An Empirical Evaluation of Rule Extraction from Recurrent Neural Networks
نویسندگان
چکیده
منابع مشابه
An Empirical Evaluation of Rule Extraction from Recurrent Neural Networks
Rule extraction from black-box models is critical in domains that require model validation before implementation, as can be the case in credit scoring and medical diagnosis. Though already a challenging problem in statistical learning in general, the difficulty is even greater when highly non-linear, recursive models, like recurrent neural networks (RNNs), are fit to data. Here, we study the ex...
متن کاملRule extraction from recurrent neural networks
This thesis investigates rule extraction from recurrent neural networks, which takes the form of automated construction of models of an underlying network. Typically the models are expressed as finite state machines and they should mimic the network while being more intelligible. It is argued that rule extraction allows a deeper and more general form of analysis than other, more or less ad hoc,...
متن کاملRethinking Rule Extraction from Recurrent Neural Networks
We will in this paper identify some of the central problems of current techniques for rule extraction from recurrent neural networks (RNN-RE). Then we will raise the expectations of future RNN-RE techniques considerably and through this, hopefully guide the research towards a common goal. Some preliminary results based on work in line with these goals, will also be presented.
متن کاملRule Extraction from Recurrent Neural Networks: A Taxonomy and Review
Rule extraction (RE) from recurrent neural networks (RNNs) refers to finding models of the underlying RNN, typically in the form of finite state machines, that mimic the network to a satisfactory degree while having the advantage of being more transparent. RE from RNNs can be argued to allow a deeper and more profound form of analysis of RNNs than other, more or less ad hoc methods. RE may give...
متن کاملRule Extraction from Neural Networks
The artificial neural networks (ANNs) are well suitable to solve a variety class of problems in a knowledge discovery field (e.g., in natural language processing) because the trained networks are more accurate at classifying the examples that represent a problem domain. However, the neural networks that consist of large number of weighted connections (called also links) and activation units oft...
متن کاملذخیره 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: Neural Computation
سال: 2018
ISSN: 0899-7667,1530-888X
DOI: 10.1162/neco_a_01111