臺灣博碩士論文加值系統

相較於傳統的機器學習演算法，現今大多數的演算法都在正確率上有顯著的提升，但是這些模型的架構也因此變得愈來愈複雜，讓我們無法從中理解預測是如何產生。這導致資料中可能潛藏的歧視難以被人類發覺，因此出現法規要求模型的可解釋性。然而，現有的可解釋模型（例如：決策樹、線性模型）太過於簡單，處理大型的複雜資料時無法產生足夠準確的預測。因此，我們從組成隨機森林的決策樹當中萃取規則，一來讓原本被認為是黑盒模型的隨機森林具有可解釋性，二來運用整體學習讓演算法能夠得到較佳的準確率。此外，借用深度學習中表徵學習的概念，我們加上深層的模型結構，讓隨機森林能學習更加複雜的特徵。在這篇論文當中，我們提出深度規則森林，同時結合可解釋性和深層模型結構，也在實驗中取得超越隨機森林等複雜模型的表現。但是這樣的結構卻導致其中的規則太過複雜不易理解，因此失去可解釋性。我們提出邏輯最佳化演算法，將萃取出的規則簡化，使之能成為易於人們閱讀且理解的形式並保留可解釋性。

Compared to traditional machine learning algorithms, most contemporary algorithms have prominent promotion in terms of accuracy, but this also complicate the model architecture, which disables human from understanding how the predictions are generated. This makes the latent discrimination in data difficult for human to discover, and thus there are legislations enforce that models should have interpretability. However, recent interpretable models (e.g. decision tree, linear model) are too simple to produce enough accurate predictions in case of dealing large and complex datasets. Therefore, we extract rules from the decision tree component in random forest, not only makes random forest, regarded as black box model, interpretable, but exploits ensemble learning to boost the accuracy. Moreover, inspired by the concept of representation learning in deep learning, we add multilayer structure to enable random forest to learn more complicated representation. In this paper, we propose Deep Rule Forest, with both interpretability and deep model architecture, and it outperform several complex models such as random forest on accuracy. Nevertheless, this structure makes the rules too complicated to understand by human and hence lose interpretability. At last, via logic optimization, we retain interpretability by simplifying the rules and making them readable and understandable to human.

摘要 ii
Abstract iii
List of Figures v
List of Table vi
1. Introduction 1
2. Background and Related Work 2
2.1. Tree-based Algorithms 2
2.2. Representation Learning 6
2.3. Deep Architecture 8
2.4. Deep Architecture Models 9
2.5. Explainable AI (XAI) 12
2.6. Logic Optimization (Logic Minimization) 13
3. Methodology 14
3.1. Building DRF 15
3.2. Interpretability of DRF 18
3.3. New Encoding for Regression Data 22
4. Experiment and Discussion 24
4.1. Experiment Setup 24
4.2. Classification with DRF 25
4.3. Regression with DRF 27
5. Conclusion 30
6. Reference 31

Bengio, Y. (2009). Learning Deep Architectures for AI. Foundations and Trends® in Machine Learning, 2(1), 1–127. https://doi.org/10.1561/2200000006
Bengio, Yoshua. (2013). Deep Learning of Representations: Looking Forward. ArXiv:1305.0445 [Cs]. Retrieved from http://arxiv.org/abs/1305.0445
Bengio, Yoshua, Courville, A., & Vincent, P. (2012). Representation Learning: A Review and New Perspectives. ArXiv:1206.5538 [Cs]. Retrieved from http://arxiv.org/abs/1206.5538
Bengio, Yoshua, Delalleau, O., & Simard, C. (2010). DECISION TREES DO NOT GENERALIZE TO NEW VARIATIONS. Computational Intelligence, 26(4), 449–467. https://doi.org/10.1111/j.1467-8640.2010.00366.x
Bengio, Yoshua, Lamblin, P., Popovici, D., & Larochelle, H. (2007). Greedy Layer-Wise Training of Deep Networks. In B. Schölkopf, J. C. Platt, & T. Hoffman (Eds.), Advances in Neural Information Processing Systems 19 (pp. 153–160). Retrieved from http://papers.nips.cc/paper/3048-greedy-layer-wise-training-of-deep-networks.pdf
Bergmeir, C., & Benítez, J. M. (2012). Neural Networks in R Using the Stuttgart Neural Network Simulator: RSNNS. Journal of Statistical Software, 46(7). https://doi.org/10.18637/jss.v046.i07
Boser, B. E., Guyon, I. M., & Vapnik, V. N. (1992). A Training Algorithm for Optimal Margin Classifiers. Proceedings of the Fifth Annual Workshop on Computational Learning Theory, 144–152. https://doi.org/10.1145/130385.130401
Bourlard, H., & Kamp, Y. (1988). Auto-association by multilayer perceptrons and singular value decomposition. Biological Cybernetics, 59(4), 291–294. https://doi.org/10.1007/BF00332918
Breiman, L. (1996). Bagging predictors. Machine Learning, 24(2), 123–140. https://doi.org/10.1007/BF00058655
Breiman, L. (2017). Classification and Regression Trees. https://doi.org/10.1201/9781315139470
Breiman, L. (n.d.). RANDOM FORESTS--RANDOM FEATURES. 29.
Bunn, A., & Korpela, M. (n.d.). An introduction to dplR. 16.
Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining - KDD ’16, 785–794. https://doi.org/10.1145/2939672.2939785
Chen, T., & He, T. (n.d.). xgboost: EXtreme Gradient Boosting. 4.
Cichocki, A., Zdunek, R., & Amari, S. (2007). Hierarchical ALS Algorithms for Nonnegative Matrix and 3D Tensor Factorization. In M. E. Davies, C. J. James, S. A. Abdallah, & M. D. Plumbley (Eds.), Independent Component Analysis and Signal Separation (pp. 169–176). Springer Berlin Heidelberg.
Cun, Y. L. (1987). Modèles connexionnistes de l’apprentissage.
Doshi-Velez, F., & Kim, B. (2017). Towards A Rigorous Science of Interpretable Machine Learning. ArXiv:1702.08608 [Cs, Stat]. Retrieved from http://arxiv.org/abs/1702.08608
Efron, B. (1979). Bootstrap Methods: Another Look at the Jackknife. The Annals of Statistics, 7(1), 1–26. https://doi.org/10.1214/aos/1176344552
Freund, Y., & Schapire, R. E. (1995). A Decision-Theoretic Generalization of on-Line Learning and an Application to Boosting.
Friedman, J. H. (2001). Greedy function approximation: A gradient boosting machine. https://doi.org/10.1214/aos/1013203451
Friedman, J., Hastie, T., & Tibshirani, R. (2010). Regularization Paths for Generalized Linear Models via Coordinate Descent. Journal of Statistical Software, 33(1). https://doi.org/10.18637/jss.v033.i01
F.R.S, K. P. (1901). LIII. On lines and planes of closest fit to systems of points in space. The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science, 2(11), 559–572. https://doi.org/10.1080/14786440109462720
Goodman, B., & Flaxman, S. (2017). European Union regulations on algorithmic decision-making and a “right to explanation.” AI Magazine, 38(3), 50. https://doi.org/10.1609/aimag.v38i3.2741
Guidotti, R., Monreale, A., Ruggieri, S., Turini, F., Giannotti, F., & Pedreschi, D. (2018). A Survey of Methods for Explaining Black Box Models. ACM Computing Surveys, 51(5), 1–42. https://doi.org/10.1145/3236009
Hahnloser, R. H. R., Sarpeshkar, R., Mahowald, M. A., Douglas, R. J., & Seung, H. S. (2000). Digital selection and analogue amplification coexist in a cortex-inspired silicon circuit. Nature, 405(6789), 947. https://doi.org/10.1038/35016072
Harrison, D., & Rubinfeld, D. L. (1978). Hedonic housing prices and the demand for clean air. Journal of Environmental Economics and Management, 5(1), 81–102. https://doi.org/10.1016/0095-0696(78)90006-2
He, K., Zhang, X., Ren, S., & Sun, J. (2015). Deep Residual Learning for Image Recognition. ArXiv:1512.03385 [Cs]. Retrieved from http://arxiv.org/abs/1512.03385
Hinton, G E, McCLELLAND, J. L., & Rumelhart, D. E. (n.d.). Distributed Representations. 33.
Hinton, Geoffrey E. (2009). Deep belief networks. Scholarpedia, 4(5), 5947. https://doi.org/10.4249/scholarpedia.5947
Hinton, Geoffrey E, & Zemel, R. S. (1994). Autoencoders, Minimum Description Length and Helmholtz Free Energy. In J. D. Cowan, G. Tesauro, & J. Alspector (Eds.), Advances in Neural Information Processing Systems 6 (pp. 3–10). Retrieved from http://papers.nips.cc/paper/798-autoencoders-minimum-description-length-and-helmholtz-free-energy.pdf
Hyvärinen, A., Karhunen, J., & Oja, E. (2004). Independent Component Analysis. John Wiley & Sons.
Karatzoglou, A., Smola, A., Hornik, K., & Zeileis, A. (2004). kernlab—An S4 Package for Kernel Methods in R. Journal of Statistical Software, 11(9). https://doi.org/10.18637/jss.v011.i09
Kuhn, M. (2008). Building Predictive Models in R Using the caret Package. Journal of Statistical Software, 28(5). https://doi.org/10.18637/jss.v028.i05
LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436–444. https://doi.org/10.1038/nature14539
Liaw, A., & Wiener, M. (2002). Classification and Regression by randomForest. 2, 5.
McCluskey, E. J. (1956). Minimization of Boolean functions. The Bell System Technical Journal, 35(6), 1417–1444. https://doi.org/10.1002/j.1538-7305.1956.tb03835.x
Miller, K., Hettinger, C., Humpherys, J., Jarvis, T., & Kartchner, D. (2017). Forward Thinking: Building Deep Random Forests. ArXiv:1705.07366 [Cs, Stat]. Retrieved from http://arxiv.org/abs/1705.07366
Peer to Peer Lending & Alternative Investing | Save with LendingClub. (n.d.). Retrieved November 15, 2018, from https://www.lendingclub.com/
Quine, W. V. (1952). The Problem of Simplifying Truth Functions. The American Mathematical Monthly, 59(8), 521–531. https://doi.org/10.2307/2308219
Quine, W. V. (1955). A Way to Simplify Truth Functions. The American Mathematical Monthly, 62(9), 627–631. https://doi.org/10.2307/2307285
Quinlan, J. R. (1986). Induction of decision trees. Machine Learning, 1(1), 81–106. https://doi.org/10.1007/BF00116251
Quinlan, J. Ross. (1993). C4.5: Programs for Machine Learning. San Francisco, CA, USA: Morgan Kaufmann Publishers Inc.
Ragin, C. C. (2014). The Comparative Method: Moving Beyond Qualitative and Quantitative Strategies. Univ of California Press.
Ribeiro, M. T., Singh, S., & Guestrin, C. (2016). “Why Should I Trust You?”: Explaining the Predictions of Any Classifier. ArXiv:1602.04938 [Cs, Stat]. Retrieved from http://arxiv.org/abs/1602.04938
RStudio Team. (2015). RStudio: Integrated Development for R. Retrieved from http://www.rstudio.com/.
Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Parallel Distributed Processing: Explorations in the Microstructure of Cognition, Vol. 1 (David E. Rumelhart, J. L. McClelland, & C. PDP Research Group, Eds.). Retrieved from http://dl.acm.org/citation.cfm?id=104279.104293
Santosa, F., & Symes, W. (1986). Linear Inversion of Band-Limited Reflection Seismograms. SIAM Journal on Scientific and Statistical Computing, 7(4), 1307–1330. https://doi.org/10.1137/0907087
Simonyan, K., & Zisserman, A. (2014). Very Deep Convolutional Networks for Large-Scale Image Recognition. ArXiv:1409.1556 [Cs]. Retrieved from http://arxiv.org/abs/1409.1556
Sirikulviriya, N., & Sinthupinyo, S. (n.d.). Integration of Rules from a Random Forest. 5.
Su, G., Wei, D., Varshney, K. R., & Malioutov, D. M. (2015). Interpretable Two-level Boolean Rule Learning for Classification. ArXiv:1511.07361 [Cs]. Retrieved from http://arxiv.org/abs/1511.07361
Szegedy, C., Ioffe, S., Vanhoucke, V., & Alemi, A. (2016). Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning. ArXiv:1602.07261 [Cs]. Retrieved from http://arxiv.org/abs/1602.07261
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., & Wojna, Z. (2016). Rethinking the Inception Architecture for Computer Vision. 2818–2826. Retrieved from https://www.cv-foundation.org/openaccess/content_cvpr_2016/html/Szegedy_Rethinking_the_Inception_CVPR_2016_paper.html
Therneau, T. M., Atkinson, E. J., & Foundation, M. (n.d.). An Introduction to Recursive Partitioning Using the RPART Routines. 60.
UCI Machine Learning Repository: Abalone Data Set. (n.d.). Retrieved June 28, 2019, from http://archive.ics.uci.edu/ml/datasets/Abalone
UCI Machine Learning Repository: Iris Data Set. (n.d.). Retrieved November 15, 2018, from http://archive.ics.uci.edu/ml/datasets/Iris
UCI Machine Learning Repository: Poker Hand Data Set. (n.d.). Retrieved July 14, 2019, from https://archive.ics.uci.edu/ml/datasets/Poker+Hand
Wickham, H. (2009). Ggplot2: Elegant graphics for data analysis. New York: Springer.
Zhou, Z.-H., & Feng, J. (2017). Deep Forest. ArXiv:1702.08835 [Cs, Stat]. Retrieved from http://arxiv.org/abs/1702.08835