LEONN
Learning and Evolution
of Neural Networks
º» ¿¬±¸´Â °úÇÐ ±â¼úºÎÀÇ ³ú °³¹ß »ç¾÷ÀÇ
ÀÏȯÀ¸·Î ÁøÇàµÇ°í ÀÖÀ¸¸ç ¸ðµ¨¸µ ÆÀÀÇ ÃÖÁ¾ ¸ñÇ¥´Â »ý¹°ÇÐÀû ½Å°æ¸ÁÀÇ
µ¿Àû Ư¼º°ú ÀÛµ¿ ¿ø¸®¿¡ ±âÃÊÇÑ »õ·Î¿î ½Ã°£Àû Á¤º¸ ó¸®ÀÇ ¿øÃµ ±â¼úÀ»
°³¹ßÇÏ´Â °ÍÀÌ´Ù
.
1Â÷³âµµÀÇ ¸ñÇ¥´Â ½Å°æ¸Á ¸ðµ¨ÀÇ º¹Àâµµ¿Í
ÇнÀ µ¥ÀÌÅÍÀÇ º¹Àâµµ ¹× À̵鰣ÀÇ »óÈ£ °ü°è¸¦ ¼öÇÐÀûÀ¸·Î ºÐ¼®Çϰí
ÀÌ ÀÌ·ÐÀû °á°ú¸¦ °Ç¼³Àû ÇнÀ ¾Ë°í¸®ÁòÀÇ ¼º´É ºÐ¼®¿¡ Àû¿ëÇϰíÀÚ ÇÏ´Â
°ÍÀÌ´Ù. ÀÌ·¯ÇÑ ¿¬±¸ÀÇ °á°ú´Â ÃßÈÄ ÀϹÝÈ ¼º´ÉÀÌ º¸´Ù ¶Ù¾î³ »õ·Î¿î
½Å°æ¸Á ¸ðµ¨°ú ÇнÀ ¾Ë°í¸®ÁòÀ» Á¦¾ÈÇϱâ À§ÇÑ Ãʼ®ÀÌ µÇ¸®¶ó ±â´ëµÈ´Ù.
º» ¿¬±¸¸¦ ¼öÇàÇϱâ À§ÇÑ ¼¼ºÎÀûÀÎ ÁøÇà ³»¿ëÀº ´ÙÀ½°ú °°´Ù.
- ÀϹÝÈ ¼º´É°ú ½Å°æ¸Á ±¸Á¶ÀÇ
º¹Àâµµ °ü°èÀÇ ¼öÇÐÀû ºÐ¼®
- Principle of Occam's Razor
- Computational Learning Theory
- Statistical Learning Theory
- ÀϹÝÈ ¼º´É°ú µ¥ÀÌÅÍ º¹ÀâµµÀÇ
°ü°è ºÐ¼®
- Probably Approximately Correct(PAC)
ÇнÀ ¸ðµ¨ ¿¬±¸
- Model Complexity¿Í Sample
Complexity °ü°è ºÐ¼®
- Statistical Learning Theory¿¡
±â¹ÝÇÑ Sample Complexity ºÐ¼®
- Regularization ±â¹ýÀ» ÀÌ¿ëÇÑ
º¹Àâµµ Ãà¼Ò ¹× ÀϹÝÈ Çâ»ó ±â¹ý
- VC Dimension ºÐ¼®
- Empirical Risk Minimization
- Structural Risk Minimization
- ÇнÀ À̷п¡ ±Ù°ÅÇÑ °Ç¼³Àû ÇнÀ
¾Ë°í¸®ÁòÀÇ ºÐ¼®
- ÀÛÀº Å©±â·Î Ãâ¹ßÇÏ¿© ¸Á
±¸Á¶ÀÇ ¼ºÀå¿¡ ÀÇÇÑ ±¸Á¶ ÃÖÀûÈ
- ERM, SRM ±â¹ÝÀÇ Statistical
Learning TheoryÀû ºÐ¼®
- ÀϹÝÈ ¼º´É Çâ»óÀ» À§ÇÑ
°Ç¼³Àû ¾Ë°í¸®ÁòÀÇ °³¼± ¹æÇâ ¿¬±¸
2Â÷³âµµÀÇ ¸ñÇ¥´Â ½Ã°ø°£ ÆÐÅÏ󸮸¦
À§ÇÑ ½Å°æ¸ÁÀ» ¸ðµ¨¸µÇϰí À̸¦ ÃÖÀûÈÇϰíÀÚ ÇÏ´Â °ÍÀÌ´Ù. À̸¦ À§ÇØ
½Ã°ø°£ µ¥ÀÌÅÍÀÇ Æ¯¼ºÀ» ºÐ¼®Çϰí, È®·üÀû, Åë°è ¹°¸®ÇÐÀû Á¢±Ù ¹æ¹ýÀ»
ÀÌ¿ëÇÏ¿© û°¢ µ¥ÀÌÅÍ Áß½ÉÀÇ ½Ã°£Àû Ư¼º ÄÚµù ¹æ¹ýÀ» ¿¬±¸ÇÑ´Ù. ±×¸®°í
ÀÌ·¯ÇÑ µ¥ÀÌÅÍ¿¡ ÀûÇÕȯ ½Å°æ¸ÁÀ» ¸ðµ¨¸µÇÏ°í ½ÇÁ¦ÀÇ Ã»°¢ µ¥ÀÌÅ͸¦
ÀÌ¿ëÇÑ ½ÇÇèÀ» ÁøÇàÇÑ´Ù. ±×¸®°í ±× ÇнÀ °á°ú¸¦ ÀÌ¿ëÇÏ¿© ½Ã°ø°£ ÆÐÅÏ
󸮸¦ À§ÇÑ ½Å°æ¸ÁÀÇ ±¸Á¶ ÃÖÀûÈ¿¡ °üÇÑ ±â¹ýÀ» ¿¬±¸ÇÑ´Ù.
³âµµ |
¸ñÇ¥ |
1998.11 ~ 1999.
7 |
½Å°æ¸ÁÀÇ º¹Àâµµ¿Í ÀϹÝÈ ¼º´É
°ü°è ºÐ¼® |
1999. 8 ~ 2000.
7 |
½Ã°ø°£ ÆÐÅÏ Ã³¸®¸¦ À§ÇÑ
»ý¹°ÇÐÀû ½Å°æ¸Á°ú ÃÖÀûÈ |
2000. 8 ~ 2001.
7 |
°¨°¢ Á¤º¸ 󸮸¦ À§ÇÑ ½Å°æ¸Á
ÇнÀ ¹× ÁøÈ ±â¹ý |
Publications
- Efficient
Model Induction by a Bayesian Evolutionary Algorithm
with Incremental Data Inheritance, Byoung-Tak Zhang,
Je-Gun Joung, 1998 (submitted to IEEE Transactions
on Evolutionary Computation)
- Bayesian
Methods for Efficient Genetic Programming, Byoung-Tak Zhang,
Genetic Programming and Evolvable Machines, vol.
1, no 3., 2000 (to appear)
- Active
Data Partitioning for Building Mixture Models, Suk-Joon Kim, Byoung-Tak
Zhang, International
Conference of Neural Information Processing, vol. 2, pp.
854-857, 1998
- Combining
Locally Trained Neural Networks by Introducing a
Reject Class, Suk-Joon Kim, Byoung-Tak Zhang, International
Joint Conference on Neural Networks 1999, (CD version)
- Temporal
Pattern Recognition using a Spiking Neural Network
with Delays, Jeongh-Woo Sohn, Byoung-Tak Zhang, International Joint Conference on Neural
Networks,
1999, (CD version)
- A Bayesian
Framework for Evolutionary Computation, Byoung-Tak Zhang,
IEEE
1999 Congress on Evolutionary Computation, pp. 722-727,
1999.
- Time
Series Prediction using Committee Machines of Evolutionary
Neural Trees, Byoung-Tak Zhang, Je-Gun Joung, IEEE 1999 Congress
on Evolutionary Computation, pp. 281-286, 1999
- Áö¿ª
À§¿øÈ¸ÀÇ ±¸Ãà°ú °áÇÕ, ±è¼®ÁØ, À庴Ź, Ãá°è Á¤º¸°úÇÐȸÁö, vol.
26, no. 1, pp. 254-256, 1999
- Áö¿ªÀû
Ư¼ºÀ» °¡Áö´Â Committee MachineÀÇ ÀϹÝÈ ¼º´É Çâ»óÀ»
À§ÇÑ ¼±ÇüÀû Áö½Ä ÃßÁ¤ ±â¹ý, ÀΰøÁö´É, ½Å°æ¸Á, ÆÛÁö ½Ã½ºÅÛ Á¾ÇÕÇмú´ëȸ
(JCEANF 99), pp. 205-212, 1999
- Time
Series Prediction Using Committee of Evolutionary
Neural Trees, Byoung-Tak Zhang, Je-Gun Joung, 1999 (submitted)
- A Bayesian
Evolutionary Approach to the Design and Learning
of Heterogeneous Neural Trees, Byoung-Tak Zhang, Je-Gun Joung, Dong-Yeon
Cho, 2000 (submitted)
- Using Stochastic Helmholtz Machine for Text Learning, Jeong-Ho Chang and Byoung-Tak Zhang, 19th International Conference on Computer Processing of Oriental Languages (ICCPOL), 2001
|
|
|