# US 12,602,597 B2 · Automatic discovery of machine learning model features > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,602,597 B2 - Document ID: US12602597B2 - Status: Granted - Kind: B2 - Application number: 17/486,770 - Filed: 2021-09-27 - Published: 2023-03-30 - Granted: 2026-04-14 - Earliest claimed: 2021-09-27 - Assignees: AT&T Intellectual Property I, L.P. - Applicants: AT&T Intellectual Property I, L.P. - Inventors: Jeremy Fix; Cuong Vo; Eric Zavesky; Abhay Dabholkar; Rudolph Mappus; James Pratt - Web page: https://patentgrep.com/patent/US12602597B2 ## CPC - G06N 5/022: Knowledge engineering; Knowledge acquisition - G06N 3/08: Learning methods - G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks - G06N 20/00: Machine learning - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract A method performed by a processing system including at least one processor includes monitoring interactions of a human user with a platform for building machine learning models, where the human user is using the platform to build a new machine learning model, detecting, within the interactions, an event that triggers a suggestion feature, performing a search of a data source for existing data from existing machine learning models which can be reused to build the new machine learning model, using information about the event, presenting a suggestion to the human user to reuse a portion of the existing data discovered in the search in the new machine learning model, receiving a user feedback in response to the suggestion, and generating an updated suggestion in response to the user feedback. ## Claims ### Claim 1 A method comprising: monitoring, by a processing system including at least one processor, interactions of a human user with a platform for building machine learning models, where the human user is using the platform to build a new machine learning model; detecting, by the processing system within the interactions, an event that triggers a suggestion feature; performing, by the processing system, a search of a data source for existing data from existing machine learning models which can be reused to build the new machine learning model, using information about the event, wherein a search parameter for the search is based on a current context of a building process for building the new machine learning model; presenting, by the processing system, a suggestion to the human user to reuse a portion of the existing data discovered in the search in the new machine learning model; receiving, by the processing system, a user feedback in response to the suggestion; and generating, by the processing system, an updated suggestion in response to the user feedback. ### Claim 2 (depends on 1) The method of claim 1, wherein the event comprises an input provided by the human user to the platform for building the new machine learning model. ### Claim 3 (depends on 2) The method of claim 2, wherein the input comprises a set of metadata describing the new machine learning model. ### Claim 4 (depends on 3) The method of claim 3, wherein the set of metadata describes at least one of: a feature the new machine learning model will take as input, a target of the new machine learning model, one or more performance criteria of the new machine learning model, a source domain of the new machine learning model, or a target domain of the new machine learning model. ### Claim 5 (depends on 2) The method of claim 2, wherein the input comprises a set of notes for building the new machine learning model. ### Claim 6 (depends on 2) The method of claim 2, wherein the input comprises code for building the new machine learning model. ### Claim 7 (depends on 2) The method of claim 2, wherein the input comprises a set of training data for training the new machine learning model. ### Claim 8 (depends on 1) The method of claim 1, wherein the existing machine learning models are directed to a similar domain space as the new machine learning model. ### Claim 9 (depends on 1) The method of claim 1, wherein the existing machine learning models are directed to a similar functional space as the new machine learning model. ### Claim 10 (depends on 1) The method of claim 1, wherein the search parameter for the search is further based on an input from the human user which triggered the event. ### Claim 11 (depends on 1) The method of claim 1, wherein metadata describing the new machine learning model matches metadata describing the existing data from the existing machine leaning models. ### Claim 12 (depends on 1) The method of claim 1, wherein individual items of the portion of the existing data are ranked. ### Claim 13 (depends on 12) The method of claim 12, wherein the individual items of the portion of the existing data are ranked according to at least one of: an assessed relevance to inputs from the human user which triggered the event, reuse of the individual items of the portion of the existing data by other users, a target performance metric for the new machine learning model, or a performance metric of an existing machine learning model from which the individual items are extracted. ### Claim 14 (depends on 12) The method of claim 12, wherein the individual items include features to be added to a set of training data for the new machine learning model, and the features are ranked according to magnitudes of effects that an addition of the features are expected to have on a performance of the new machine learning model. ### Claim 15 (depends on 1) The method of claim 1, wherein the user feedback comprises a selection by the human user of a suggested existing machine learning model or a feature of an existing machine learning model for reuse in the new machine learning model. ### Claim 16 (depends on 1) The method of claim 1, wherein the user feedback comprises the human user ignoring the suggestion. ### Claim 17 (depends on 1) The method of claim 1, wherein the user feedback comprises the human user selecting a graphical user interface feature associated with the suggestion. ### Claim 18 (depends on 1) The method of claim 1, further comprising: recording the user feedback. ### Claim 19 A non-transitory computer-readable medium storing instructions which, when executed by a processing system including at least one processor, cause the processing system to perform operations, the operations comprising: monitoring interactions of a human user with a platform for building machine learning models, where the human user is using the platform to build a new machine learning model; detecting, within the interactions, an event that triggers a suggestion feature; performing a search of a data source for existing data from existing machine learning models which can be reused to build the new machine learning model, using information about the event, wherein a search parameter for the search is based on a current context of a building process for building the new machine learning model; presenting a suggestion to the human user to reuse a portion of the existing data discovered in the search in the new machine learning model; receiving a user feedback in response to the suggestion; and generating an updated suggestion in response to the user feedback. ### Claim 20 A device comprising: a processing system including at least one processor; and a non-transitory computer-readable medium storing instructions which, when executed by the processing system, cause the processing system to perform operations, the operations comprising: monitoring interactions of a human user with a platform for building machine learning models, where the human user is using the platform to build a new machine learning model; detecting, within the interactions, an event that triggers a suggestion feature; performing a search of a data source for existing data from existing machine learning models which can be reused to build the new machine learning model, using information about the event, wherein a search parameter for the search is based on a current context of a building process for building the new machine learning model; presenting a suggestion to the human user to reuse a portion of the existing data discovered in the search in the new machine learning model; receiving a user feedback in response to the suggestion; and generating an updated suggestion in response to the user feedback.