Abstract: A cyclic rotary dispensing device and a vending device with the same are provided. The dispensing device includes: a housing, a closed bent slideway disposed on the housing, multiple dispensing assemblies disposed along the closed bent slideway and abutted against each other, a driving assembly and a detection assembly disposed below the closed bent slideway; bottoms of the dispensing assemblies are provided with protrusions; the driving assembly includes a rotary pulling piece disposed below the closed bent slideway and a driving member for driving the rotary pulling piece to rotate; the rotary pulling piece is rotated to drive the protrusions to move, to thereby make the dispensing assemblies cyclically move along the closed bent slideway; the detection assembly is electrically connected to the driving assembly, and is configured to detect positions of the dispensing assemblies. The dispensing device improves the efficiency for selling, and is convenient to maintain.

Abstract: The present technology is directed to training and the use of a machine learning model to measure the safety of an autonomous vehicle (AV) driving. An AV management system can identify driving data including sensor data from an AV that is descriptive of an environment around the AV, a path of the AV, kinematic data of the AV, a path of at least one object in the environment, and in-memory data pertaining to data output by one or more algorithms in an autonomous driving stack. As follows, the AV management system can output a safety score for the path of the AV indicating a probability of a collision between the AV and the at least one object.

Abstract: The present technology is directed to training and using a machine learning model to predict a likelihood of a counterfactual safety critical event in autonomous vehicle (AV) driving in a projected scenario occurring after a human takes over control of an AV. An AV management system can identify driving data collected from periods around an occurrence of a human take over event where a human takes over control of an AV. The AV management system can project a scenario that would have resulted if the human did not take over control of the AV based on the driving data and output a counterfactual safety score for the projected scenario. The counterfactual safety score can indicate a probability of a counterfactual collision between the AV and the at least one object in the projected scenario.

Abstract: Disclosed herein are systems and method including a method for managing an autonomous vehicle. The method includes providing input associated with an autonomous vehicle to a machine learning model, wherein the machine learning model is trained to predict what a planning stack of the autonomous vehicle will choose with respect to selecting a low cost branch of a tree structure in which a plurality of branches of the tree structure are evaluated to determine the low cost branch associated with a future route for the autonomous vehicle. The method further includes generating an output of the machine learning model to predict an output of the planning stack and inputting the output of the machine learning model into the planning stack. The planning stack can traverse a tree structure of possible routes more efficiently with a predicted outcome based on the output of the machine learning model.

Abstract: To select the content to be presented to the user, a first latent vector is determined for a content item based on a first object associated with the content item. A second latent vector is determined for the content item based on a second object associated with the content item. A content item vector is then determined based on the first and second latent vectors. Furthermore, a user vector is determined based on interactions of the user with the first set of content objects and the second set of content objects. A score indicative of the likelihood of the user interacting with the content item is determined based on the content item vector and the user vector.

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 one embodiment, a method includes accessing an input vector representing an input post, wherein: the vector space comprises clusters each associated with a topic; each cluster was determined based on a clustering of training-page vectors corresponding to training pages that each comprise training posts, each training post submitted by a user to a training page and comprises content selected by the user; and each training-page vector was generated by an ANN that was trained, based on the training posts of training pages associated with the ANN, to receive a post and then output a probability that the received post is related to the training posts of the training pages; determining that the input vector is located within a particular cluster in the vector space; and determining a topic of the input post based on the topic associated with the particular cluster that the input vector is located within.

Abstract: The present disclosure is directed to a high-capacity training and prediction machine learning platform that can support high-capacity parameter models (e.g., with 10 billion parameters). The platform generates a model for a metric of interest based on a known training set. The model includes parameters indicating importances of different features of the model, taken both singly and in pairs. The model may be applied to predict a value for the metric for given sets of objects, such as for a pair consisting of a user object and a content item object.

Abstract: An optimized computer architecture for training an neural network includes a system having multiple GPUs. The neural network may be divided into separate portions, and a different portion is assigned to each of the multiple GPUs. Within each GPU, its portion is further divided across multiple training worker threads in multiple processing cores, and each processing core has lock-free access to a local parameter memory. The local parameter memory of each GPU is separately, and individually, synchronized with a remote master parameter memory by lock memory access. Each GPU has a separate set of communication worker threads dedicated to data transfer between the GPU and the remote parameter memory so that the GPU's training worker threads are not involved with cross GPU communications.

Abstract: An online system receives content items from a third party content provider. For each content item, the online system inputs an image into a neural network and extracts a feature vector from a hidden layer of the neural network. The online system compresses each feature vector by assigning a label to each feature value representing whether the feature value was above a threshold value. The online system identifies a set of content items that the user has interacted with and determines a user feature vector by aggregating feature vectors of the set of content items. For a new set of content items, the online system compares the compressed feature vectors of the content item with the user feature vector. The online system selects one or more of the new content items based on the comparison and sends the selected content items to the user.

Abstract: An online system receives content items from a third party content provider. For each content item, the online system inputs an image into a neural network and extracts a feature vector from a hidden layer of the neural network. The online system compresses each feature vector by assigning a label to each feature value representing whether the feature value was above a threshold value. The online system identifies a set of content items that the user has interacted with and determines a user feature vector by aggregating feature vectors of the set of content items. For a new set of content items, the online system compares the compressed feature vectors of the content item with the user feature vector. The online system selects one or more of the new content items based on the comparison and sends the selected content items to the user.

Abstract: In one embodiment, a method includes accessing an input vector representing an input post, wherein the input post includes one or more n-grams and an image, the input vector corresponds to a point in a d-dimensional vector space, the input vector was generated by an artificial neural network (ANN) based on a text vector representing the one or more n-grams of the input post and an image vector representing the image of the input post; and the ANN was jointly trained to receive a text vector representing one or more n-grams of a post and an image vector representing an image of the post and then output a probability that the received post is related to the training posts of a training page; and determining a topic of the input post based on the input vector.

Abstract: In one embodiment, a method includes accessing a user profile associated with a user of an online social network, wherein the user profile identifies one or more topics that the user is interested in; accessing post vectors, wherein each post vector represents one of a plurality of posts, indicates one or more topics, and for each of the topics, indicates a probability that the post is related to the corresponding topic; ranking the posts based on comparisons between the user profile and the post vectors; and providing for display to the user posts based on the ranking.

Abstract: In one embodiment, a method includes accessing an input vector representing an input post, wherein: the vector space comprises clusters each associated with a topic; each cluster was determined based on a clustering of training-page vectors corresponding to training pages that each comprise training posts, each training post submitted by a user to a training page and comprises content selected by the user; and each training-page vector was generated by an ANN that was trained, based on the training posts of training pages associated with the ANN, to receive a post and then output a probability that the received post is related to the training posts of the training pages; determining that the input vector is located within a particular cluster in the vector space; and determining a topic of the input post based on the topic associated with the particular cluster that the input vector is located within.

Abstract: The present disclosure is directed to a high-capacity training and prediction machine learning platform that can support high-capacity parameter models (e.g., with 10 billion weights). The platform implements a generic feature transformation layer for joint updating and a distributed training framework utilizing shard servers to increase training speed for the high-capacity model size. The models generated by the platform can be utilized in conjunction with existing dense baseline models to predict compatibilities between different groupings of objects (e.g., a group of two objects, three objects, etc.).

Abstract: An optimized computer architecture for training an neural network includes a system having multiple GPUs. The neural network may be divided into separate portions, and a different portion is assigned to each of the multiple GPUs. Within each GPU, its portion is further divided across multiple training worker threads in multiple processing cores, and each processing core has lock-free access to a local parameter memory. The local parameter memory of each GPU is separately, and individually, synchronized with a remote master parameter memory by lock memory access. Each GPU has a separate set of communication worker threads dedicated to data transfer between the GPU and the remote parameter memory so that the GPU's training worker threads are not involved with cross GPU communications.

Abstract: The present disclosure is directed to a high-capacity training and prediction machine learning platform that can support high-capacity parameter models (e.g., with 10 billion parameters). The platform generates a model for a metric of interest based on a known training set. The model includes parameters indicating importances of different features of the model, taken both singly and in pairs. The model may be applied to predict a value for the metric for given sets of objects, such as for a pair consisting of a user object and a content item object.

Abstract: To select the content to be presented to the user, a first latent vector is determined for a content item based on a first object associated with the content item. A second latent vector is determined for the content item based on a second object associated with the content item. A content item vector is then determined based on the first and second latent vectors. Furthermore, a user vector is determined based on interactions of the user with the first set of content objects and the second set of content objects. A score indicative of the likelihood of the user interacting with the content item is determined based on the content item vector and the user vector.

Abstract: An online system simplifies modification of features used by machine learned models used by the online system, such as machined learned models with high dimensionality. The online system obtains a superset of features including features used by at least one machine learned model and may include additional features. From the superset of features, the online system generates various groups of features for a machine learned model. The groups of features may be a group including features currently used by the machine learned model, a group including all available features, and one or more intermediate groups. Intermediate groups include various numbers of features from the set selected based on measures of feature impact on the machine learned model associated with various features. A user may select a group of features, test the machine learning model using the selected group, and then launch the tested model based on the results.

Abstract: Provided are techniques for a Data-Partitioned Secondary Index (DPSI) partition level join. While using a Data-Partitioned Secondary Index (DPSI) to perform a join of an outer table and an inner table, a different task from multiple tasks is assigned to each partition of the inner table. With each task, a join is performed of the outer table and the assigned partition of the inner table using the DPSI to generate results. The results from each different task are merged.