The information processing device inputs data into a machine learning model, acquires a first value output from the machine learning model in response to the inputting, a second value output from the machine learning model based on a variable obtained by modifying a latent variable that is calculated by the machine learning model in response to the inputting, and information entropy of the latent variable, and trains the machine learning model based on the first value, the second value and the information entropy of the latent variable.

The information processing device inputs data into a machine learning model, acquires a first value output from the machine learning model in response to the inputting, a second value output from the machine learning model based on a variable obtained by modifying a latent variable that is calculated by the machine learning model in response to the inputting, and information entropy of the latent variable, and trains the machine learning model based on the first value, the second value and the information entropy of the latent variable.

Provided are methods, systems, and computer storage media for determining a command (e.g., intent) of an image based on image data features. A task associated with the determined command is generated based on a portion of the image data features. Task entities corresponding to the task are determined. The task and the corresponding task entities are generated and configured for use in a computer productivity application. Accordingly, present embodiments provide an improved technique for generating command-specific tasks and task entities that may be integratable for use in a computer productivity application to enhance functionality of a computer productivity application and reduce computational resources utilized by manually creating these tasks and task entities.

Various embodiments discussed herein are directed to improving existing technologies by causing certain characters to be replaced at a document if such characters are likely to be an error. For example, documents generated using speech-to-text technology or Optical Character Recognition (OCR) technology often contain character errors. A scoring threshold may be utilized to determine one or more characters are not being correctly represented in the document. Alternatively or additionally, various embodiments recommend multiple character sequences as candidates to replace other characters and a user may select which of the candidates will be used for replacement.

A method of updating a trained cardinality estimation model includes receiving a cardinality estimation model with cardinality labels and detecting a drift in underlying data or predicates of the cardinality estimation model. The type of the detected drift is determined and new test queries that mimic test queries for the detected drift are synthesized. A portion of the synthesized test queries is selected to reduce annotation cost and used to update the cardinality estimation model.

Generally discussed herein are devices, systems, and methods for improving architecture search and identification with constraints. A method can include receiving, at a compute device, a request for a transformer-based autoregressive language model (TBALM), the request specifying a maximum latency, identifying TBALM architectures that satisfies the maximum latency, identifying a TBALM architecture of the identified TBALM architectures that has a greatest number of decoder parameters resulting in an identified TBALM architecture, and providing the identified TBALM architecture.

The proposed model is a Variational Autoencoder having a learnable prior that is parametrized with a Tensor Train (VAE-TTLP). The VAE-TTLP can be used to generate new objects, such as molecules, that have specific properties and that can have specific biological activity (when a molecule). The VAE-TTLP can be trained in a way with the Tensor Train so that the provided data may omit one or more properties of the object, and still result in an object with a desired property.

The proposed model is a Variational Autoencoder having a learnable prior that is parametrized with a Tensor Train (VAE-TTLP). The VAE-TTLP can be used to generate new objects, such as molecules, that have specific properties and that can have specific biological activity (when a molecule). The VAE-TTLP can be trained in a way with the Tensor Train so that the provided data may omit one or more properties of the object, and still result in an object with a desired property.

A system and method for implementing a machine learning-based virtual dialogue agent includes computing an input embedding based on receiving a user input; computing, via a pre-trained machine learning language model, an embedding response inference based on the input embedding; searching, based on the embedding response inference, a response imprintation embedding space that includes a plurality of distinct embedding representations of potential text-based responses to the user input, wherein each of the plurality of distinct embedding representations is tethered to a distinct human-imprinted media response, and searching the response imprintation embedding space includes: searching the response imprintation embedding space based on an embedding search query, and returning a target embedding representation from the response imprintation embedding space based on the searching of the response imprintation embedding space; and executing, via a user interface of the machine learning-based virtual dialogue agent, a human-imprinted media response tethered to the target embedding representation.

To suppress an increase in processing time due to a load of inference processing while improving reading accuracy by the inference processing of machine learning. An optical information reading device includes a processor including: an inference processing part that inputs a code image to a neural network and executes inference processing of generating an ideal image corresponding to the code image; and a decoding processing part that executes first decoding processing of decoding the code image and second decoding processing of decoding the ideal image generated by the inference processing part. The processor executes the inference processing and the first decoding processing in parallel, and executes the second decoding processing after completion of the inference processing.