Method of machine learning and information processing apparatus

    公开(公告)号:US12039275B2

    公开(公告)日:2024-07-16

    申请号:US17683406

    申请日:2022-03-01

    申请人: FUJITSU LIMITED

    IPC分类号: G06F40/295 G06N20/00

    CPC分类号: G06F40/295 G06N20/00

    摘要: A non-transitory computer-readable recording medium stores a program for causing a computer to execute a process, the process includes identifying first named entities and first verbs in a dependent relationship respectively with the first named entities from each of a plurality of sentences, vectorizing each of the plurality of sentences based on the first named entities and the first verbs, identifying, out of the plurality of sentences, one or more sentences each having a similarity greater than or equal to a threshold with respect to a specific sentence based on a plurality of vectors generated by the vectorizing, the similarity between two sentences being defined as a similarity between vectors generated by vectorizing the two sentences, and executing training of a machine learning model based on the one or more sentences.

    Optimization apparatus and optimization method

    公开(公告)号:US12039233B2

    公开(公告)日:2024-07-16

    申请号:US17183408

    申请日:2021-02-24

    申请人: FUJITSU LIMITED

    摘要: A method includes: accessing first storage configured to store a first weight coefficient group which is at least some of a plurality of weight coefficients indicating a magnitude of interaction between a plurality of state variables in an evaluation function representing energy of an Ising model; accessing a plurality of second storages each of the plurality of second storage being configured to store a second weight coefficient group related to a state variable having a value of 1 in any of a plurality of state variable groups respectively including the plurality of state variables among the plurality of weight coefficients; outputting, for each of the plurality of state variable groups, a search result obtained by performing searching processing configured to perform processing of searching for an optimum solution by repeatedly performing a first update process with a first constraint or a second update process with a second constraint.

    Ethics-based multi-modal user post monitoring

    公开(公告)号:US12001489B1

    公开(公告)日:2024-06-04

    申请号:US18159103

    申请日:2023-01-25

    申请人: Fujitsu Limited

    摘要: In an embodiment, a set of user posts and associated metadata are received. A parameterized logistic function is fit on the metadata. Each user post associated with the corresponding metadata is categorized, based on the fitted parameterized logistic function. A category-prior associated with each user post is determined, based on the clustered set of user posts. A set of multi-modal features associated with each user post is determined. Further, a feature likelihood for each user post is determined, based on the set of multi-modal features. A final category associated with each posts is determined, based on the category-prior and on the feature likelihood. Post categorization information is rendered based on the final category associated with each user post. The set of user posts is filtered based on the final category of each user post.

    Computer-readable recording medium storing position identification program, position identification method, and information processing apparatus

    公开(公告)号:US11967108B2

    公开(公告)日:2024-04-23

    申请号:US17475982

    申请日:2021-09-15

    申请人: FUJITSU LIMITED

    发明人: Shun Kohata

    摘要: A recording medium stores a program for causing a computer to execute processing including acquiring a captured image and a depth image that represents a distance from an imaging position, identifying a road region and another region in contact with the road region from the captured image, calculating a change in a depth of a first region that corresponds to the road region and a change in a depth of a second region that corresponds to the another region included in the depth image, determining whether or not the another region is a detection target based on the change in the depth of the first region and the change in the depth of the second region, and identifying a position of a subject included in the another region based on the depth of the second region and the imaging position when the another region is the detection target.

    Storage medium, pattern search device, and pattern search method

    公开(公告)号:US11960491B2

    公开(公告)日:2024-04-16

    申请号:US17676926

    申请日:2022-02-22

    申请人: FUJITSU LIMITED

    发明人: Hiroaki Iwashita

    摘要: A non-transitory computer-readable storage medium storing a pattern search program that causes a computer to execute a process, the process includes designating priority ranks of each of attributes included in a plurality of attribute patterns that indicate one or more attributes, based on an inclusion relationship between occurrence sets of a plurality of samples with respect to each of the attributes included in the plurality of attribute patterns; and determining whether each of the plurality of attribute patterns is an emerging pattern according to a search order based on the priority ranks, when a frequency of occurrence of the plurality of samples in a first attribute pattern is less than the frequency of occurrence of the plurality of samples in a second attribute pattern that has all the attributes except one of the attributes with lowest one of the priority ranks among the attributes included in the first attribute pattern.

    Quantizer and optical transmitter
    19.
    发明授权

    公开(公告)号:US11956018B2

    公开(公告)日:2024-04-09

    申请号:US17899663

    申请日:2022-08-31

    申请人: FUJITSU LIMITED

    IPC分类号: H04B10/50 H04B10/40 H04B10/60

    CPC分类号: H04B10/50 H04B10/40 H04B10/60

    摘要: A quantizer includes: a clipper configured to clip a portion exceeding a quantization range of a sample value sampled at a predetermined rate; and a noise shaper configured to determine a plurality of candidates for a quantization level based on the clipped sample value, and outputs, as a quantization value, a value obtained by adding a minimum noise in which a noise in a low-frequency region is minimum among noises generated in each candidate to the sample value before clipping.