TY - GEN
T1 - Visualization of feature evolution during convolutional neural network training
AU - Punjabi, Arjun
AU - Katsaggelos, Aggelos K.
N1 - Funding Information:
The authors would like to acknowledge the Integrated Data Driven Discovery in Earth and Astrophysical Sciences (IDEAS) program at Northwestern University for support (NSF Research Traineeship Grant 1450006).
Publisher Copyright:
© EURASIP 2017.
PY - 2017/10/23
Y1 - 2017/10/23
N2 - Convolutional neural networks (CNNs) are a staple in the fields of computer vision and image processing. These networks perform visual tasks with state-of-the-art accuracy; yet, the understanding behind the success of these algorithms is still lacking. In particular, the process by which CNNs learn effective task-specific features is still unclear. This work elucidates such phenomena by applying recent deep visualization techniques during different stages of the training process. Additionally, this investigation provides visual justification to the benefits of transfer learning. The results are in line with previously discussed notions of feature specificity, and show a new facet of a particularly vexing machine learning pitfall: overfitting.
AB - Convolutional neural networks (CNNs) are a staple in the fields of computer vision and image processing. These networks perform visual tasks with state-of-the-art accuracy; yet, the understanding behind the success of these algorithms is still lacking. In particular, the process by which CNNs learn effective task-specific features is still unclear. This work elucidates such phenomena by applying recent deep visualization techniques during different stages of the training process. Additionally, this investigation provides visual justification to the benefits of transfer learning. The results are in line with previously discussed notions of feature specificity, and show a new facet of a particularly vexing machine learning pitfall: overfitting.
KW - Convolutional neural network
KW - Deep learning
KW - Feature visualization
KW - Transfer learning
UR - http://www.scopus.com/inward/record.url?scp=85041518685&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85041518685&partnerID=8YFLogxK
U2 - 10.23919/EUSIPCO.2017.8081219
DO - 10.23919/EUSIPCO.2017.8081219
M3 - Conference contribution
AN - SCOPUS:85041518685
T3 - 25th European Signal Processing Conference, EUSIPCO 2017
SP - 311
EP - 315
BT - 25th European Signal Processing Conference, EUSIPCO 2017
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 25th European Signal Processing Conference, EUSIPCO 2017
Y2 - 28 August 2017 through 2 September 2017
ER -