Abstract: In an example embodiment, a skip logic using downsampling is applied to negative signals on a training data set fed to a machine-learning algorithm to train a machine-learned model. By downsampling the negatively labeled pieces of training data, the technical problem encountered in biasing the machine-learned model towards negative cases is overcome.

Abstract: The disclosed embodiments provide a system for performing activity-based inference of title preferences. During operation, the system determines features and labels related to first title preferences for jobs sought by a first set of candidates. Next, the system inputs the features and the labels as training data for a machine learning model. The system then applies the machine learning model to additional features for a second set of candidates to produce predictions of second title preferences for the second set of candidates. Finally, the system stores the predictions in association with the second set of candidates.

Abstract: The disclosed embodiments provide a system for processing data. During operation, the system receives, over a set of event streams, a set of logging events for actions performed between members and jobs over multiple channels. Next, the system aggregates a subset of the logging events spanning a logging window by a reference identifier (ID) generated based on a user session of a member, a first member ID for the member, and a first job ID for a job. The system then creates, based on a unified data logic, a record containing a subset of the actions represented by the logging events and contexts for the subset of the actions. Finally, the system outputs the record for use in subsequent analysis associated with the member and the job.

Abstract: In some embodiments, a computer system selects a first subset of candidate content items based on their filter scores that are generated based on a partial generalized linear mixed model comprising a baseline model and a user-based model, with the baseline model being a generalized linear model, and the user-based model being a random effects model based on user actions by the target user directed towards reference content items related to the candidate content items. In some embodiments, the computer system then selects a second subset from the first subset based on recommendation scores that are generated based on a full generalized linear mixed model comprising the baseline model, the user-based model, and an item-based model, with the item-based model being a random effects model based on user actions directed towards the candidate online content item by reference users related to the target user.

Abstract: Described herein are methods and systems for using weak labels to train a model for use in identifying job listings that are relevant to a user of an online job hosting service. The weak labels correspond with various user actions that a user has undertaken with respect to job listings presented to the user. By way of example, the relevant user actions may include: Job Applies, Job Saves, Job Views, Job Skips and Job Dismisses.

Abstract: The disclosed embodiments provide a system for processing data. During operation, the system receives, over a set of event streams, a set of logging events for actions performed between members and jobs over multiple channels. Next, the system aggregates a subset of the logging events spanning a logging window by a reference identifier (ID) generated based on a user session of a member, a first member ID for the member, and a first job ID for a job. The system then creates, based on a unified data logic, a record containing a subset of the actions represented by the logging events and contexts for the subset of the actions. Finally, the system outputs the record for use in subsequent analysis associated with the member and the job.

Abstract: In some embodiments, a computer system selects a first subset of candidate content items based on their filter scores that are generated based on a partial generalized linear mixed model comprising a baseline model and a user-based model, with the baseline model being a generalized linear model, and the user-based model being a random effects model based on user actions by the target user directed towards reference content items related to the candidate content items. In some embodiments, the computer system then selects a second subset from the first subset based on recommendation scores that are generated based on a full generalized linear mixed model comprising the baseline model, the user-based model, and an item-based model, with the item-based model being a random effects model based on user actions directed towards the candidate online content item by reference users related to the target user.

Abstract: The disclosed embodiments provide a system for selecting recommendations based on title transition embeddings. During operation, the system obtains a word embedding model of a set of job histories. Next, the system calculates similarities between pairs of the embeddings produced by the word embedding model from attributes associated with titles in the set of job histories. The system then identifies, based on the similarities, job titles with high similarity to a current title of the candidate. Finally, the system outputs the job titles for use in selecting job recommendations for the candidate.

Abstract: In some embodiments, a computer system determines that online postings belong to a cohort based on the postings having an attribute of the cohort, identifies skills from the postings, determines that a user belongs to the cohort based on a determination that a profile of the user includes the attribute(s) of the cohort, determines that one or more of the skills is stored in association with the profile, determines a user confidence score that indicates a relevance level of the skill to the user for each one of the one or more of the skills, determines a cohort confidence score for each one of the one or more of the skills based on how many of the postings include the skill, and displays a recommendation associated based on a combination of the user confidence score and the cohort confidence score for at least a portion of the skills.

Abstract: An approach for continual learning in slowly-varying environments is provided. The approach receives one or more action requests from a decision agent. The approach deploys a first model to a decision engine. The approach initiates an observation period. The approach builds a second model, in which the second model comprises collected transaction data from the observation period. The approach initiates a test period. The approach determines a performance score for the first model and a performance score for the second model. The approach selects the model providing an optimized action.

Abstract: The disclosed embodiments provide a system for performing activity-based inference of title preferences. During operation, the system determines features and labels related to first title preferences for jobs sought by a first set of candidates. Next, the system inputs the features and the labels as training data for a machine learning model. The system then applies the machine learning model to additional features for a second set of candidates to produce predictions of second title preferences for the second set of candidates. Finally, the system stores the predictions in association with the second set of candidates.

Abstract: The disclosed embodiments provide a system that ranks job recommendations based on title preferences. During operation, the system determines features related to applications for jobs by a candidate, wherein the features include a title preference for the candidate and a similarity between a first set of attribute values for the candidate and a second set of attribute values for a job. Next, the system applies a machine learning model to the features to produce scores representing likelihoods of the candidate applying to the jobs. The system then generates a ranking of the jobs by the scores. Finally, the system outputs, to the candidate, at least a portion of the ranking as a set of recommendations.

Abstract: The disclosed embodiments provide a system for recommending jobs based on title transition embeddings. During operation, the system obtains a word embedding model of job histories of members of an online network. Next, the system applies the word embedding model to a first set of attributes associated with a title of a candidate to produce a first embedding. The system also applies the word embedding model to a second set of attributes associated with a job title of a job to produce a second embedding. The system then calculates a similarity between the first and second embeddings. Finally, the system outputs the similarity for use in recommending the job to the candidate.

Abstract: An approach for spending allocation, executed by one or more processors to provide one or more monetary output values in response to a request for determining spending allocation in a digital marketing channel, is provided. The approach fits one or more models to train a business environment simulator. The approach generates a supervised learning policy. The approach evolves a supervised learning policy into a distribution estimator policy by adjusting network weights of the supervised learning policy. The approach generates an optimized policy by evolving the distribution estimator policy through interaction with the business environment simulator. The approach determines a profit uplift of the optimized policy by comparing the optimized policy and the supervised learning policy. Further, in response to the optimized policy outperforming the supervised learning policy, the approach deploys the optimized policy in a live environment.

Abstract: An approach for distributed stream computing in non-idempotent output operations is provided. The approach assigns an eventid to a corresponding entityid. The approach determines a minibatchid and a partitionid for a partition. The approach determines whether the partition was previously processed. The approach generates a new minibatchid and a new partitionid for a new partition based upon determining the partition was not previously processed. The approach determines whether a record was previously processed based upon determining the partition was previously processed. The approach processes the record of the partition based upon determining the record was not previously processed.

Abstract: An approach for spending allocation, executed by one or more processors to provide one or more monetary output values in response to a request for determining spending allocation in a digital marketing channel, is provided. The approach fits one or more models to train a business environment simulator. The approach generates a supervised learning policy. The approach evolves a supervised learning policy into a distribution estimator policy by adjusting network weights of the supervised learning policy. The approach generates an optimized policy by evolving the distribution estimator policy through interaction with the business environment simulator. The approach determines a profit uplift of the optimized policy by comparing the optimized policy and the supervised learning policy. Further, in response to the optimized policy outperforming the supervised learning policy, the approach deploys the optimized policy in a live environment.

Abstract: An approach for distributed stream computing in non-idempotent output operations is provided. The approach assigns an eventid to a corresponding entityid. The approach determines a minibatchid and a partitionid for a partition. The approach determines whether the partition was previously processed. The approach generates a new minibatchid and a new partitionid for a new partition based upon determining the partition was not previously processed. The approach determines whether a record was previously processed based upon determining the partition was previously processed. The approach processes the record of the partition based upon determining the record was not previously processed.

Abstract: An approach for continual learning in slowly-varying environments is provided. The approach receives one or more action requests from a decision agent. The approach deploys a first model to a decision engine. The approach initiates an observation period. The approach builds a second model, in which the second model comprises collected transaction data from the observation period. The approach initiates a test period. The approach determines a performance score for the first model and a performance score for the second model. The approach selects the model providing an optimized action.