Á¤º¸°úÇÐȸ³í¹®Áö (Journal of KIISE)
ÇѱÛÁ¦¸ñ(Korean Title) |
CNN ¸ðµ¨ÀÇ ±×·¡µð¾ðÆ® ÇÃ·Î¿ì ºÐ¼®°ú ¼º´É ºñ±³ |
¿µ¹®Á¦¸ñ(English Title) |
Gradient Flow Analysis and Performance Comparison of CNN Models |
ÀúÀÚ(Author) |
¹Ú½½±â
È«¸í´ö
Á¶±Ù½Ä
Seulgi Park
Myungduk Hong
Geunsik Jo
³ë¼³Çö
Seol-Hyun Noh
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 48 NO. 01 PP. 0100 ~ 0106 (2021. 01) |
Çѱ۳»¿ë (Korean Abstract) |
CNNs(Convolutional Neural Networks)Àº ÄÄÇ»ÅÍ ½Ã°¢ ÀνÄ(Computer vision)°ú ÀÚ¿¬¾î ó¸®(Natural language processing) ºÐ¾ß¿¡¼ ¶Ù¾î³ ¼º´ÉÀ» º¸¿© °¡Àå ³Î¸® »ç¿ëµÇ°í ÀÖ´Â µö·¯´× ¹æ¹ýÀÌ ´Ù. CNNsÀº ÀԷµ¥ÀÌÅÍ¿¡ ÄÁº¼·ç¼Ç ·¹À̾ ¿¬¼ÓÀûÀ¸·Î Àû¿ëÇÏ´Â ±¸Á¶¸¦ ÅëÇØ ÀÔ·Â µ¥ÀÌÅÍÀÇ locality¿Í correlationÀ» È¿°úÀûÀ¸·Î ÃßÃâÇÏ¿© CNNsÀÇ ±íÀÌ°¡ ±í¾îÁú¼ö·Ï ½Å°æ¸ÁÀÇ ¼º´ÉÀÌ Çâ»óµÇ¾î¿Ô´Ù. ±×·¯³ª CNNsÀÇ ±íÀÌ°¡ ±í¾îÁú¼ö·Ï ½Å°æ¸ÁÀÇ Á¤È®µµ°¡ ¹Ýµå½Ã ³ô¾ÆÁö´Â °ÍÀº ¾Æ´Ï´Ù. ±×·¡µð¾ðÆ® ¼Ò½Ç ¹®Á¦ (Gradient vanishing problem)À¸·Î ÀÎÇØ weighted layersÀÇ °¡ÁßÄ¡µéÀÌ ¼ö·ÅÇÏÁö ¾Ê´Â Çö»óÀÌ ¹ß»ýÇÒ ¼ö Àֱ⠶§¹®ÀÌ´Ù. ÀÌ¿¡ º» ¿¬±¸¿¡¼´Â VGGNet ¸ðµ¨, ResNet ¸ðµ¨, DenseNet ¸ðµ¨ÀÇ gradient flow¸¦ ºÐ¼®ÇÏ°í ºñ±³ÇÔÀ¸·Î½á °¢ ¸ðµ¨ÀÇ error rate ¼º´É¿¡ Â÷ÀÌ°¡ ³ª´Â ±Ù°Å¸¦ µµÃâÇÏ¿´´Ù.
|
¿µ¹®³»¿ë (English Abstract) |
Among the various deep learning techniques available, convolutional neural networks (CNNs) are widely used due to their superior performance in the fields of computer vision and natural language processing. CNNs can effectively extract the locality and correlation of input data using structures wherein convolutional layers are successively applied to the input data. The performance of neural networks has generally been improved as the depth of CNNs has increased. However, an increase in the depth of a CNN is not always accompanied by a corresponding increase in the accuracy of the neural network. This is because the gradient vanishing problem may occur, thereby causing the weights of the weighted layers to fail to converge. Accordingly, in the present study, the gradient flows of the VGGNet model, ResNet model, and DenseNet model were analyzed and compared, and reasons for the differences in the error rate performances of the models was derived.
|
Å°¿öµå(Keyword) |
ÀÌ»ó ŽÁö
±¤ÇÐ È帧
°´Ã¼ Áß½É
µö·¯´×
ÀΰøÁö´É
anomaly detection
optical flow
object-centric
deep learning
artificial intelligence
CNN
±×·¡µð¾ðÆ® ¼Ò½Ç
±×·¡µð¾ðÆ® Ç÷οì
¼º´É ºñ±³
¿À·ùÀ²
CNN
gradient vanishing problem
gradient flow
performance comparison
error rate
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|