction event 206 is invoked with first algorithm 204 and mutation parameter 205.
  • the reproduction event comprises software for generating an agent attribute comprising an algorithm capable of processing data based on (1 ) the agent algorithm capable of processing data and a second agent algorithm capable of processing data, and, optionally, a mutation parameter or (2) the agent algorithm capable of processing data and a mutation parameter.
  • the mutation parameter if present, determines the degree of randomization in the generated agent attribute algorithm.
  • a reproduction event generates an algorithm capable of processing data. As described above, the generation is based on (1 ) the agent algorithm capable of processing data and a second agent algorithm capable of processing data, and, optionally, a mutation parameter or (2) the agent algorithm capable of processing data and a mutation parameter. [0062]
  • the algorithm capable of processing data comprises algorithmic
  • computational operators e.g., logical statements like IF, AND, OR
  • mathematical operators e.g., arithmetic operations like multiplication, division, subtraction, and addition; trigonometric operations;
  • logistic functions e.g., a constant numerical value, including integers or values like pi
  • a predictor e.g., observed or measured values or formulas
  • features e.g., characteristics
  • variables e.g., ternary operators (e.g., an operator that takes three arguments where the first argument is a comparison argument, the second is the result upon a true comparison, and the third is the result upon a false comparison)
  • algorithms formulas, literals, functions (e.g., unary functions, binary functions, etc.), binary operators (e.g., an operator that operates on two operands and manipulates them to return a result), weights and weight vectors, nodes and hidden nodes, gradient descent, sigmoidal activation functions, hyper- parameters, and biases.
  • algorithmic components of the first agent algorithm and algorithmic components of the second agent algorithm are grouped in a common pool.
  • the generated agent algorithm is based on algorithmic components selected from this common pool, together with algorithmic components selected from a second set of algorithmic components, where drawing of components from the second set of algorithmic components is based on the mutation parameter.
  • algorithmic components of the first agent algorithm are selected for the generated agent algorithm, together with algorithmic components selected from a second set of algorithmic components, where drawing of components from the second set of algorithmic components is based on the mutation parameter.
  • reproduction event 206 generates an algorithm wherein algorithmic components of the first algorithm 204 are combined or modified with algorithmic components of a broader set of algorithmic components, wherein the likelihood of adding an
  • algorithmic component from the common pool is determined by mutation parameter 205.
  • reproduction event 206a generates an algorithm comprising generated algorithmic components, wherein the generated algorithmic components are selected from the set of algorithmic components comprising the first algorithm 204a's algorithmic components and the second algorithm 208a's algorithmic components.
  • the invention further comprises a data processing event, which comprises software code for applying an agent algorithm to data, which comprises one or more features and an outcome.
  • the agent algorithm generates a predicted outcome based on the one or more features, and the predicted outcome is then compared to the outcome of the data.
  • An agent attribute of the agent comprising the agent algorithm is then modified, based on the result of the comparison. For example, if the predicted outcome is equal to the outcome of the data, the energy attribute of the agent may be increased.
  • Figure 3 depicts a data processing event process flow.
  • the data processing event accesses both data 303 and algorithm 302 of agent 301 .
  • Algorithm 302 is applied to the features of data 303 to generate a predicted outcome, which is then compared to the outcome of data 303 at step 305. In the affirmative case of a comparison result, attribute 306 of agent 301 is modified thereby as a result of the comparison. [0070] VISUALIZATION MODULE
  • the invention may comprise a visualization module to depict a visualization of the simulation.
  • the visualization module comprises both a visualizer module, comprising software for converting data collected by the data collector into a graphical depiction.
  • the visualizer further comprises a hardware display device, to display the graphical depiction, e.g., a liquid crystal display, light emitting diode display, cathode ray tube display, electroluminescent display, electronic paper display, etc.
  • the visualizer causes grid coordinates to be translated to screen coordinates, displaying a visual depiction of the contents of grid cells in a corresponding location on screen.
  • Step 401 comprises creating a first agent and a second agent within a digital environment.
  • the first agent comprises a first algorithm capable of processing data and an agent attribute.
  • the second agent comprises a second algorithm capable of processing data.
  • the first agent may be associated with a digital spatial coordinate.
  • Step 402 comprises generating a predicted outcome based on applying the first algorithm to a feature of a set of data, wherein the set of data comprises the feature and an outcome.
  • Step 403 comprises comparing the predicted outcome to the outcome of the set of data.
  • Step 404 comprises modifying an attribute of the first agent based on the comparison.
  • Step 405 comprises combining, based on the modified attribute of the first agent, the first and the second algorithms to create a third algorithm.
  • steps 402, 403, 404, and 405 may be caused by a scheduler activating a first behavior of the first agent.
  • the first agent which may be associated with a digital spatial coordinate, may modify the digital spatial coordinate according to a behavior associated with the first agent.
  • the agent attribute of the first agent may be an energy attribute. The energy attribute may be added by the modification in step 404.
  • Step 501 comprises creating an agent within a digital environment, wherein the agent comprises an algorithm capable of processing data, and wherein the algorithm comprises algorithmic components.
  • Step 502 comprises processing a set of data comprising a feature and an outcome with the algorithm, thereby generating a predicted outcome.
  • Step 503 comprises comparing the predicted outcome to the outcome of the set of data.
  • Step 504 comprises determining, based on the comparison result, whether to generate, based on the algorithmic components and a mutation parameter, a second algorithm capable of processing data.
  • Step 504 may further comprise determining, based on the comparison result, whether to modify an attribute of the first agent, and determining, based on the attribute of the first agent, whether to generate, based on the algorithmic components and the mutation parameter, the second algorithm.
  • the process depicted in Figure 5 may further comprise generating, based on the algorithmic components and the mutation parameter, the second algorithm capable of processing data.
  • the set of data of step 502 may be associated with digital spatial coordinates.
  • another agent may comprise the set of data and digital spatial coordinates.
  • Step 601 comprises creating a first and second agent within a digital environment, wherein the first agent comprises a first algorithm capable of processing data and an agent attribute, and the second agent comprises a second algorithm capable of processing data.
  • Step 602 comprises processing a set of data comprising a feature and an outcome with the first algorithm, thereby generating a predicted outcome.
  • Step 603 comprises comparing the predicted outcome to the outcome of the set of data.
  • Step 604 comprises determining, based on the comparison result, whether to modify an attribute of the first agent.
  • Step 605 comprises determining, based on a set of factors comprising the attribute of the first agent, whether to generate a third algorithm based on the first algorithm.
  • process of Figure 6 can additionally include generating the third algorithm based on the first algorithm.
  • a computer configured to implement the process of Figure 6 can be capable also of generating the third algorithm based on the first algorithm, the second algorithm, and a mutation parameter.
  • the digital environment may comprise digital spatial coordinates capable of being associated with agents.
  • the set of data may be associated with a set of digital spatial coordinates.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Biophysics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Physiology (AREA)
  • Genetics & Genomics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

La présente invention comprend un procédé de simulation d'un environnement écologique, des agents numériques à l'intérieur de l'environnement pouvant traiter des données, et des agents qui traitent avec succès des données étant autorisés à se reproduire pour générer de nouveaux algorithmes. L'invention constitue une avancée novatrice dans l'intelligence artificielle et l'apprentissage automatique et permet des processus auparavant considérés impossibles du point de vue informatique.
PCT/US2017/056218 2017-10-11 2017-10-11 Procédés de génération automatique de modèles précis en temps réduit WO2019074504A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/US2017/056218 WO2019074504A1 (fr) 2017-10-11 2017-10-11 Procédés de génération automatique de modèles précis en temps réduit

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/US2017/056218 WO2019074504A1 (fr) 2017-10-11 2017-10-11 Procédés de génération automatique de modèles précis en temps réduit

Publications (1)

Publication Number Publication Date
WO2019074504A1 true WO2019074504A1 (fr) 2019-04-18

Family

ID=66101605

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2017/056218 WO2019074504A1 (fr) 2017-10-11 2017-10-11 Procédés de génération automatique de modèles précis en temps réduit

Country Status (1)

Country Link
WO (1) WO2019074504A1 (fr)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6081766A (en) * 1993-05-21 2000-06-27 Axys Pharmaceuticals, Inc. Machine-learning approach to modeling biological activity for molecular design and to modeling other characteristics
US20070088534A1 (en) * 2005-10-18 2007-04-19 Honeywell International Inc. System, method, and computer program for early event detection
WO2008027912A2 (fr) * 2006-08-28 2008-03-06 Dan Theodorescu Prédiction de l'activité d'agents sur différents types de cellules et de tissus
US20160189558A1 (en) * 2014-12-31 2016-06-30 Genesys Telecommunications Laboratories, Inc. Learning Based on Simulations of Interactions of a Customer Contact Center

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6081766A (en) * 1993-05-21 2000-06-27 Axys Pharmaceuticals, Inc. Machine-learning approach to modeling biological activity for molecular design and to modeling other characteristics
US20070088534A1 (en) * 2005-10-18 2007-04-19 Honeywell International Inc. System, method, and computer program for early event detection
WO2008027912A2 (fr) * 2006-08-28 2008-03-06 Dan Theodorescu Prédiction de l'activité d'agents sur différents types de cellules et de tissus
US20160189558A1 (en) * 2014-12-31 2016-06-30 Genesys Telecommunications Laboratories, Inc. Learning Based on Simulations of Interactions of a Customer Contact Center

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
STEVEN L. PECK: "Simulation as experiment: a philosophical reassessment for biological modeling", TRENDS IN ECOLOGY AND EVOLUTION, vol. 19, no. 10, October 2004 (2004-10-01), pages 530 - 534, XP004570792, Retrieved from the Internet <URL:https://www.sciencedirect.com/science/article/pii/S0169534704002162> DOI: doi:10.1016/j.tree.2004.07.019 *

Similar Documents

Publication Publication Date Title
US20210350056A1 (en) Method and system for quantum computing
Wang In search of denotational mathematics: Novel mathematical means for contemporary intelligence, brain, and knowledge sciences
Li et al. Development and investigation of efficient artificial bee colony algorithm for numerical function optimization
Fister et al. Memetic artificial bee colony algorithm for large-scale global optimization
Huizinga et al. Evolving neural networks that are both modular and regular: Hyperneat plus the connection cost technique
Shen et al. A bi-population EDA for solving the no-idle permutation flow-shop scheduling problem with the total tardiness criterion
Agarwal et al. Improvement heuristic for the flow-shop scheduling problem: An adaptive-learning approach
Fox et al. Parametrized hierarchical procedures for neural programming
Alcan et al. A genetic algorithm application using fuzzy processing times in non-identical parallel machine scheduling problem
WO2022153290A1 (fr) Modélisation de circuit quantique
Aguilar et al. A near-optimal control method for stochastic boolean networks
Aguilar-Justo et al. A local cooperative approach to solve large-scale constrained optimization problems
Nápoles et al. A computational tool for simulation and learning of Fuzzy Cognitive Maps
WO2015044655A1 (fr) Génération d&#39;un programme d&#39;ordinateur
Jawaddi et al. Integrating OpenAI Gym and CloudSim Plus: A simulation environment for DRL Agent training in energy-driven cloud scaling
EP3764250A2 (fr) Appareil d&#39;optimisation, procédé de commande d&#39;appareil d&#39;optimisation et programme d&#39;optimisation
US10510010B1 (en) Methods for automatically generating accurate models in reduced time
Wu et al. Regret-based multi-agent coordination with uncertain task rewards
WO2019074504A1 (fr) Procédés de génération automatique de modèles précis en temps réduit
Vranjković et al. Coarse-grained reconfigurable hardware accelerator of machine learning classifiers
Chaves et al. Qualitative modeling, analysis and control of synthetic regulatory circuits
EP3834094A1 (fr) Conception assistée par ordinateur de noyaux de réseaux cellulaires personnalisés, selon des propriétés de matériaux
Amaya et al. Neurorobotic reinforcement learning for domains with parametrical uncertainty
Kuzmin et al. Hierarchical reinforcement learning with options and united neural network approximation
Samarasinghe et al. Grammar‐based autonomous discovery of abstractions for evolution of complex multi‐agent behaviours

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17928202

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17928202

Country of ref document: EP

Kind code of ref document: A1