Á¤º¸°úÇÐȸ ÄÄÇ»ÆÃÀÇ ½ÇÁ¦ ³í¹®Áö (KIISE Transactions on Computing Practices)
Current Result Document :
ÇѱÛÁ¦¸ñ(Korean Title) |
½ºÆÄÅ© ±â¹Ý µö ·¯´× ºÐ»ê ÇÁ·¹ÀÓ¿öÅ© ¼º´É ºñ±³ ºÐ¼® |
¿µ¹®Á¦¸ñ(English Title) |
A Comparative Performance Analysis of Spark-Based Distributed Deep-Learning Frameworks |
ÀúÀÚ(Author) |
ÀåÀçÈñ
¹ÚÀçÈ«
±èÇÑÁÖ
À±¼º·Î
Jaehee Jang
Jaehong Park
Hanjoo Kim
Sungroh Yoon
|
¿ø¹®¼ö·Ïó(Citation) |
VOL 23 NO. 05 PP. 0299 ~ 0303 (2017. 05) |
Çѱ۳»¿ë (Korean Abstract) |
µö ·¯´×(Deep learning)Àº ±âÁ¸ Àΰø ½Å°æ¸Á ³» °èÃþ ¼ö¸¦ Áõ°¡½ÃÅ´°ú µ¿½Ã¿¡ È¿°úÀûÀÎ ÇнÀ ¹æ¹ý·ÐÀ» Á¦½ÃÇÔÀ¸·Î½á °´Ã¼/À½¼º ÀÎ½Ä ¹× ÀÚ¿¬¾î ó¸® µî °í¼öÁØ ¹®Á¦ ÇØ°á¿¡ ÀÖ¾î °ý¸ñÇÒ¸¸ÇÑ ¼º°ú¸¦ º¸ÀÌ°í ÀÖ´Ù. ±×·¯³ª ÇнÀ¿¡ ÇÊ¿äÇÑ ½Ã°£°ú ¸®¼Ò½º°¡ Å©´Ù´Â ÇѰ踦 Áö´Ï°í ÀÖ¾î, À̸¦ ÁÙÀ̱â À§ÇÑ ¿¬±¸°¡ È°¹ßÈ÷ ÁøÇàµÇ°í ÀÖ´Ù. º» ¿¬±¸¿¡¼´Â ¾ÆÆÄÄ¡ ½ºÆÄÅ© ±â¹Ý Ŭ·¯½ºÅÍ ÄÄÇ»Æà ÇÁ·¹ÀÓ¿öÅ© »ó¿¡¼ µö ·¯´×À» ºÐ»êÈÇÏ´Â µÎ °¡Áö Åø(DeepSpark, SparkNet)ÀÇ ¼º´ÉÀ» ÇнÀ Á¤È®µµ¿Í ¼Óµµ Ãø¸é¿¡¼ ÃøÁ¤ÇÏ°í ºÐ¼®ÇÏ¿´´Ù. CIFAR-10/CIFAR-100 µ¥ÀÌÅ͸¦ »ç¿ëÇÑ ½ÇÇè¿¡¼ SparkNetÀº ÇнÀ °úÁ¤ÀÇ Á¤È®µµ º¯µ¿ ÆøÀÌ ÀûÀº ¹Ý¸é DeepSpark´Â ÇнÀ Ãʱâ Á¤È®µµ´Â º¯µ¿ ÆøÀÌ Å©Áö¸¸ Á¡Â÷ º¯µ¿ ÆøÀÌ ÁÙ¾îµé¸é¼ SparkNet ´ëºñ ¾à 15% ³ôÀº Á¤È®µµ¸¦ º¸¿´°í, Á¶°Ç¿¡ µû¶ó ´ÜÀÏ ¸Ó½Åº¸´Ùµµ ³ôÀº Á¤È®µµ·Î º¸´Ù ºü¸£°Ô ¼ö·ÅÇÏ´Â ¾ç»óÀ» È®ÀÎÇÒ ¼ö ÀÖ¾ú´Ù. |
¿µ¹®³»¿ë (English Abstract) |
By piling up hidden layers in artificial neural networks, deep learning is delivering outstanding performances for high-level abstraction problems such as object/speech recognition and natural language processing. Alternatively, deep-learning users often struggle with the tremendous amounts of time and resources that are required to train deep neural networks. To alleviate this computational challenge, many approaches have been proposed in a diversity of areas. In this work, two of the existing Apache Spark-based acceleration frameworks for deep learning SparkNet and DeepSpark, are compared and analyzed in terms of the training accuracy and the time demands. In the authors' experiments with the CIFAR-10 and CIFAR-100 benchmark datasets, SparkNet showed a more stable convergence behavior than DeepSpark; but in terms of the training accuracy, DeepSpark delivered a higher classification accuracy of approximately 15%. For some of the cases, DeepSpark also outperformed the sequential implementation running on a single machine in terms of both the accuracy and the running time. |
Å°¿öµå(Keyword) |
µö ·¯´×
½ºÆÄÅ©
Caffe
ºÐ»ê ÄÄÇ»ÆÃ
Ŭ·¯½ºÅÍ ÄÄÇ»ÆÃ
deep learning
Apache Spark
Caffe
parallel computing
cluster computing
|
ÆÄÀÏ÷ºÎ |
PDF ´Ù¿î·Îµå
|