Abstract: Disclosed herein are a parameter server and a method for sharing distributed deep-learning parameters using the parameter server. The method for sharing distributed deep-learning parameters using the parameter server includes initializing a global weight parameter in response to an initialization request by a master process; performing an update by receiving a learned local gradient parameter from the worker process, which performs deep-learning training after updating a local weight parameter using the global weight parameter; accumulating the gradient parameters in response to a request by the master process; and performing an update by receiving the global weight parameter from the master process that calculates the global weight parameter using the accumulated gradient parameters of the one or more worker processes.

Abstract: Disclosed herein are a parameter server and a method for sharing distributed deep-learning parameters using the parameter server. The method for sharing distributed deep-learning parameters using the parameter server includes initializing a global weight parameter in response to an initialization request by a master process; performing an update by receiving a learned local gradient parameter from the worker process, which performs deep-learning training after updating a local weight parameter using the global weight parameter; accumulating the gradient parameters in response to a request by the master process; and performing an update by receiving the global weight parameter from the master process that calculates the global weight parameter using the accumulated gradient parameters of the one or more worker processes.

Abstract: Disclosed herein are a parameter server and a method for sharing distributed deep-learning parameters using the parameter server. The method for sharing distributed deep-learning parameters using the parameter server includes initializing a global weight parameter in response to an initialization request by a master process; performing an update by receiving a learned local gradient parameter from the worker process, which performs deep-learning training after updating a local weight parameter using the global weight parameter; accumulating the gradient parameters in response to a request by the master process; and performing an update by receiving the global weight parameter from the master process that calculates the global weight parameter using the accumulated gradient parameters of the one or more worker processes.

Abstract: Disclosed herein are an apparatus and method for interfacing with common memory. The apparatus for interfacing with common memory includes a computer-input/output (I/O)-interface-protocol-processing unit for receiving a packet for accessing common memory from a computer; a direct memory access unit for transforming the packet into an instruction for performing any one of reading from and writing to the common memory; and a common memory interface unit for transmitting the instruction to the common memory and receiving information about whether execution of the instruction is completed from the common memory.

Abstract: A parameter-sharing apparatus includes a memory allocation unit for managing allocation of a memory area, in which a parameter is to be stored, to a memory box, and updating a mapping table stored in the memory box based on allocation management of the memory area, and an operation processing unit for providing the memory allocation unit with parameter information required for the allocation management of the memory area in which the parameter is to be stored and sharing the parameter stored in the memory box.

Abstract: Disclosed herein are an apparatus and method for interfacing with common memory. The apparatus for interfacing with common memory includes a computer-input/output (I/O)-interface-protocol-processing unit for receiving a packet for accessing common memory from a computer; a direct memory access unit for transforming the packet into an instruction for performing any one of reading from and writing to the common memory; and a common memory interface unit for transmitting the instruction to the common memory and receiving information about whether execution of the instruction is completed from the common memory.

Abstract: Disclosed herein are a parameter server and a method for sharing distributed deep-learning parameters using the parameter server. The method for sharing distributed deep-learning parameters using the parameter server includes initializing a global weight parameter in response to an initialization request by a master process; performing an update by receiving a learned local gradient parameter from the worker process, which performs deep-learning training after updating a local weight parameter using the global weight parameter; accumulating the gradient parameters in response to a request by the master process; and performing an update by receiving the global weight parameter from the master process that calculates the global weight parameter using the accumulated gradient parameters of the one or more worker processes.

Abstract: Disclosed herein are an apparatus and method for sharing memory between computers. The apparatus for sharing memory between computers includes multiple memory adapters, installed in corresponding ones of multiple computers, for receiving an address corresponding to an instruction from the computers and transforming the received address into an instruction in the form of a packet; and shared memory for transforming the instruction in the form of the packet, received from the multiple memory adapters, into an address and performing an operation corresponding to the instruction for a memory cell corresponding to the address.

Abstract: Disclosed herein are a parameter-sharing apparatus and method. The parameter-sharing apparatus includes a memory allocation unit for managing allocation of a memory area, in which a parameter is to be stored, to a memory box, and updating a mapping table stored in the memory box based on allocation management of the memory area, and an operation processing unit for providing the memory allocation unit with parameter information required for the allocation management of the memory area in which the parameter is to be stored and sharing the parameter stored in the memory box.

Abstract: Provided herein is a method for dynamic operating of a multi-attribute memory cache based on a distributed memory integration framework, the method including setting a predetermined memory area to be divided into a first area and a second area; in response to obtaining information from the cache client on a request for a service to transceive data regarding the predetermined file, determining a type of a function that the cache client requested; and in response to determining that the cache client requested a cache function, specifying the first area as an area to be used by the cache client for data transceiving regarding the predetermined file, and in response to determining that the cache client requested a storing function, specifying the second area as an area to be used by the cache client for data transceiving regarding the predetermined file.

Abstract: The present invention relates to technology for providing a remote memory, and more particularly, to a system for providing a remote memory which may enable an application in a high performance computing system to use a physical memory of a remote computing node like a local memory of a computing node in which the corresponding application is executed, and a temporary page pool operating method for providing a remote memory.

Abstract: Disclosed herein are a resource manager node and a resource management method. The resource manager node includes a resource management unit, a resource policy management unit, a shared resource capability management unit, a shared resource status monitoring unit, and a shared resource allocation unit. The resource management unit performs an operation necessary for resource allocation when a resource allocation request is received. The resource policy management unit determines a resource allocation policy based on the characteristic of the task, and generates resource allocation information. The shared resource capability management unit manages the topology of nodes, information about the capabilities of resources, and resource association information. The shared resource status monitoring unit monitors and manages information about the status of each node and the use of allocated resources. The shared resource allocation unit sends a resource allocation request to at least one of the plurality of nodes.

Abstract: Disclosed herein are a resource manager node and a resource management method. The resource manager node includes a resource management unit, a resource policy management unit, a shared resource capability management unit, a shared resource status monitoring unit, and a shared resource allocation unit. The resource management unit performs an operation necessary for resource allocation when a resource allocation request is received. The resource policy management unit determines a resource allocation policy based on the characteristic of the task, and generates resource allocation information. The shared resource capability management unit manages the topology of nodes, information about the capabilities of resources, and resource association information. The shared resource status monitoring unit monitors and manages information about the status of each node and the use of allocated resources. The shared resource allocation unit sends a resource allocation request to at least one of the plurality of nodes.

Abstract: Provided is a hyperlink display method based on a visit history accumulation, which is capable of providing search convenience between web documents by providing a self-updating hyperlink through accumulation of a visit history. The hyperlink display method includes displaying a plurality of hyperlink included in a web document when the web document is selected, recording a weight value of a selected hyperlink when any one of the plurality of hyperlinks is selected, and displaying the weight value of the selected hyperlink on the web document.

Abstract: Provided is a method of pairing terminals with each other, and a terminal for the method. The method includes sensing a physical motion of a terminal caused by a user and outputting a sensing value, comparing a reception value received from an external terminal with the sensing value, and establishing a communication path with the external terminal according to the comparison result.

Abstract: A user manual supporting method for use in an electronic appliance includes converting a series of operations performed by the user to operate the electronic appliance and converting the operations into a pattern of user operation sequence, and checking if an error is present in the pattern of user operation sequence to retrieve a pattern of erroneous operation sequence corresponding to the pattern of user operation sequence having the error. Thereafter, a manual content associated with the pattern of erroneous operation pattern is extracted and the extracted manual content is provided to the user. The manual content the manual content includes text and/or graphics information for notifying the user of a missing operation in the pattern of user operation sequence, or for guiding a normal operation against the pattern of erroneous operation sequence.

Abstract: A system for a user authentication includes an odor sensor unit for sensing an odor of a user's body to generate an odor biometric information vector, and a learning unit for performing an initial learning using the odor biometric information vector to generate a comparative odor biometric information vector. An authentication unit performs the user authentication by comparing an odor biometric information vector of the user's body to be authenticated from the odor sensor unit with the comparative biometric information vector if the user authentication is required. The authentication unit further performs an incremental learning of the comparative odor biometric information vector using the odor biometric information vector used in the authentication to create an incrementally learned odor biometric information vector. The comparative odor biometric information vector is updated with the incrementally learned odor biometric information vector.

Abstract: A multimedia data streaming server for effectively changing a transmit rate of multimedia data according to a network bandwidth that irregularly changes, and a method thereof are provided. The multimedia data streaming server includes a data storing unit, a monitoring unit, and a transmit rate controller. The data storing unit stores the multimedia data according to transmission levels, and the monitoring unit regularly monitors a network bandwidth state of a network streaming the multimedia data. The transmit rate controller increases the transmission level of the multimedia data to a default level if the network bandwidth state is getting bad while transmitting the multimedia data. Then, the transmit rate controller gradually increases or decreases the transmission level of the multimedia data from the default level gradually according to the network bandwidth state.