• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.13 no.6
• /
• pp.470-476
• /
• 2020

In this paper, we present the design of modular replaceable AI server for image deep learning that separates the server from the Edge Device so as to drive the AI block and the method of data transmission and reception. The modular replaceable AI server for image deep learning can reduce the dependency between social robots and edge devices where the robot's platform will be operated to improve drive stability. When a user requests a function from an AI server for interaction with a social robot, modular functions can be used to return only the results. Modular functions in AI servers can be easily maintained and changed by each module by the server manager. Compared to existing server systems, modular replaceable AI servers produce more efficient performance in terms of server maintenance and scale differences in the programs performed. Through this, more diverse image deep learning can be included in robot scenarios that allow human-robot interaction, and more efficient performance can be achieved when applied to AI servers for image deep learning in addition to robot platforms.

• The Transactions of the Korea Information Processing Society
• /
• v.4 no.2
• /
• pp.531-542
• /
• 1997

In this paper,we propose applicable methodology using simple scenarios for the architecture and its procedure of sibnalling at the NNI for B-ISDN in Korea.By the recent technologies,the flexible and intergrated networks of the future are designed, and the infrastructure hardware and software of those networks are sdsigned based on the modular concepts.Also the ATM-based B-ISDN must be able to support the broadband transmission function for providing various services with diverse bandwidths such as multimedia serivce and the function for controlling services and bandwidth is necessaty.Currently,the trquirements at the NNI for B-ISDN in korea atr based on the network mode connection type which is connection setup strucuture.They must show the upper/lower layer conformance to have the flexibility in new services and guarantee the interoperability between the network functions.But,network functions and protocol specifications described in this paper are limity to the first stage target system in korea.Therefore,we present the modular architecture which is corresponding to the B-ISUP of HAN/B-ISDN with those characteristics.Currently applicable architecture and procedure for B-ISUP and the functions of expandable modular architecture are also proposed.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.39 no.1
• /
• pp.16-23
• /
• 2002

In this paper, we propose the growing and pruning algorithm to design the optimal structure of modular wavelet neural network(MWNN) with F-projection and geometric growing criterion. Geometric growing criterion consists of estimated error criterion considering local error and angle criterion which attempts to assign wavelet function that is nearly orthogonal to all other existing wavelet functions. These criteria provide a methodology which a network designer can construct MWNN according to one's intention. The proposed growing algorithm increases in number of module or the size of modules of MWNN. Also, the pruning algorithm eliminates unnecessary node of module or module from constructed MWNN to overcome the problem due to localized characteristic of wavelet neural network which is used to modules of MWNN. We apply the proposed constructing algorithm of the optimal structure of MWNN to approximation problems of 1-D function and 2-D function, and evaluate the effectiveness of the proposed algorithm.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2001.12a
• /
• pp.10-13
• /
• 2001

Nonlinear Function Approximation of Moduled Neural Network Using Genetic Algorithm Neural Network consists of neuron and synapse. Synapse memorize last pattern and study new pattern. When Neural Network learn new pattern, it tend to forget previously learned pattern. This phenomenon is called to catastrophic inference or catastrophic forgetting. To overcome this phenomenon, Neural Network must be modularized. In this paper, we propose Moduled Neural Network. Modular Neural Network consists of two Neural Network. Each Network individually study different pattern and their outputs is finally summed by net function. Sometimes Neural Network don't find global minimum, but find local minimum. To find global minimum we use Genetic Algorithm.

• East Asian mathematical journal
• /
• v.35 no.1
• /
• pp.67-75
• /
• 2019

We investigate the averaging value of a random sampling of a Dirichlet series with some condition using Poisson distribution. Our result is the following: Let $L(s)={\sum}^{\infty}_{n=1}{\frac{a_n}{n^s}}$ be a Dirichlet series that converges absolutely for Re(s) > 1. If $X_t$ is an increasing random sampling with Poisson distribution and there exists a number $0<{\alpha}<{\frac{1}{2}}$ such that ${\sum}_{n{\leq}u}a_n{\ll}u^{\alpha}$, then we have $${\mathbb{E}}L(1/2+iX_t)=O(t^{\alpha}{\sqrt{{\log}t}})$$, for all sufficiently large t in ${\mathbb{R}}$. As a result, we get the behaviour of $L({\frac{1}{2}}+it)$ such that L is a Dirichlet L-function or a modular L-function, when t is sampled by the Poisson distribution.

• Korean Journal of Mathematics
• /
• v.25 no.1
• /
• pp.87-97
• /
• 2017

In [9], we computed shadows of the second order mock theta functions and showed that they are essentially same with the shadow of a mock theta function related to the Mathieu moonshine phenomenon. In this paper, we further survey the second order mock theta functions on their quantum modularity and their behavior in the lower half plane.

• Journal of the Chungcheong Mathematical Society
• /
• v.22 no.4
• /
• pp.799-814
• /
• 2009

A class invariant is the value of a modular function that generates a ring class field of an imaginary quadratic number field such as the singular moduli of level 1. In this paper, using Shimura Reciprocity Law, we compute the Galois actions of some class invariants from the generalized Weber functions $\mathfrak{g}_0,\mathfrak{g}_1,\mathfrak{g}_2$ and $\mathfrak{g}_3$ over quadratic number fields with discriminant $D{\equiv}1$ (mod 12).

• Journal of the Korean Society for Precision Engineering
• /
• v.17 no.5
• /
• pp.34-40
• /
• 2000

최근 CNC에는 또 하나의 전기를 마련할 수 있는 여러 가지 새로운 이슈들이 등장하고 있다. 첫 번째 이슈는 고도로 발달해가고 있는 반도체 산업의 영항으로 CPU의 속도가 기하급수적으로 달라지고 있고 메모리 용량이 상상을 초월할 정도로 급속한 발전을 거듭하고 있다. 이런 환경 속에서 자연스럽게 CNC도 성능이 우수한 하드웨어와 그 성능에 어울리는 고급 소프트웨어 탑재가 요구된다. 두 번째 이슈는 기존의 공작기계에서 생각하지 못했던 개념의 변화이다.(중략)

• Proceedings of the IEEK Conference
• /
• 2002.06c
• /
• pp.127-130
• /
• 2002

Human's brain action can divide by recognition and intelligence. recognition is sensing voice, image and smell and Intelligence is logical judgment, inference, decision. To this concept, Define function of cerebral cortex, and apply the result. Current expert system is lack, that reasoning by cerebral cortex and thalamus, hoppocampal and so on. In this paper, With human's brain action, wish to embody human's action artificially Embody brain mechanism using Modular Neural Network, Applied this result to snake robot.

• Journal of the Korean Mathematical Society
• /
• v.51 no.6
• /
• pp.1269-1289
• /
• 2014

We first find a sufficient condition for a product of theta constants to be a Siegel modular function of a given even level. And, when $K_{(2p)}$ denotes the ray class field of $K=\mathbb{Q}(e^{2{\pi}i/5})$ modulo 2p for an odd prime p, we describe a subfield of $K_{(2p)}$ generated by the special value of a certain theta constant by using Shimura's reciprocity law.