CGXplain: Rule-Based Deep Neural Network Explanations Using Dual Linear Programs
نویسندگان
چکیده
Rule-based surrogate models are an effective and interpretable way to approximate a Deep Neural Network’s (DNN) decision boundaries, allowing humans easily understand deep learning models. Current state-of-the-art decompositional methods, which those that consider the DNN’s latent space extract more exact rule sets, manage derive sets at high accuracy. However, they a) do not guarantee model has learned from same variables as DNN (alignment), b) only allow optimising for single objective, such accuracy, can result in excessively large (complexity), c) use tree algorithms intermediate models, different explanations (stability). This paper introduces Column Generation eXplainer address these limitations – method using dual linear programming rules hidden representations of DNN. approach allows any number objectives empowers users tweak explanation their needs. We evaluate our results on wide variety tasks show CGX meets all three criteria, by having reproducibility guarantees stability reduces set size >80% (complexity) improved accuracy fidelity across (alignment).
منابع مشابه
Using Deep Neural Network Approximate Bayesian Network
We present a new method to approximate posterior probabilities of Bayesian Network using Deep Neural Network. Experiment results on several public Bayesian Network datasets shows that Deep Neural Network is capable of learning joint probability distribution of Bayesian Network by learning from a few observation and posterior probability distribution pairs with high accuracy. Compared with tradi...
متن کاملRule Based Ensembles Using Pair Wise Neural Network Classifiers
In value estimation, the inexperienced people's estimation average is good approximation to true value, provided that the answer of these individual are independent. Classifier ensemble is the implementation of mentioned principle in classification tasks that are investigated in two aspects. In the first aspect, feature space is divided into several local regions and each region is assigned wit...
متن کاملscour modeling piles of kambuzia industrial city bridge using hec-ras and artificial neural network
today, scouring is one of the important topics in the river and coastal engineering so that the most destruction in the bridges is occurred due to this phenomenon. whereas the bridges are assumed as the most important connecting structures in the communications roads in the country and their importance is doubled while floodwater, thus exact design and maintenance thereof is very crucial. f...
ذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: Lecture Notes in Computer Science
سال: 2023
ISSN: ['1611-3349', '0302-9743']
DOI: https://doi.org/10.1007/978-3-031-39539-0_6