-
公开(公告)号：US11720071B2
公开(公告)日：2023-08-08
申请号：US17815541
申请日：2022-07-27
Applicant: Microsoft Technology Licensing, LLC
Inventor： Damian Silvio Steiger , Helmut Gottfried Katzgraber , Matthias Troyer , Christopher Anand Pattison
IPC: G05B13/04 , G06F30/25 , G06F111/10
CPC classification number: G05B13/042 , G06F30/25 , G06F2111/10
Abstract: A computing device is provided, including memory storing a cost function of a plurality of variables. The computing device may further include a processor configured to, for a stochastic simulation algorithm, compute a control parameter upper bound. The processor may compute a control parameter lower bound. The processor may compute a plurality of intermediate control parameter values within a control parameter range between the control parameter lower bound and the control parameter upper bound. The processor may compute an estimated minimum or an estimated maximum of the cost function using the stochastic simulation algorithm with the control parameter upper bound, the control parameter lower bound, and the plurality of intermediate control parameter values. A plurality of copies of the cost function may be simulated with a respective plurality of seed values.
-
公开(公告)号：US11562273B2
公开(公告)日：2023-01-24
申请号：US16272851
申请日：2019-02-11
Applicant: Microsoft Technology Licensing, LLC
Abstract: A computing device, including memory, an accelerator device, and a processor. The processor may generate a plurality of data packs that each indicate an update to a variable of one or more variables of a combinatorial cost function. The processor may transmit the plurality of data packs to the accelerator device. The accelerator device may, for each data pack, retrieve a variable value of the variable indicated by the data pack and generate an updated variable value. The accelerator device may generate an updated cost function value based on the updated variable value. The accelerator device may be further configured to determine a transition probability using a Monte Carlo algorithm and may store the updated variable value and the updated cost function value with the transition probability. The accelerator device may output a final updated cost function value to the processor.
-
公开(公告)号：US20210065037A1
公开(公告)日：2021-03-04
申请号：US16446511
申请日：2019-06-19
Applicant: Microsoft Technology Licensing, LLC
Inventor： Nathan O. Wiebe , Alexei Bocharov , Paul Smolensky , Matthias Troyer , Krysta Svore
Abstract: Embodiments of a new approach for training a class of quantum neural networks called quantum Boltzmann machines are disclosed. in particular examples, methods for supervised training of a quantum Boltzmann machine are disclosed using an ensemble of quantum states that the Boltzmann machine is trained to replicate. Unlike existing approaches to Boltzmann training, example embodiments as disclosed herein allow for supervised training even in cases where only quantum examples are known (and not probabilities from quantum measurements of a set of states). Further, this approach does not require the use of approximations such as the Golden-Thompson inequality.
-
公开(公告)号：US12124452B1
公开(公告)日：2024-10-22
申请号：US18321415
申请日：2023-05-22
Applicant: Microsoft Technology Licensing, LLC
Inventor： Chi Chen , Hongbin Liu , Andrea Cepellotti , Mark A Woodlief , Nihit Pokhrel , Adrian Dumitrascu , Matthias Troyer , Nathan Andrew Baker
IPC: G06F16/24 , G06F16/2453 , G06F16/2455 , G06F16/248
CPC classification number: G06F16/2455 , G06F16/24542 , G06F16/248
Abstract: Examples are disclosed that relate to materials discovery using machine learning models. One example provides a method enacted on a computing system. The method comprises receiving a query comprising one or more of element information and material property information, and, based on the query, retrieving material data from a materials information database. The material data comprises structural information for each material within a set of materials matching the query, the set comprising one or more materials, and for one or more materials in the set of materials, one or more predicted material properties determined using one or more trained machine learning models. The method further comprises outputting the material data.
-
公开(公告)号：US11694103B2
公开(公告)日：2023-07-04
申请号：US16530916
申请日：2019-08-02
Applicant: Microsoft Technology Licensing, LLC
Inventor： Matthias Troyer , David Poulin , Bettina Heim , Jessica Lemieux
CPC classification number: G06N10/00 , G06F18/295 , G06F30/20
Abstract: Example circuit implementations of Szegedy's quantization of the Metropolis-Hastings walk are presented. In certain disclosed embodiments, a quantum walk procedure of a Markov chain Monte Carlo simulation is implemented in which a quantum move register is reset at every step in the quantum walk. In further embodiments, a quantum walk procedure of a Markov chain Monte Carlo simulation is implemented in which an underlying classical walk is obtained using a Metropolis-Hastings rotation or a Glauber dynamics rotation. In some embodiments, a quantum walk procedure is performed in the quantum computing device to implement a Markov Chain Monte Carlo method; during the quantum walk procedure, an intermediate measurement is obtained; and a rewinding procedure of one or more but not all steps of the quantum walk procedure is performed if the intermediate measurement produces an incorrect outcome.
-
Patent Agency Ranking
公开(公告)号：US11630703B2
公开(公告)日：2023-04-18
申请号：US16743386
申请日：2020-01-15
Applicant: Microsoft Technology Licensing, LLC
Abstract: A computing device is provided, including a cluster update accelerator circuit configured to receive signals encoding a combinatorial cost function of a plurality of variables and a connectivity graph for the combinatorial cost function. In an energy sum phase, the cluster update accelerator circuit may determine a respective plurality of accumulated energy change values for the combinatorial cost function based at least in part on the connectivity graph. In an update phase, the cluster update accelerator circuit may determine a respective update indicator bit for each accumulated energy change value. In an encoder phase, based on the plurality of update indicator bits, the cluster update accelerator circuit may select a largest update-indicated cluster of the variables included in the connectivity graph. The cluster update accelerator circuit may output an instruction to update the variables included in the largest update-indicated cluster.
-
公开(公告)号：US11922337B2
公开(公告)日：2024-03-05
申请号：US18157339
申请日：2023-01-20
Applicant: Microsoft Technology Licensing, LLC
CPC classification number: G06N7/01 , G06F7/582 , G06F9/3877 , G06F17/11
Abstract: A computing device, including memory, an accelerator device, and a processor. The processor may generate a plurality of data packs that each indicate an update to a variable of one or more variables of a combinatorial cost function. The processor may transmit the plurality of data packs to the accelerator device. The accelerator device may, for each data pack, retrieve a variable value of the variable indicated by the data pack and generate an updated variable value. The accelerator device may generate an updated cost function value based on the updated variable value. The accelerator device may be further configured to determine a transition probability using a Monte Carlo algorithm and may store the updated variable value and the updated cost function value with the transition probability. The accelerator device may output a final updated cost function value to the processor.
-
公开(公告)号：US11783222B2
公开(公告)日：2023-10-10
申请号：US16446511
申请日：2019-06-19
Applicant: Microsoft Technology Licensing, LLC
Inventor： Nathan O. Wiebe , Alexei Bocharov , Paul Smolensky , Matthias Troyer , Krysta Svore
Abstract: A method of training a quantum computer employs quantum algorithms. The method comprises loading, into the quantum computer, a description of a quantum Boltzmann machine, and training the quantum Boltzmann machine according to a protocol, wherein a classification error is used as a metric for the protocol.
-
公开(公告)号：US20200090072A1
公开(公告)日：2020-03-19
申请号：US16530916
申请日：2019-08-02
Applicant: Microsoft Technology Licensing, LLC
Inventor： Matthias Troyer , David Poulin , Bettina Heim , Jessica Lemieux
Abstract: Example circuit implementations of Szegedy's quantization of the Metropolis-Hastings walk are presented. In certain disclosed embodiments, a quantum walk procedure of a Markov chain Monte Carlo simulation is implemented in which a quantum move register is reset at every step in the quantum walk. In further embodiments, a quantum walk procedure of a Markov chain Monte Carlo simulation is implemented in which an underlying classical walk is obtained using a Metropolis-Hastings rotation or a Glauber dynamics rotation. In some embodiments, a quantum walk procedure is performed in the quantum computing device to implement a Markov Chain Monte Carlo method; during the quantum walk procedure, an intermediate measurement is obtained; and a rewinding procedure of one or more but not all steps of the quantum walk procedure is performed if the intermediate measurement produces an incorrect outcome.
-
公开(公告)号：US20170161612A1
公开(公告)日：2017-06-08
申请号：US14961605
申请日：2015-12-07
Applicant: Microsoft Technology Licensing, LLC
Inventor： Matthew B. Hastings , Nathan Wiebe , Ilia Zintchenk , Matthias Troyer
CPC classification number: G06N5/022 , G06F17/11 , G06F17/50 , G06F2217/08 , G06N20/00
Abstract: In some examples, techniques and architectures for solving combinatorial optimization or statistical sampling problems use a recursive hierarchical approach that involves reinitializing various subsets of a set of variables. The entire set of variables may correspond to a first level of a hierarchy. In individual steps of the recursive process of solving an optimization problem, the set of variables may be partitioned into subsets corresponding to higher-order levels of the hierarchy, such as a second level, a third level, and so on. Variables of individual subsets may be randomly initialized. Based on the objective function, a combinatorial optimization operation may be performed on the individual subsets to modify variables of the individual subsets. Reinitializing subsets of variables instead of reinitializing the entire set of variables may allow for preservation of information gained in previous combinatorial optimization operations.
-
-
-
-
-
-
-
-
-
Search Results
10
records
(took 0.017 seconds)
