Abstract: In one embodiment, a method for training a machine-learning model having multiple parameters includes instantiating trainers each associated with at least a worker thread, a synchronization thread, and a local version of the parameters, using the worker threads to perform training operations that comprise generating an updated local version of the parameters for each trainer using its associated worker thread, while the worker threads are performing training operations, using the synchronization threads to perform synchronization operations that comprise generating a global version of the parameters based on the updated local versions of the parameters and generating a synchronized local version of the parameters for each trainer based on the global version, continuing performing training operations based on the synchronized local versions of the parameters, and determining the parameters at the end of training based on at least a final local version of the parameters associated with one trainer.

Abstract: In particular embodiments, a method includes receiving, from a client system of a first user, a structured query comprising references to one or more objects associated with the online social network, identifying a set of target objects corresponding to the structured query, wherein each target object in the set is an object that is connected within the online social network to at least one of the objects referenced in the structured query, filtering the set of target objects based on a privacy setting of each of the connections to the target object, and generating search results comprising references to each target object from the post-filtered set that is connected to the first user within the online social network by a series of connections that have a visibility that is visible to the first user.

Abstract: An online system identifies an additional feature to evaluate for inclusion in a machine learned model. The additional feature is based on characteristics of one or more dimensions of information maintained by the online system. To generate data for evaluating the additional feature, the online system generates various partitions of stored data, where each partition includes characteristics associated with one or more dimensions on which the additional feature is based. Using values of characteristics in a partition, the online system generates values for the additional feature and includes the values of the additional feature in the partition. Values for the additional feature are generated for various partitions based on the values of characteristics in each partition. The online system combines multiple partitions that include values for the additional feature to generate a training set for evaluating a machine learned model including the additional feature.

Abstract: An online system identifies an additional feature to evaluate for inclusion in a machine learned model. The additional feature is based on characteristics of one or more dimensions of information maintained by the online system. To generate data for evaluating the additional feature, the online system generates various partitions of stored data, where each partition includes characteristics associated with one or more dimensions on which the additional feature is based. Using values of characteristics in a partition, the online system generates values for the additional feature and includes the values of the additional feature in the partition. Values for the additional feature are generated for various partitions based on the values of characteristics in each partition. The online system combines multiple partitions that include values for the additional feature to generate a training set for evaluating a machine learned model including the additional feature.

Abstract: Some embodiments include a workflow authoring tool that accesses a text string representation of a workflow and a text string representation of at least a data processing operator type. The workflow authoring tool enables definition of one or more data processing operator types that can be referenced in defining the machine learning workflow. When scheduling a workflow, the text string representation of the workflow can be parsed and traversed to generate an interdependency graph of one or more data processing operators. The text string representation of the data processing operator type can identify operator attributes associated with the data processing operator type.

Abstract: In particular embodiments, a method includes receiving, from a client device associated with a first user of an online social network, a text query comprising one or more character strings, identifying one or more data objects that match at least a portion of one or more of the character strings, wherein each of the identified data objects is connected to the first user within the online social network, generating one or more recommended queries that each comprise the character strings of the text query and references to one or more of the identified data objects, and sending, to the client device associated with the first user in response to receiving the text query, one or more of the recommended queries for display to the first user.

Abstract: Some embodiments include an experiment management interface for a machine learning system. The experiment management interface can manage one or more workflow runs related to building or testing machine learning models. The experiment management interface can receive an experiment initialization command to create a new experiment associated with a new workflow. A workflow can be represented by an interdependency graph of one or more data processing operators. The experiment management interface enables definition of the new workflow from scratch or by cloning and modifying an existing workflow. The workflow can define a summary format for its inputs and outputs. In some embodiments, the experiment management interface can automatically generate a comparative visualization at the conclusion of running the new workflow based on an input schema or an output schema of the new workflow.

Abstract: Some embodiments include a method of machine learner workflow processing. For example, a workflow execution engine can receive an interdependency graph of operator instances for a workflow run. The operator instances can be associated with one or more operator types. The workflow execution engine can assign one or more computing environments from a candidate pool to execute the operator instances based on the interdependency graph. The workflow execution engine can generate a schedule plan of one or more execution requests associated with the operator instances. The workflow execution engine can distribute code packages associated the operator instances to the assigned computing environments. The workflow execution engine can maintain a memoization repository to cache one or more outputs of the operator instances upon completion of the execution requests.

Abstract: In particular embodiments, a method includes receiving, from a client system of a first user, a structured query comprising references to one or more objects associated with the online social network, identifying a set of target objects corresponding to the structured query, wherein each target object in the set is an object that is connected within the online social network to at least one of the objects referenced in the structured query, filtering the set of target objects based on a privacy setting of each of the connections to the target object, and generating search results comprising references to each target object from the post-filtered set that is connected to the first user within the online social network by a series of connections that have a visibility that is visible to the first user.

Abstract: In particular embodiments, a method includes receiving, from a client device associated with a first user of an online social network, an unstructured text query, parsing the text query to identify one or more n-grams, determining a score for each n-gram, wherein the score for each n-gram is a probability that the n-gram corresponds to a particular data object of a plurality of data objects associated with the online social network, identifying one or more data objects based on the scores of their corresponding n-grams, each identified data object corresponding to at least one of the n-grams, and generating one or more structured queries that each comprise references to one or more of the identified data objects.

Abstract: In particular embodiments, a method includes receiving an unstructured text query, identifying nodes and edges from a social graph that correspond to n-grams in the text query, generating structured queries that include references to the identified nodes and edges, receiving a selection of a structured query, identifying target nodes that correspond to the structured query, and then generating search results that include target nodes with privacy settings where the nodes and edges along the path connecting the target node and the querying user are all visible to the user.

Abstract: In one embodiment, a method includes receiving a text query inputted by a first user, identifying objects associated with the online social network, each identified object having a name string matching at least in part a portion of the text query, determining a probability score for each identified object that the identified object corresponds to a respective portion of the text query, selecting one or more of the identified objects based on the probability score of the identified object, identifying one or more grammars of a grammar model, each identified grammar having query tokens corresponding to at least one of the selected objects, and sending instructions for presenting suggested queries to the first user, each suggested query comprising the text string, a query token corresponding to one of the selected objects, and one or more additional query tokens of the identified grammar corresponding to the respective selected object.

Abstract: In one embodiment, a method includes accessing a social graph that includes a plurality of nodes and edges, receiving an unstructured text query, identifying nodes and edges that correspond to n-grams of the text query, accessing a context-free grammar model, identifying grammars having query tokens that correspond to the identified nodes and edges, determining a score for each identified grammar, and then generating structured queries based on the identified grammars based on strings generated by the grammars.

Abstract: In particular embodiments, a method includes receiving, from a client device associated with a first user of an online social network, an unstructured text query, parsing the text query to identify one or more n-grams, determining a score for each n-gram, wherein the score for each n-gram is a probability that the n-gram corresponds to a particular data object of a plurality of data objects associated with the online social network, identifying one or more data objects based on the scores of their corresponding n-grams, each identified data object corresponding to at least one of the n-grams, and generating one or more structured queries that each comprise references to one or more of the identified data objects.

Abstract: Some embodiments include a workflow authoring tool that accesses a text string representation of a workflow and a text string representation of at least a data processing operator type. The workflow authoring tool enables definition of one or more data processing operator types that can be referenced in defining the machine learning workflow. When scheduling a workflow, the text string representation of the workflow can be parsed and traversed to generate an interdependency graph of one or more data processing operators. The text string representation of the data processing operator type can identify operator attributes associated with the data processing operator type.

Abstract: Some embodiments include a method of machine learner workflow processing. For example, a workflow execution engine can receive an interdependency graph of operator instances for a workflow run. The operator instances can be associated with one or more operator types. The workflow execution engine can assign one or more computing environments from a candidate pool to execute the operator instances based on the interdependency graph. The workflow execution engine can generate a schedule plan of one or more execution requests associated with the operator instances. The workflow execution engine can distribute code packages associated the operator instances to the assigned computing environments. The workflow execution engine can maintain a memoization repository to cache one or more outputs of the operator instances upon completion of the execution requests.

Abstract: Some embodiments include an experiment management interface for a machine learning system. The experiment management interface can manage one or more workflow runs related to building or testing machine learning models. The experiment management interface can receive an experiment initialization command to create a new experiment associated with a new workflow. A workflow can be represented by an interdependency graph of one or more data processing operators. The experiment management interface enables definition of the new workflow from scratch or by cloning and modifying an existing workflow. The workflow can define a summary format for its inputs and outputs. In some embodiments, the experiment management interface can automatically generate a comparative visualization at the conclusion of running the new workflow based on an input schema or an output schema of the new workflow.

Abstract: In particular embodiments, a method includes receiving an unstructured text query, parsing the text query to identify n-grams; determining a score that the n-grams correspond to particular nodes and edges from a social graph, identifying those nodes and edges with a score greater than a threshold score, and then generating structured queries that include references to the identified nodes and edges.

Abstract: In particular embodiments, a method includes receiving, from a client device associated with a first user of an online social network, a text query comprising one or more character strings, identifying one or more data objects that match at least a portion of one or more of the character strings, wherein each of the identified data objects is connected to the first user within the online social network, generating one or more recommended queries that each comprise the character strings of the text query and references to one or more of the identified data objects, and sending, to the client device associated with the first user in response to receiving the text query, one or more of the recommended queries for display to the first user.

Abstract: An online system identifies an additional feature to evaluate for inclusion in a machine learned model. The additional feature is based on characteristics of one or more dimensions of information maintained by the online system. To generate data for evaluating the additional feature, the online system generates various partitions of stored data, where each partition includes characteristics associated with one or more dimensions on which the additional feature is based. Using values of characteristics in a partition, the online system generates values for the additional feature and includes the values of the additional feature in the partition. Values for the additional feature are generated for various partitions based on the values of characteristics in each partition. The online system combines multiple partitions that include values for the additional feature to generate a training set for evaluating a machine learned model including the additional feature.