Abstract: One or more systems and/or methods for combining vectors output by multiple different mechanisms for content item retrieval are provided. An image encoder may output a first set of vectors generated by an image model using an input image as input. A text encoder may output a second set of vectors generated by a text model using input text as input. A vector combination module may combine the first set of vectors and the second set of vectors to create a vector output. A weight is applied to the vector output to create a weighted output. An output vector is generated based upon a combination of the first set of vectors, the second set of vectors, and the weighted output. The output vector is used to query a catalog to identify a content item related to the input image and the input text.

Abstract: One or more systems and/or methods for product similarity detection and recommendation are provided. Users may view articles and/or other content that includes images depicting products that may be of interest to the users. These images are processed using image processing functionality such as computer vision to identify the products depicted by the images. A vector embedding model is used to generate product vector representations of the products. Catalog items that are available from a catalog to supplement the articles and other content may be processed to generate catalog item vector representations. When content (an article) with an image depicting a product is to be displayed to the user, similarity between a product vector representation of the product and the catalog item vector representations is determined in order to identify and display catalog items depicting products that are similar to the product depicted by the image in the content.

Abstract: One or more computing devices, systems, and/or methods for a model training framework are provided. A definition and configuration of a model are received. Computations that the model will perform during training of the model are setup based upon the definition and configuration. Summary statistics to be tracked during training are specified. A batch of training data is input into the model to train the model based upon hyper parameters specified in the configuration of the model, and the summary statistics are tacked during the training. The parameters of the model are updated based upon a function corresponding to accuracy of the model processing the training data. The summary statistics are outputted.