Intelligently modifying digital calendars utilizing a graph neural network and reinforcement learning

Patent metadata

Field	Value
Publication number	US-12536429-B2
Application number	US-202117337998-A
Country	US
Kind code	B2
Filing date	Jun 3, 2021
Priority date	Apr 26, 2021
Publication date	Jan 27, 2026
Grant date	Jan 27, 2026

Abstract

This disclosure describes methods, non-transitory computer readable storage media, and systems that intelligently generate and modify schedules of task sequences utilizing a graph neural network and/or reinforcement learning model. For example, the disclosed system utilizes a graph neural network to generate performance efficiency scores indicating predicted performances of the sets of tasks. Additionally, the disclosed systems utilizes the performance efficiency scores to rank sets of tasks and then determine a schedule including an ordered sequence of tasks. Furthermore, disclosed system generates modified schedules in response to detecting a modification to the schedule. For example, the disclosed system utilizes a reinforcement learning model to provide recommendations of new tasks or task sequences deviating from the schedule in the event of an interruption. The disclosed system also utilizes the reinforcement learning model to learn from user choices to inform future scheduling of tasks.

First claim

What is claimed is:

1 . A non-transitory computer readable storage medium comprising instructions that, when executed by at least one processor, cause a computing device to:
generate, utilizing a graph neural network, a performance efficiency score for a set of tasks from a plurality of candidate tasks corresponding to a user, wherein the performance efficiency score indicates a difference between an estimated number of completed tasks of the set of tasks and an estimated number of uncompleted tasks of the set of tasks; generate, from edge weights between a set of task nodes and a set of user nodes of the graph neural network, a plurality of task contribution scores for the set of tasks in connection with the performance efficiency score by:
determining, for a task of the set of tasks utilizing an inferencer model with the graph neural network, a norm of edge weights for a task node corresponding to the task in the graph neural network by a backpropagation step for a final layer in the graph neural network via normalizing a combination of edge weights for the task node corresponding to the task in the graph neural network,
wherein the plurality of task contribution scores indicate a contribution of the task of the set of tasks to the performance efficiency score according to a feature of the task and a dependency between the task and the user;
determine, based on the performance efficiency score and the plurality of task contribution scores, a schedule comprising an ordered sequence of the set of tasks for a time period; generate, according to a prior user input within a graphical user interface indicating selection of a prior modified schedule, trajectory data associated with the user; and in response to detecting a user input within the graphical user interface requesting modification to the schedule during the time period and according to the trajectory data, generate, for display via a client device of the user and utilizing the edge weights between the set of task nodes and the set of user nodes, a recommended task in a modified schedule utilizing a reinforcement learning model.

2 . The non-transitory computer readable storage medium as recited in claim 1 , further comprising instructions that, when executed by the at least one processor, cause the computing device to generate the recommended task in a modified schedule by:
generating an updated graph neural network by providing the trajectory data to the graph neural network to the edge weights between the set of task nodes and the set of user nodes; and generating, according to the updated graph neural network, the recommended task in the modified schedule utilizing the reinforcement learning model.

3 . The non-transitory computer readable storage medium as recited in claim 1 , further comprising instructions that, when executed by the at least one processor, cause the computing device to generate the performance efficiency score for the set of tasks by:
determining the edge weights between the set of user nodes and the set of task nodes comprising annotations in a bipartite graph of the graph neural network; and generating the performance efficiency score for the set of tasks based on the edge weights.

4 . The non-transitory computer readable storage medium as recited in claim 3 , further comprising instructions that, when executed by the at least one processor, cause the computing device to:
generate, for the user, the plurality of task contribution scores for the set of tasks in connection with the performance efficiency score by determining norm values of weights associated with the set of task nodes of the graph neural network; and determine the schedule comprising the ordered sequence of the set of tasks further based on the plurality of task contribution scores.

5 . The non-transitory computer readable storage medium as recited in claim 4 , further comprising instructions that, when executed by the at least one processor, cause the computing device to provide a recommendation to add a new task to the set of tasks or remove a task from the set of tasks based on the plurality of task contribution scores.

6 . The non-transitory computer readable storage medium as recited in claim 1 , further comprising instructions that, when executed by the at least one processor, cause the computing device to determine the schedule by:
determining a ranked list of a plurality of sets of tasks based on a plurality of performance efficiency scores for the plurality of sets of tasks; and determining the schedule by selecting the set of tasks from the ranked list of the plurality of sets of tasks.

7 . The non-transitory computer readable storage medium as recited in claim 1 , further comprising instructions that, when executed by the at least one processor, cause the computing device to determine the schedule by:
generating a plurality of schedules comprising a plurality of different ordered sequences of the set of tasks according to one or more constraints corresponding to the user; and determining the schedule from the plurality of schedules in response to a user input selecting the schedule.

8 . The non-transitory computer readable storage medium as recited in claim 1 , further comprising instructions that, when executed by the at least one processor, cause the computing device to provide the recommended task by:
detecting performance of an additional task not in the ordered sequence of the set of tasks during the time period; and determining, utilizing the reinforcement learning model, the recommended task in a modified ordered sequence of tasks in response to detecting the performance of the additional task.

9 . The non-transitory computer readable storage medium as recited in claim 8 , further comprising instructions that, when executed by the at least one processor, cause the computing device to:
determine a plurality of candidate modified ordered sequences of tasks in response to detecting the performance of the additional task; and select the modified ordered sequence of tasks utilizing a Markov decision process in the reinforcement learning model according to one or more constraints associated with the user.

10 . A system comprising:
one or more memory devices comprising a graph neural network having a bipartite graph including a set of user nodes and a set of task nodes, the set of user nodes distinct from the set of task nodes; and one or more computing devices configured to cause the system to:
generate, utilizing edge weights between the set of user nodes and the set of task nodes of the graph neural network, a first performance efficiency score for a first set of tasks from a plurality of candidate tasks corresponding to a user, wherein the first performance efficiency score indicates a difference between an estimated number of completed tasks of the first set of tasks and an estimated number of uncompleted tasks of the first set of tasks;
generate, from the edge weights between the set of task nodes and the set of user nodes of the graph neural network, a first plurality of task contribution scores for the first set of tasks in connection with the first performance efficiency score by:
determining, for a task of the first set of tasks utilizing an inferencer model with the graph neural network, a norm of edge weights for a task node corresponding to the task in the graph neural network by a backpropagation step for a final layer in the graph neural network via normalizing a combination of edge weights for the task node corresponding to the task in the graph neural network,
wherein the first plurality of task contribution scores indicate a contribution of the task of the first set of tasks to the first performance

Assignees

• Adobe Inc

Inventors

• Mitra Saayan
• Wu Gang
• Theocharous Georgios
• Whitehead Richard
• Swaminathan Viswanathan
• Parekh Zahraa
• Tepfer Ben

Classifications

• G06V10/751

  Comparing pixel values or logical combinations thereof, or feature values having positional relevance, e.g. template matching · CPC title

• G06F18/29

  Graphical models, e.g. Bayesian networks · CPC title

• G06F18/2113

  by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation · CPC title

• G06F9/4881

  Scheduling strategies for dispatcher, e.g. round robin, multi-level priority queues · CPC title

• G06N3/08 Primary

  Learning methods · CPC title