Patents by Inventor Mohammad SALAMEH
Mohammad SALAMEH has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).
-
Patent number: 11914672Abstract: A method and system for generating neural architectures to perform a particular task. An actor neural network, as part of a continuous action reinforcement learning (RL) agent, generates a randomized continuous actions parameters to encourage exploration of a search space to generate candidate architectures without bias. The continuous action parameters are discretized and applied to a search space to generate candidate architectures, the performance of which for performing the particular task is evaluated. Corresponding reward and state are determined based on the performance. A critic neural network, as part of the continuous action RL agent, learns a mapping of the continuous action to a reward using modified Deep Deterministic Policy Gradient (DDPG) with quantile loss function by sampling a list of top performing architectures. The actor neural network is updated with the learned mapping.Type: GrantFiled: September 29, 2021Date of Patent: February 27, 2024Assignee: HUAWEI TECHNOLOGIES CO., LTD.Inventors: Mohammad Salameh, Keith George Mills, Di Niu
-
Publication number: 20230096654Abstract: A method and system for generating neural architectures to perform a particular task. An actor neural network, as part of a continuous action reinforcement learning (RL) agent, generates a randomized continuous actions parameters to encourage exploration of a search space to generate candidate architectures without bias. The continuous action parameters are discretized and applied to a search space to generate candidate architectures, the performance of which for performing the particular task is evaluated. Corresponding reward and state are determined based on the performance. A critic neural network, as part of the continuous action RL agent, learns a mapping of the continuous action to a reward using modified Deep Deterministic Policy Gradient (DDPG) with quantile loss function by sampling a list of top performing architectures. The actor neural network is updated with the learned mapping.Type: ApplicationFiled: September 29, 2021Publication date: March 30, 2023Inventors: Mohammad SALAMEH, Keith George MILLS, Di NIU
-
Patent number: 11113608Abstract: Implementations are directed to receiving communication data from a device, the communication data including data input by a user of the device, determining a context based on an extended finite state machine that defines contexts and transitions between contexts, transmitting a service request to at least one cloud-hosted service, the service request being provided at least partially based on masking sensitive information included in the communication data, receiving a service response from the at least one cloud-hosted service, the service response including one or more of an intent, and an entity, determining at least one action that is to be performed by at least one back-end source system based on the service response, providing a response at least partially based on an action results received from the at least one back-end source system, and transmitting the result data to the device.Type: GrantFiled: October 30, 2017Date of Patent: September 7, 2021Assignee: Accenture Global Solutions LimitedInventors: Tariq Mohammad Salameh, Michele Tornielli
-
Publication number: 20200344998Abstract: The present disclosure concerns a modified antifreeze protein having the formula AFP-PEG, where AFP is an antifreeze protein, PEG is a poly(alkylene glycol) unit, and the PEG is linked to an amino acid residue in the AFP that is not involved in direct ice-surface binding and that has a functional group selected from an amine, a thiol, a hydroxy, a carboxylate, an amide and a guanidine in its side chain. A formulation including the same, a method of protecting a biological tissue, organ or body using the same, and a method of synthesizing a modified antifreeze protein are also disclosed.Type: ApplicationFiled: July 20, 2020Publication date: November 5, 2020Inventors: Yong BA, Mohammad SALAMEH, Adiel PEREZ
-
Patent number: 10719701Abstract: Implementations for optical character recognition of a language script can include actions of receiving an image comprising a graphical representation of a word written in the language script, segmenting the word into two or more segments, each segment being determined based on one or more of a variation in a height of the word and a variation in a width of the word, and including at least one character, providing a boundary for each segment of the two or more segments, the boundary enclosing the at least one character of a respective segment, each boundary having an edge with respect to an axis of the image, normalizing boundaries of the two or more segments by aligning edges of the boundaries, and labeling each segment of the two or more segments with a respective label, the respective label indicating a language character within the respective boundary.Type: GrantFiled: April 25, 2018Date of Patent: July 21, 2020Assignee: Accenture Global Solutions LimitedInventor: Tariq Mohammad Salameh
-
Patent number: 10616036Abstract: Implementations are directed to integrating a computer-executed service platform with multiple channels, and include actions of providing a transition layer as an application executed on a client-side device, which receives input from a channel of the multiple channels, and providing a unified input that is in a unified format for processing by the service platform, receiving, by the service platform, the unified input, identifying, by the service platform, a set of actions based on the unified input, the set of actions including to be executed by one or more back-end services, providing a service layer including a plurality of bots, each bot interacting with a back-end service based on an action to receive at least one result from the respective back-end service, receiving result data including data describing the at least one result, and transmitting an output based on the result data through the transition layer of the client-side device.Type: GrantFiled: June 7, 2017Date of Patent: April 7, 2020Assignee: Accenture Global Solutions LimitedInventors: Tariq Mohammad Salameh, Marco Intermite, Matteo Luca Maga, Federica Rossi
-
Publication number: 20190332860Abstract: Implementations for optical character recognition of a language script can include actions of receiving an image comprising a graphical representation of a word written in the language script, segmenting the word into two or more segments, each segment being determined based on one or more of a variation in a height of the word and a variation in a width of the word, and including at least one character, providing a boundary for each segment of the two or more segments, the boundary enclosing the at least one character of a respective segment, each boundary having an edge with respect to an axis of the image, normalizing boundaries of the two or more segments by aligning edges of the boundaries, and labeling each segment of the two or more segments with a respective label, the respective label indicating a language character within the respective boundary.Type: ApplicationFiled: April 25, 2018Publication date: October 31, 2019Inventor: Tariq Mohammad Salameh
-
Publication number: 20190130286Abstract: Implementations are directed to receiving communication data from a device, the communication data including data input by a user of the device, determining a context based on an extended finite state machine that defines contexts and transitions between contexts, transmitting a service request to at least one cloud-hosted service, the service request being provided at least partially based on masking sensitive information included in the communication data, receiving a service response from the at least one cloud-hosted service, the service response including one or more of an intent, and an entity, determining at least one action that is to be performed by at least one back-end source system based on the service response, providing a response at least partially based on an action results received from the at least one back-end source system, and transmitting the result data to the device.Type: ApplicationFiled: October 30, 2017Publication date: May 2, 2019Inventors: Tariq Mohammad Salameh, Michele Tornielli
-
Publication number: 20180359132Abstract: Implementations are directed to integrating a computer-executed service platform with multiple channels, and include actions of providing a transition layer as an application executed on a client-side device, which receives input from a channel of the multiple channels, and providing a unified input that is in a unified format for processing by the service platform, receiving, by the service platform, the unified input, identifying, by the service platform, a set of actions based on the unified input, the set of actions including to be executed by one or more back-end services, providing a service layer including a plurality of bots, each bot interacting with a back-end service based on an action to receive at least one result from the respective back-end service, receiving result data including data describing the at least one result, and transmitting an output based on the result data through the transition layer of the client-side device.Type: ApplicationFiled: June 7, 2017Publication date: December 13, 2018Inventors: Tariq Mohammad Salameh, Marco Intermite, Matteo Luca Maga, Federica Rossi
-
Publication number: 20180253638Abstract: Implementations are directed to receiving communication data from a device, the communication data including data input by a user of the device, receiving text data based on the communication data, providing an intent set and an entity set based on processing the text data through an artificial intelligence service, the intent set including one or more intents indicated in the text data, the entity set including one or more entities indicated in the text data, the artificial intelligence service implementing a convolution neural networks (CNN), identifying a set of actions based on one or more of the text data, the intent set, and the entity set, receiving a set of results including at least one result from executing an action of the set of actions, providing result data, and transmitting the result data to the device.Type: ApplicationFiled: March 2, 2017Publication date: September 6, 2018Inventors: Matteo Luca Maga, Tariq Mohammad Salameh, Federica Rossi