Abstract: A computer-implemented method is provided for policy evaluation. In the method, the utility of the given decision-making policy is estimated based on a dataset of state-action-reward-state tuples, a set of candidate bootstrapping estimators of the fitted Q-evaluation (FQE) algorithm, and a criterion function. The method automatically selects the best bootstrapping estimator from the candidates based on the criterion function and, when the criterion function is appropriately designed, it produces a good policy-value estimate such that the estimation error is small (below a threshold).

Abstract: A deep neural network training accelerator includes an operational unit sequentially performing first and second operations on a plurality of input data of a sub-set according to a mini-batch gradient descent, a determination unit determining each of the input data as one of skip data and training data based on a confidence matrix obtained by the first operation, and a control unit controlling the operational unit to skip the second operation with respect to the skip data.

Abstract: Dialogue summarization is challenging due to its multi-speaker standpoints, casual spoken language style, and limited labelled data. The embodiments are directed to a coarse-to-fine dialogue summarization model that improves abstractive dialogue summarization quality and enables granular controllability. A summary draft that includes key words for turns in a dialogue conversation history is created. The summary draft includes pseudo-labelled interrogative pronoun categories and noisy key phrases. The dialogue conversation history is divided into segments. A generate language model is trained to generate a segment summary for each dialogue segment using a portion of the summary draft that corresponds to at least one dialogue turn in the dialogue segment. A dialogue summary is generated using the generative language model trained using the summary draft.

Abstract: An integrated circuit with arrays of convolution units that include hardware convolution neural network units configured to perform efficient expansion process by parallelization. The integrated circuit includes multiple building blocks that include the arrays of convolution units, inputs, and outputs t.

Abstract: A system and method enhance original media including a first audio using generative artificial intelligence, including large language models and media conversion modules. The system includes a graphic user interface (GUI) including a media player and a display region for outputting an enhanced media including the original media, a summary of the original media, and the plurality of chapter headings of the original media. The media player plays the original media, and the display region displays the summary and the plurality of chapter headings. The summary, the plurality of chapter headings, and each of a translation into a selected language and a second audio generated from the summary and the plurality of chapter headings are automatically generated from the original media. The method implements the system.

Abstract: A contract platform may receive, from data structures, historical contract data; read, from the historical contract data, numerical categorical data; process the numerical categorical data, with one or more artificial intelligence models, to convert the numerical categorical data into vector format; generate a taxonomy in vector format based on the converted numerical categorical data; read, from the historical contract data, nonnumerical noncategorical data; process the nonnumerical noncategorical data, with the models, to convert the nonnumerical noncategorical data into vector format; generate a knowledge graph in vector format based on the converted nonnumerical noncategorical data.

Abstract: Techniques for using a deep generative model to generate synthetic data sets that can be used to boost the performance of a discriminative model are described. In an example, an autoencoding generative adversarial network (AEGAN) is trained to generate the synthetic data sets. The AEGAN includes an autoencoding network and a generative adversarial network (GAN) that share a generator. The generator learns how to the generate synthetic data sets based on a data distribution from a latent space. Upon training the AEGAN, the generator generates the synthetic data sets. In turn, the synthetic data sets are used to train a predictive model, such as a convolutional neural network for gaze prediction.

Abstract: A decision support method and system is provided for monitoring and treating pediatric obesity. Embodiments include generating obesity risk curves corresponding to obesity risk levels, for example, severe and morbid obesity risk levels. Generating obesity risk curves depends on predicting at least one health proxy such as, for example, spend data and chronic conditions. Generating severe obesity curves depends on an age-dependent multiplier. An obesity risk level is assigned to a target pediatric patient using the obesity risk curves dependent on the age-dependent multiplier. In some aspects, an intervening response is initiated based on the assigned obesity risk level.

Abstract: Systems, computer-implemented methods, and computer program products to facilitate mapping conditional execution logic to different quantum computing resources are provided. According to an embodiment, a system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise a compiler component that maps a logical reference to a quantum bit data structure in an instruction, to a first engine component, and a deployment component that deploys the first engine component to a first block controller component operatively connected to a first quantum computing resource, wherein the first engine component controls the first quantum computing resource based on the instruction.

Abstract: Methods and apparatus for determining which displayed content, e.g., text, of a PDF document is selected and then taking an action based on the text selection, e.g., highlighting, coping, or editing the selected text. The methods and apparatus take into consideration potentially differences between text position and/or size differences between rendered and displayed page content and text position and/or size information obtained from performing a text extraction operation on the content of a PDF document. By combining extracted text information with post-render information corresponding to individual DOM elements in synthesized text elements and then storing the information in an easily access data structure, accurate identification of selected text characters can be made in a relatively fast manner using a browser implemented application.

Abstract: A method of displaying an electronic form on a mobile device comprises creating one or more repeatable items comprising a plurality of fields to be filled or displayed wherein one or more of the fields may include another one or more repeatable item. A method of displaying an electronic form on a mobile device having a screen and managed by an operating system comprises attaching a visual priority to one or more fields used in the electronic form; obtaining the current mobile device screen size; and adapting the number of fields to be displayed on the screen based on the screen size and on visual priority attached to one or more of the fields. In one implementation, a minimum width is set for each of the one or more fields, and the fields that can meet the minimum width based on the current mobile device screen size are displayed.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for validating interactions with false rendered elements. In one aspect, a method includes receiving a rendering notification and a declaration of a rendered element defined in an active window on a client device, detecting interaction with the rendered element at the client device, determining whether the interaction occurred at a declared location of the rendered element within the active window, and processing the interaction including: in response to determining that the interaction occurred: capturing a screenshot of the active window on the client device; verifying a visual appearance of the rendered element in the screenshot with a declared appearance of the rendered element, and generating an interaction attestation, thereby validating the interaction. In response to determining that the interaction did not occur, refraining from generating the interaction attestation.

Abstract: A technology for the optimized capturing of resource file content for resources referred in recorded user interaction sequences is disclosed. Individual resource files are typically referred in multiple recorded resources, therefore it is desired to capture those resources only once and reuse them for all recorded session capturing them. As user interaction sequences are executed and captured in independently operating web-browsers, a direct coordination between recording web-browsers to avoid multiple captures of the same resource is not possible. Data about the global resource capturing and demand situation is generated on a monitoring server that receives all session recording data and transferred to session recording browsers in form of lists identifying resources that are referred in sessions but are still unresolved and should therefore be captured, and for resources that should not captured, because they are already available and capturing them again should be avoided.

Abstract: An apparatus that may include a neural network processor, the neural network processor comprises multiple building blocks. Each of the at least some of the building blocks may include, may consist or may consist essentially of a channel split unit, a convolution unit, a concatenation unit, and a shuffle unit.

Abstract: A system is configured to provide a dashboard creation system. Panels associated with queries for retrieving information from a database are shown in a user interface. Various other user interfaces show query code, panel display settings, function code, dependencies, etc. Queries in a first query format access a data source. Queries in a second query format access a cache database that is synchronized with the data source.

Abstract: Layouts for elements of a flowchart or flow diagram may use recursive algorithms that trace along each parent-child pathway to ensure that child elements always occur after parent elements in a horizontal spacing. Vertical spacing requirements may then be calculated for each element, along with a vertical space above a center line and a vertical space below a center line for each element. These vertical spacing requirements may then be used by a recursive algorithm to assign vertical positions to each of the elements. The vertical spacing may ensure that lanes created for each branch element are not crossed by later elements in the process flow, which ensures that connections between elements may be clearly distinguished from each other.

Abstract: A character input device according to one or more embodiments may include: an output unit configured to output a first character string to an application program having a suggestion function; a detector configured to detect selection of a second character string that is presented in correspondence with the first character string by the application program; and a registration unit configured to register, in a dictionary database, the second character string that is detected, by the detector, to have been selected.

Abstract: An apparatus includes a processor configured to receive, during editing of a first design, user input indicating that a first design element has a first content role. The processor is configured to generate a content signature of the first design indicating that the first design element has the first content role, to generate a second design based on a design template, and to update the second design by applying the content signature to the second design. Generating the second design includes, based on determining that the design template includes a second design element having the first content role, adding a third design element having the first content role to the second design. Applying the content signature to the second design includes transferring content from the first design element to the third design element. The processor is configured to generate a graphical user interface including an image of the second design.

Abstract: Provided is an information processing device including a control unit that performs: temporary label addition processing of adding a temporary label to a predetermined position on input data for machine learning; processing of generating label data in accordance with input of correction of the temporary label by a person in charge of labeling; and processing of comparing label data corrected by the person in charge of labeling with correct-answer label data that is acquired in advance, and calculating labeling accuracy of the person in charge of labeling.

Abstract: Methods and systems for optimized template data management are disclosed. By enabling direct access to the shared universal content data, the template management system can replicate the selected universal content either within one template or across several templates. Furthermore, the universal content can be a section that comprise sub-components or a block that comprises subblocks. The present subject matter can improve computer and network efficiency by eliminating duplicative universal content data.