公开(公告)号：KR101356786B1

公开(公告)日：2014-01-28

申请号：KR1020120027638

申请日：2012-03-19

Applicant: 한국과학기술원

Inventor： 유회준 ,  김경훈 ,  이규호 ,  홍인준

IPC: G06N3/02

Abstract: 본 발명의 시각피질의 복합기둥을 모방한 인지 프로세서는 시각 정보를 처리하는 기본 단위인 복합기둥 모방 연산기를 복수개 포함하는 복합기둥 모방 연산기 묶음; 및 상기 복합기둥 모방 연산기 묶음 내의 상기 복수개의 복합기둥 모방 연산기 사이를 연결하는 회백질 네트워크를 포함하며, 상기 회백질 네트워크를 통해 상기 복합기둥 모방 연산기 각각은 자신과 인접한 복합기둥 모방 연산기들과 연결되어 국지적 정보 전달이 가능하다.

公开(公告)号：KR1020130106039A

公开(公告)日：2013-09-27

申请号：KR1020120027638

申请日：2012-03-19

Applicant: 한국과학기술원

Inventor： 유회준 ,  김경훈 ,  이규호 ,  홍인준

IPC: G06N3/02

CPC classification number: G06N3/06

Abstract: PURPOSE: A cognitive processor is provided to process visual information with low power in real time by connecting each composite column through a network using a composite column emulator as a basic operator. CONSTITUTION: A composite column emulator bundle (100) includes a plurality of composite column emulators which is a basic unit processing visual information. A gray matter network connects between the plurality of composite column emulator within the composite column emulator bundle. Each of the composite column emulators transmits local information by being connected to the adjacent composite column emulators. A white matter network connects between the plurality of composite column emulator bundles. Each of composite emulators (10) prepares information about color information of stimulus, directivity information of stimulus or origin of stimulus. [Reference numerals] (10) Composite column emulator; (100) Composite column emulator bundle; (AA) Local information; (BB) Universal information; (CC) L composite column emulator bundles

Abstract translation: 目的：通过使用复合列仿真器作为基本操作符通过网络连接每个复合列，提供认知处理器来实时处理低功率的视觉信息。 构成：复合列仿真器束（100）包括多个复合列仿真器，它们是处理视觉信息的基本单元。 灰色物质网络连接在复合列模拟器束内的多个复合柱模拟器之间。 每个复合列仿真器通过连接到相邻的复合列仿真器来发送本地信息。 白色网络连接在多个复合柱模拟器束之间。 复合仿真器（10）中的每一个准备关于刺激的颜色信息，刺激的方向性信息或刺激起源的信息。 （附图标记）（10）复合列模拟器; （100）复合柱模拟器束; （AA）本地信息; （BB）通用信息; （CC）L复合柱模拟器束

公开(公告)号：KR1020020054228A

公开(公告)日：2002-07-06

申请号：KR1020000083261

申请日：2000-12-27

Applicant: 한국전자통신연구원 ,  한국과학기술원

Inventor： 최길영 ,  이종권 ,  김탁곤 ,  이규호

IPC: H04L12/18

CPC classification number: H04L47/10 ,  H04L12/1886 ,  H04L45/16

Abstract: PURPOSE: A method for controlling flow by using an active network technique in multipoint communication is provided to provide the scalability even though receivers are increased or are in a wide area, and minimize packet loss in even case that congested situation is generated within a network. CONSTITUTION: Data packets arrive at receivers and intermediate nodes(201). It is determined whether the packets are the first arrived packets of multicast communication(202). A soft storage in the node is initialized to start controlling multicast flow included in a program(203). Data packets arrived at the nodes, are saved at a buffer(204). An output port enabling transmission is searched(207). If data packet is in priority to be transmitted to the corresponding output port(208), data packet is duplicated ad transmitted to bottom nodes in case that the number of allowable credit to the corresponding bottom node is larger than 0(209). If the current node is the final receiver(205), packets are transmitted to an application layer and counter is increased by 1(206). If the counted value becomes a reference counter of flow control and a double number of credit update unit(211), the current node decides the amount of packets receivable from the top node based on the size of virtual buffer to each bottom node and generates credit packets including related information to the top node(213). It is determined whether data packets were transmitted to all corresponding output ports or the application layer(214).

Abstract translation: 目的：提供一种通过在多点通信中使用活动网络技术来控制流量的方法，以便即使接收机增加或处于广泛区域也能提供可扩展性，并且即使在网络内产生拥塞情况的情况下也可最小化分组丢失。 规定：数据包到达接收器和中间节点（201）。 确定分组是否是第一到达的多播通信分组（202）。 初始化节点中的软存储器以开始控制程序（203）中包括的多播流程。 到达节点的数据包被保存在缓冲区（204）。 搜索启用传输的输出端口（207）。 如果数据包优先发送到对应的输出端口（208），则在相应底部节点的允许信用数量大于0（209）的情况下，将数据包复制广播发送到底层节点。 如果当前节点是最终的接收者（205），则分组被传送到应用层，计数器增加1（206）。 如果计数值成为流量​​控制的参考计数器和双倍的信用更新单元（211），则当前节点根据每个底层节点的虚拟缓冲区的大小来确定从顶层节点接收的分组的数量并生成信用 分组包括相关信息到顶层节点（213）。 确定数据分组是否被传送到所有对应的输出端口或应用层（214）。

公开(公告)号：KR100383665B1

公开(公告)日：2003-05-16

申请号：KR1020000083261

申请日：2000-12-27

Applicant: 한국전자통신연구원 ,  한국과학기술원

Inventor： 최길영 ,  이종권 ,  김탁곤 ,  이규호

IPC: H04L12/18

Abstract: PURPOSE: A method for controlling flow by using an active network technique in multipoint communication is provided to provide the scalability even though receivers are increased or are in a wide area, and minimize packet loss in even case that congested situation is generated within a network. CONSTITUTION: Data packets arrive at receivers and intermediate nodes(201). It is determined whether the packets are the first arrived packets of multicast communication(202). A soft storage in the node is initialized to start controlling multicast flow included in a program(203). Data packets arrived at the nodes, are saved at a buffer(204). An output port enabling transmission is searched(207). If data packet is in priority to be transmitted to the corresponding output port(208), data packet is duplicated ad transmitted to bottom nodes in case that the number of allowable credit to the corresponding bottom node is larger than 0(209). If the current node is the final receiver(205), packets are transmitted to an application layer and counter is increased by 1(206). If the counted value becomes a reference counter of flow control and a double number of credit update unit(211), the current node decides the amount of packets receivable from the top node based on the size of virtual buffer to each bottom node and generates credit packets including related information to the top node(213). It is determined whether data packets were transmitted to all corresponding output ports or the application layer(214).

Abstract translation: 目的：提供一种在多点通信中通过使用主动网络技术来控制流的方法，以提供即使接收器增加或处于广域中也具有可扩展性，并且即使网络中产生拥塞情况也能使包丢失最小化。 构成：数据包到达接收机和中间节点（201）。 确定分组是否是多播通信的第一个到达的分组（202）。 初始化节点中的软存储器以开始控制包含在程序中的组播流（203）。 到达节点的数据包被保存在缓冲区（204）。 搜索（207）输出端口使能传输。 如果数据分组优先被发送到相应的输出端口（208），则在对应的底部节点的允许信用数量大于0的情况下，数据分组被复制和传输到底部节点（209）。 如果当前节点是最终接收器（205），则将分组发送到应用层并且将计数器增加1（206）。 如果计数值变为流量控制的参考计数器和双倍信用更新单元（211），则当前节点基于到每个底部节点的虚拟缓冲区的大小来确定从顶端节点可接收的分组量，并生成信用 将包括相关信息的分组发送给顶层节点（213）。 确定数据分组是否被发送到所有对应的输出端口或应用层（214）。