Searchable database of trained artificial intelligence objects that can be reused, reconfigured, and recomposed, into one or more subsequent artificial intelligence models

• US 10,586,173 B2
• Filed: 01/26/2017
• Issued: 03/10/2020
• Est. Priority Date: 01/27/2016
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Bonsai AI, Inc. (Microsoft Corporation)
• Inventors: Hammond, Mark Isaac, Browne, Keen McEwan, Campos, Marcos, Brown, Matthew James, Kong, Ruofan, Guss, William, Story, Ross
• Primary Examiner(s): Hicks, Austin
• Application Number:
  US15/417,075
• Publication Number:
  US 20170213155A1
• Time in Patent Office: 1,139 Days
• CPC Class Codes: G06F 15/80 comprising an array of processing units with common control, e.g. single instruction multiple data processors G06F15/82 takes precedence ; for correlation function computation G06F17/15
• CPC Class Codes: G06F 16/2228 Indexing structures
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 18/2148 characterised by the process organisation or structure, e.g. boosting cascade
• CPC Class Codes: G06F 3/03543 Mice or pucks G06F3/03541 takes precedence
• CPC Class Codes: G06F 3/0482 Interaction with lists of selectable items, e.g. menus
• CPC Class Codes: G06F 30/20 Design optimisation, verification or simulation optimisation, verification or simulation of circuit designs G06F30/30
• CPC Class Codes: G06F 30/333 Design for testability [DFT], e.g. scan chain or built-in self-test [BIST]
• CPC Class Codes: G06F 40/166 Editing, e.g. inserting or deleting
• CPC Class Codes: G06F 8/31 Programming languages or programming paradigms
• CPC Class Codes: G06F 8/311 Functional or applicative languages; Rewrite languages
• CPC Class Codes: G06F 8/38 for implementing user interfaces
• CPC Class Codes: G06F 8/437 Type checking
• CPC Class Codes: G06F 9/451 Execution arrangements for user interfaces
• CPC Class Codes: G06F 9/4881 Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues
• CPC Class Codes: G06F 9/5066 Algorithms for mapping a plurality of inter-dependent sub-tasks onto a plurality of physical CPUs mappping at compile time, see G06F8/451
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/008 based on physical entities controlled by simulated intelligence so as to replicate intelligent life forms, e.g. based on robots replicating pets or humans in their appearance or behaviour
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/08 : Learning methods
• CPC Class Codes: G06N 3/105 : Shells for specifying net layout
• CPC Class Codes: G06N 5/04 : Inference or reasoning models
• CPC Class Codes: G06Q 10/00 : Administration; Management
• CPC Class Codes: H04L 67/01 : Protocols
• CPC Class Codes: H04L 67/02 : based on web technology, e.g. hypertext transfer protocol [HTTP]

Abstract

An AI database hosted on cloud platform is configured to cooperate with a search engine and an AI engine. The AI database stores and indexes trained AI objects and its class of AI objects have searchable criteria. The AI database cooperates with the search engine to utilize search criteria supplied from a user, from either or both 1) via scripted software code and 2) via data put into defined fields of a user interface. The search engine utilizes the search criteria in order for the search engine to retrieve one or more AI data objects that have already been trained as query results. The AI database is coupled to an AI engine to allow any of reuse, reconfigure ability, and recomposition of the one or more trained AI data objects from the AI database into a new trained AI model.

Claims

1. An artificial intelligence (“

AI”

) database hosted on cloud platform and configured to cooperate with a search engine, comprising;

where the AI database is configured to store and index trained AI objects and its class of AI objects have searchable criteria, where the AI database is configured to cooperate with the search engine to utilize search criteria supplied from a user, from either or both

     1) via scripted software code and

     2) via data put into defined fields of a user interface, in order for the search engine to retrieve one or more AI data objects that have already been trained as query results, where the AI database is coupled to an AI engine to allow any of reuse, reconfigure ability, and recomposition of the one or more trained AI data objects into a new trained AI model, where reuse, reconfigure ability, and recomposition of the one or more trained AI data objects improves the functioning of the AI engine by reducing an amount of memory used and processing cycles used as compared to start training an AI model from scratch, which has not previously been through training sessions to be trained.

2. The AI database configured to cooperate with the search engine of claim 1,

where the search engine is a component within the AI database of the AI system, where the AI database further consists of a storage layer which is configured to know how to efficiently store database objects, in this case AI objects, an indexing mechanism to speed retrieval of the stored AI objects, where the search engine is configured to translate a query request into a retrieval strategy to retrieve AI objects that satisfy a query, and a query language which describes to the AI database what AI objects are desired to be retrieved.

3. The AI database configured to cooperate with the search engine of claim 1,

a signature module is configured to calculate a signature for a mental model of concepts in a first trained AI model, where the signature is a form of hashing such that mental models of different trained AI models that have similar machine learning algorithmic characteristics have similar signatures.

4. The AI database configured to cooperate with the search engine of claim 3,

where the search engine utilizes the user supplied criteria to query for relevant trained AI objects by referring to 1) the signatures of the stored AI objects as well as 2) any indexed parameters for the AI objects indexed by the AI database.

5. The AI database configured to cooperate with the search engine of claim 1,

where the AI database is composed of a set of one or more databases in which each database has a different profile and indexing, where the set of databases are configured to operate in a parallel to then send back accurate, fast, and efficient returns of trained AI objects that satisfy the search query.

6. The AI database configured to cooperate with the search engine of claim 1,

where each trained artificial intelligence model itself is a collection of trained AI objects corresponding to a main concept and a set of sub concepts feeding parameters into the main concept, where the AI database indexes AI objects corresponding to the main concept and the set of sub concepts making up a given trained artificial intelligence model so that reuse, recomposition, and reconfiguration of all or part of a trained artificial intelligence model is possible.

7. The AI database configured to cooperate with the search engine of claim 1, further comprising:

a user interface to the AI database and search engine that is configured to present a population of known trained AI objects and the search engine cooperating with the AI database is configured to search the population of known trained AI objects to return a set of one or more already trained AI objects similar to a problem trying to be solved by the user supplying the search criteria.

8. The AI database configured to cooperate with the AI engine to supply one or more AI objects of claim 1, wherein the AI engine includes

an architect module configured to create and optimize learning topologies of neural networks for the AI objects;

an instructor module configured to carrying out a training plan codified in a pedagogical software programming language; and

a learner module configured to carrying out an actual execution of underlying AI learning algorithms during a training session, where the architect module, when reconfiguring or recomposing the AI objects, composes the one or more trained AI data objects into a new AI model and the instructor module and learner module cooperate with one or more data sources to train the new AI model.

9. The AI database configured to cooperate with the search engine of claim 1,

where the search engine is configured to import the data put into defined fields of the user interface to use as searchable criteria to find relevant trained AI objects indexed in the AI database.

10. The AI database configured to cooperate with the search engine of claim 9,

where the search engine is also configured to parse scripted software code written in a pedagogical software programming language and then map that to one or more searchable criteria.

11. A method of causing an artificial intelligence (“

AI”

) database to cooperate with a search engine, comprising;

causing the AI database to store and index trained AI objects and its class of AI objects to have searchable criteria; and

utilizing search criteria supplied from a user, from either or both

     1) via scripted software code and

     2) via data put into defined fields of a user interface, in order for the search engine to retrieve one or more AI data objects that have already been trained as query results, where the AI database cooperates with an AI engine to cause any of reuse, reconfigure ability, and recomposition of the one or more trained AI data objects into a new trained AI model.

12. The method of claim 11, further comprising:

1) parsing scripted software code written in a pedagogical software programming language with the search engine and then mapping that to one or more searchable criteria as well as

     2) importing the data put into defined fields of the user interface to use as searchable criteria to find relevant trained AI objects indexed in the AI database.

13. The method of claim 11, further comprising:

utilizing the user supplied criteria to query for relevant trained AI objects by referring to

     1) the signatures of the stored AI objects as well as

     2) any indexed parameters for the AI objects indexed by the AI database.

14. The method of claim 11, further comprising:

configuring a set of one or more databases in which each database has a different profile and indexing to be the AI database, where the set of databases are configured to operate in a parallel to then send back accurate, fast, and efficient returns of trained AI objects that satisfy the search query.

15. The method of claim 11, further comprising:

storing and indexing parts of a trained artificial intelligence model as a collection of trained AI objects corresponding to a main concept and a set of sub concepts feeding parameters into the main concept so that reuse, recomposition, and reconfiguration of all or part of a trained artificial intelligence model is possible.

16. The method of claim 11, further comprising:

configuring the AI database to cooperate with the AI engine to supply one or more AI objects, wherein the AI engine includes

an architect module configured to create and optimize learning topologies of neural networks for the AI objects;

an instructor module configured to carrying out a training plan codified in a pedagogical software programming language; and

a learner module configured to carrying out an actual execution of underlying artificial intelligence learning algorithms during a training session, where the architect module, when reconfiguring or recomposing the AI objects, composes the one or more trained AI data objects into a new AI model and the instructor module and learner module cooperate with one or more data sources to train the new AI model.

17. The method of claim 11, further comprising:

presenting a population of known trained AI objects; and

searching the population of known trained AI objects to return a set of one or more already trained AI objects similar to a problem trying to be solved by the user supplying the search criteria.

18. A non-transitory machine-readable medium configured to store data and instructions, which, when executed by one or more processors on a computing device, causes the following operations, comprising:

causing an artificial intelligence (“

AI”

) database to cooperate with a search engine;

causing the AI database to store and index trained AI objects and its class of AI objects to have searchable criteria; and

utilizing search criteria supplied from a user, from either or both

     1) via scripted software code and

     2) via data put into defined fields of a user interface, in order for the search engine to retrieve one or more AI data objects that have already been trained as query results, where the AI database cooperates with an AI engine to cause any of reuse, reconfigure ability, and recomposition of the one or more trained AI data objects into a new trained AI model.

19. The non-transitory machine-readable medium of claim 18 storing instructions to cause further following operations, comprising:

either or both

     1) parsing scripted software code written in a pedagogical software programming language with the search engine and then mapping that to one or more searchable criteria as well as

     2) importing the data put into defined fields of the user interface to use as searchable criteria to find relevant trained AI objects indexed in the AI database.

20. The non-transitory machine-readable medium of claim 18 storing instructions to cause further following operations, comprising:

calculating a signature for a mental model of concepts in a first trained AI model, where the signature is a form of hashing such that mental models of different trained AI models that have similar machine learning algorithmic characteristics have similar signatures.

21. The non-transitory machine-readable medium of claim 18 storing instructions to cause further following operations, comprising:

utilizing the user supplied criteria to query for relevant trained AI objects by referring to

     1) the signatures of the stored AI objects as well as

     2) any indexed parameters for the AI objects indexed by the AI database.

22. The non-transitory machine-readable medium of claim 18 storing instructions to cause further following operations, comprising:

storing and indexing parts of a trained artificial intelligence model as a collection of trained AI objects corresponding to a main concept and a set of sub concepts feeding parameters into the main concept, where the AI database indexes AI objects corresponding to the main concept and the set of sub concepts making up a given trained artificial intelligence model so that reuse, recomposition, and reconfiguration of all or part of a trained artificial intelligence model is possible.

23. The non-transitory machine-readable medium of claim 18 storing instructions to cause further following operations, comprising:

presenting a population of known trained AI objects; and

searching the population of known trained AI objects to return a set of one or more already trained AI objects similar to a problem trying to be solved by the user supplying the search criteria.