公开(公告)号：US10380498B1

公开(公告)日：2019-08-13

申请号：US14720474

申请日：2015-05-22

Applicant: Amazon Technologies, Inc.

Inventor： Vineet Shashikant Chaoji ,  Aswin Natarajan ,  Seyit Ismail Parsa ,  Rajeev Ramnarain Rastogi

IPC: G06F15/18 ,  G06N20/00 ,  G06N7/00

Abstract: This disclosure is directed to the automated generation of Machine Learning (ML) models. The system receives a user directive containing one or more requirements for building the ML model. The system further identifies common requirements between the user directive and one or more prior user directives and associates characteristics of the prior user directive, or model generated therefrom, with the user directive. The system further associates performance values generated by continuous monitoring of deployed ML models to individual characteristics of the user directive used to generate each of the deployed ML models. The system continuously improves model generation efficiency, model performance, and first run performance of individual ML models by learning from the improvements made to one or more prior ML models having similar characteristics.

公开(公告)号：US10354201B1

公开(公告)日：2019-07-16

申请号：US14990171

申请日：2016-01-07

Applicant: Amazon Technologies, Inc.

Inventor： Gourav Roy ,  Amit Chandak ,  Prateek Gupta ,  Srujana Merugu ,  Aswin Natarajan ,  Sathish Kumar Palanisamy ,  Gowda Dayananda Anjaneyapura Range ,  Jagannathan Srinivasan ,  Bharath Venkatesh

IPC: G06F16/28 ,  G06N20/00

Abstract: A number of attributes of different attribute types, to be used to assign observation records of a data set to clusters, are identified. Attribute-type-specific distance metrics for the attributes, which can be combined to obtain a normalized aggregated distance of an observation record from a cluster representative, are selected. One or more iterations of a selected clustering methodology are implemented on the data set using resources of a machine learning service until targeted termination criteria are met. A given iteration includes assigning the observations to clusters of a current version of a clustering model based on the aggregated distances from the cluster representatives of the current version, and updating the cluster representatives to generate a new version of the clustering model.

公开(公告)号：US10268749B1

公开(公告)日：2019-04-23

申请号：US14990161

申请日：2016-01-07

Applicant: Amazon Technologies, Inc.

Inventor： Gourav Roy ,  Amit Chandak ,  Prateek Gupta ,  Srujana Merugu ,  Aswin Natarajan ,  Sathish Kumar Palanisamy ,  Gowda Dayananda Anjaneyapura Range ,  Jagannathan Srinivasan ,  Bharath Venkatesh

IPC: G06N5/04 ,  G06N99/00 ,  G06F17/30

Abstract: An approximate data structure to represent clusters of observation records of a data set is identified. A hierarchical representation of a plurality of clusters, including the targeted number of clusters among which the observation records are to be distributed, is generated. Each node of the hierarchy comprises an instance of the approximate data structure. Until a set of termination criteria are met, iterations of a selected clustering methodology are run. In a given iteration, distances of observation records from the cluster representatives of a current version of the model are computed using the hierarchical representation, and a new version of the model with modified cluster representatives is generated.

公开(公告)号：US10073892B1

公开(公告)日：2018-09-11

申请号：US14738097

申请日：2015-06-12

Applicant: Amazon Technologies, Inc.

Inventor： Vineet Khare ,  Aswin Natarajan

IPC: G06F17/30

CPC classification number: G06F16/2465 ,  G06F16/26

Abstract: Data mining systems and methods are disclosed for item recommendation based on frequent attribute-values associated with items. The system may determine commonalities in item attribute-values based on user transactions and identify frequent attribute-value tuples that include attribute-values that frequently co-occur in user transactions. The system may associate user interests with the frequent attribute-value tuples and recommend items to target users based thereon. A user-interface for presenting the recommendation allows users to explore item recommendations based on modifications to one or more frequent attribute-value tuples initially recommended to the user