TY - GEN
T1 - Incorporating Prior Domain Knowledge into Deep Neural Networks
AU - Muralidhar, Nikhil
AU - Islam, Mohammad Raihanul
AU - Marwah, Manish
AU - Karpatne, Anuj
AU - Ramakrishnan, Naren
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/7/2
Y1 - 2018/7/2
N2 - In recent years, the large amount of labeled data available has also helped tend research toward using minimal domain knowledge, e.g., in deep neural network research. However, in many situations, data is limited and of poor quality. Can domain knowledge be useful in such a setting? In this paper, we propose domain adapted neural networks (DANN) to explore how domain knowledge can be integrated into model training for deep networks. In particular, we incorporate loss terms for knowledge available as monotonicity constraints and approximation constraints. We evaluate our model on both synthetic data generated using the popular Bohachevsky function and a real-world dataset for predicting oxygen solubility in water. In both situations, we find that our DANN model outperforms its domain-agnostic counterpart yielding an overall mean performance improvement of 19.5% with a worst- and best-case performance improvement of 4% and 42.7%, respectively.
AB - In recent years, the large amount of labeled data available has also helped tend research toward using minimal domain knowledge, e.g., in deep neural network research. However, in many situations, data is limited and of poor quality. Can domain knowledge be useful in such a setting? In this paper, we propose domain adapted neural networks (DANN) to explore how domain knowledge can be integrated into model training for deep networks. In particular, we incorporate loss terms for knowledge available as monotonicity constraints and approximation constraints. We evaluate our model on both synthetic data generated using the popular Bohachevsky function and a real-world dataset for predicting oxygen solubility in water. In both situations, we find that our DANN model outperforms its domain-agnostic counterpart yielding an overall mean performance improvement of 19.5% with a worst- and best-case performance improvement of 4% and 42.7%, respectively.
KW - Deep Learning
KW - Domain Knowledge
KW - Limited Training Data
KW - Neural Networks
KW - Noisy Data
UR - http://www.scopus.com/inward/record.url?scp=85062590504&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85062590504&partnerID=8YFLogxK
U2 - 10.1109/BigData.2018.8621955
DO - 10.1109/BigData.2018.8621955
M3 - Conference contribution
AN - SCOPUS:85062590504
T3 - Proceedings - 2018 IEEE International Conference on Big Data, Big Data 2018
SP - 36
EP - 45
BT - Proceedings - 2018 IEEE International Conference on Big Data, Big Data 2018
A2 - Abe, Naoki
A2 - Liu, Huan
A2 - Pu, Calton
A2 - Hu, Xiaohua
A2 - Ahmed, Nesreen
A2 - Qiao, Mu
A2 - Song, Yang
A2 - Kossmann, Donald
A2 - Liu, Bing
A2 - Lee, Kisung
A2 - Tang, Jiliang
A2 - He, Jingrui
A2 - Saltz, Jeffrey
T2 - 2018 IEEE International Conference on Big Data, Big Data 2018
Y2 - 10 December 2018 through 13 December 2018
ER -