Abstract: Generally, the present disclosure is directed to systems and methods that provide a simple, scalable, yet effective strategy to perform transfer learning with a mixture of experts (MoE). In particular, the transfer of pre-trained representations can improve sample efficiency and reduce computational requirements for new tasks. However, representations used for transfer are usually generic, and are not tailored to a particular distribution of downstream tasks. In contrast, example systems and methods of the present disclosure use expert representations for transfer with a simple, yet effective, strategy.

Abstract: Aspects of the subject technology relate to dynamically adjusting a UI based on the current modality. Layout features of a UI may be determined based on an input modality of a computing device. The arrangement of the UI elements may be determined based on the layout features and respective importance scores of the UI elements. The UI elements arranged based on the arrangement may be provided for display of the computing device.

Abstract: Aspects of the subject technology relate to dynamically adjusting a UI based on the current modality. Layout features of a UI may be determined based on an input modality of a computing device. The arrangement of the UI elements may be determined based on the layout features and respective importance scores of the UI elements. The UI elements arranged based on the arrangement may be provided for display of the computing device.

Abstract: Provided are systems and methods for predicting machine learning model performance from the model parameter values, including for use in making improved decisions with regard to early stopping of training procedures. As one example, the present disclosure discusses the prediction of the accuracy (e.g., relative to a defined task and testing dataset such as a computer vision task) of trained neural networks (e.g., convolutional neural networks (CNNs)), using only the parameter values (e.g., the values of the network's weights) as inputs. As such, one example aspect of the present disclosure is directed to computing systems that include and use a machine-learned performance prediction model that has been trained to predict performance values of machine-learned models based on their parameter values (e.g., weight values and/or hyperparameter values).

Abstract: Aspects of the subject technology relate to dynamically adjusting a UI based on the current modality. Layout features of a UI may be determined based on an input modality of a computing device. The arrangement of the UI elements may be determined based on the layout features and respective importance scores of the UI elements. The UI elements arranged based on the arrangement may be provided for display of the computing device.

Abstract: The present disclosure provides systems and methods that leverage machine-learned models (e.g., neural networks) to provide enhanced communication assistance. In particular, the systems and methods of the present disclosure can include or otherwise leverage a machine-learned communication assistance model to detect problematic statements included in a communication and/or provide suggested replacement statements to respectively replace the problematic statements. In one particular example, the communication assistance model can include a long short-term memory recurrent neural network that detects an inappropriate tone or unintended meaning within a user-composed communication and provides one or more suggested replacement statements to replace the problematic statements.

Abstract: A computing device is described that receives first input, at an initial time, of a first textual character and a second input, at a subsequent time, of a second textual character. The computing device determines, based on the first and second textual characters, a first character sequence that does not include a space character between the first and second textual characters and a second character sequence that includes the space character between the first and second textual characters. The computing device determines a first score associated with the first character sequence and a second score associated with the second character sequence. The computing device adjusts, based on a duration of time between the initial and subsequent times, the second score to determine a third score, and responsive to determining that the third score exceeds the first score, the computing device outputs the second character sequence.

Abstract: A computer-implemented method includes: receiving, at a user device, user input corresponding to handwritten text to be recognized using a recognition engine; and receiving, at the user device, a representation of the handwritten text. The representation includes the handwritten text parsed into individual handwritten characters. The method further includes: displaying, on a display of the user device, the handwritten characters using a first indicator; receiving, at the user device, an identification of a text character recognized as one of the handwritten characters; displaying, on the display, the text character; and adjusting, at the user device, the one of the handwritten characters from being displayed using the first indicator to using a second indicator in response to the received identification. The first and second indicators are different.

Abstract: The present disclosure provides systems and methods that leverage machine-learned models (e.g., neural networks) to provide enhanced communication assistance. In particular, the systems and methods of the present disclosure can include or otherwise leverage a machine-learned communication assistance model to detect problematic statements included in a communication and/or provide suggested replacement statements to respectively replace the problematic statements. In one particular example, the communication assistance model can include a long short-term memory recurrent neural network that detects an inappropriate tone or unintended meaning within a user-composed communication and provides one or more suggested replacement statements to replace the problematic statements.

Abstract: Aspects of the subject technology relate to dynamically adjusting a UI based on the current modality. Layout features of a UI may be determined based on an input modality of a computing device. The arrangement of the UI elements may be determined based on the layout features and respective importance scores of the UI elements. The UI elements arranged based on the arrangement may be provided for display of the computing device.

Abstract: The present disclosure provides systems and methods that leverage machine-learned models (e.g., neural networks) to provide enhanced communication assistance. In particular, the systems and methods of the present disclosure can include or otherwise leverage a machine-learned communication assistance model to detect problematic statements included in a communication and/or provide suggested replacement statements to respectively replace the problematic statements. In one particular example, the communication assistance model can include a long short-term memory recurrent neural network that detects an inappropriate tone or unintended meaning within a user-composed communication and provides one or more suggested replacement statements to replace the problematic statements.

Abstract: A computing device is described that receives first input, at an initial time, of a first textual character and a second input, at a subsequent time, of a second textual character. The computing device determines, based on the first and second textual characters, a first character sequence that does not include a space character between the first and second textual characters and a second character sequence that includes the space character between the first and second textual characters. The computing device determines a first score associated with the first character sequence and a second score associated with the second character sequence. The computing device adjusts, based on a duration of time between the initial and subsequent times, the second score to determine a third score, and responsive to determining that the third score exceeds the first score, the computing device outputs the second character sequence.

Abstract: A computing device is described that receives first input, at an initial time, of a first textual character and a second input, at a subsequent time, of a second textual character. The computing device determines, based on the first and second textual characters, a first character sequence that does not include a space character between the first and second textual characters and a second character sequence that includes the space character between the first and second textual characters. The computing device determines a first score associated with the first character sequence and a second score associated with the second character sequence. The computing device adjusts, based on a duration of time between the initial and subsequent times, the second score to determine a third score, and responsive to determining that the third score exceeds the first score, the computing device outputs the second character sequence.

Abstract: A computer-implemented method includes: receiving, at a user device, user input corresponding to handwritten text to be recognized using a recognition engine; and receiving, at the user device, a representation of the handwritten text. The representation includes the handwritten text parsed into individual handwritten characters. The method further includes: displaying, on a display of the user device, the handwritten characters using a first indicator; receiving, at the user device, an identification of a text character recognized as one of the handwritten characters; displaying, on the display, the text character; and adjusting, at the user device, the one of the handwritten characters from being displayed using the first indicator to using a second indicator in response to the received identification. The first and second indicators are different.

Abstract: A method includes receiving from a user a message for sharing with others, identifying message attributes of the received message, the message attributes comprising message content, identifying a group of recipients with whom the user has previously interacted, computing a relevance score for each recipient in the group of recipients, ranking each recipient in the group of recipients according to the relevance score and determining one or more target recipients for the received message from the group of target recipients ranked according to the relevance score, the determined target recipients having a ranking within a predetermined threshold of highest relevance scores.

Abstract: A computer-implemented method includes: receiving, at a user device, user input corresponding to handwritten text to be recognized using a recognition engine; and receiving, at the user device, a representation of the handwritten text. The representation includes the handwritten text parsed into individual handwritten characters. The method further includes: displaying, on a display of the user device, the handwritten characters using a first indicator; receiving, at the user device, an identification of a text character recognized as one of the handwritten characters; displaying, on the display, the text character; and adjusting, at the user device, the one of the handwritten characters from being displayed using the first indicator to using a second indicator in response to the received identification. The first and second indicators are different.

Abstract: A social network server system may receive a social media message that is to be posted at the social network server system, the social media message being authored by a user of the social network server system. Prior to posting the social media message at the social network server system, the social network server system may determine, based at least in part on applying one or more rules to content of the social media message, a likelihood that the user would modify the content of the social media message after it is posted at the social network server system, wherein the one or more rules are generated based at least in part on previous actions taken by the user on previous social media messages authored by the user and posted at the social network server system and may, responsive to determining that the likelihood exceeds a threshold, generate an alert message.

Abstract: The present disclosure provides systems and methods that leverage machine-learned models (e.g., neural networks) to provide enhanced communication assistance. In particular, the systems and methods of the present disclosure can include or otherwise leverage a machine-learned communication assistance model to detect problematic statements included in a communication and/or provide suggested replacement statements to respectively replace the problematic statements. In one particular example, the communication assistance model can include a long short-term memory recurrent neural network that detects an inappropriate tone or unintended meaning within a user-composed communication and provides one or more suggested replacement statements to replace the problematic statements.

Abstract: Forwarding wireless signals comprises a user and a counterpart opening secure applications on a user computing device and a counterpart computing device, respectively. The user places the user computing device within range of a wireless signal, such as a wireless signal provided by a point of sale (“POS”) terminal. The user computing device forwards the wireless signal from the POS terminal to the counterpart computing device. The user computing device forwards the wireless signal from the counterpart computing device to the POS terminal. Thus, the counterpart computing device may conduct a transaction with the POS terminal as if the counterpart computing device were at the location of the POS terminal. The counterpart computing device may also receive a forwarded beacon signal comprising data, such as an offer, provided by the POS terminal or another suitable beacon transmission device at the merchant location.

Abstract: Forwarding wireless signals comprises a user and a counterpart opening secure applications on a user computing device and a counterpart computing device, respectively. The user places the user computing device within range of a wireless signal, such as a wireless signal provided by a point of sale (“POS”) terminal. The user computing device forwards the wireless signal from the POS terminal to the counterpart computing device. The user computing device forwards the wireless signal from the counterpart computing device to the POS terminal. Thus, the counterpart computing device may conduct a transaction with the POS terminal as if the counterpart computing device were at the location of the POS terminal. The counterpart computing device may also receive a forwarded beacon signal comprising data, such as an offer, provided by the POS terminal or another suitable beacon transmission device at the merchant location.