One way to interpret the behavior of a blackbox recurrent neural network (RNN) is to extract from it a more interpretable discrete computational model, like a finite state machine, that captures its behavior. In this work, we propose a new method for extracting finite automata from RNNs inspired by the state merging paradigm from grammatical inference. We demonstrate the effectiveness of our method on the Tomita languages benchmark, where we find that it is able to extract faithful automata from RNNs trained on all languages in the benchmark. We find that extraction performance is aided by the number of data provided during the extraction process, as well as, curiously, whether the RNN model is trained for additional epochs after perfectly learning its target language. We use our method to analyze this phenomenon, finding that training beyond convergence is useful because it leads to compression of the internal state space of the RNN. This finding demonstrates how our method can be used for interpretability and analysis of trained RNN models.
Amazon（亚马逊）；NEC Laboratories Europe（NEC欧洲实验室）；University of Stuttgart（斯图加特大学）
Recurrent neural networks are a widely used class of neural architectures. They have, however, two shortcomings. First, they are often treated as black-box models and as such it is difficult to understand what exactly they learn as well as how they arrive at a particular prediction. Second, they tend to work poorly on sequences requiring long-term memorization, despite having this capacity in principle. We aim to address both shortcomings with a class of recurrent networks that use a stochastic state transition mechanism between cell applications. This mechanism, which we term state-regulariza…
Recent neural network architectures such as the basic recurrent neural network (RNN) and Gated Recurrent Unit (GRU) have gained prominence as end-to-end learning architectures for natural language processing tasks. But what is the computational power of such systems? We prove that finite precision RNNs with one hidden layer and ReLU activation and finite precision GRUs are exactly as computationally powerful as deterministic finite automata. Allowing arbitrary precision, we prove that RNNs with one hidden layer and ReLU activation are at least as computationally powerful as pushdown automata. …
Mila；Université de Montréal（蒙特利尔大学）；McGill University（麦吉尔大学）
In this paper, we unravel a fundamental connection between weighted finite automata~(WFAs) and second-order recurrent neural networks~(2-RNNs): in the case of sequences of discrete symbols, WFAs and 2-RNNs with linear activation functions are expressively equivalent. Motivated by this result, we build upon a recent extension of the spectral learning algorithm to vector-valued WFAs and propose the first provable learning algorithm for linear 2-RNNs defined over sequences of continuous input vectors. This algorithm relies on estimating low rank sub-blocks of the so-called Hankel tensor, from whi…
Recurrent Neural Networks (RNNs) have achieved tremendous success in sequential data processing. However, it is quite challenging to interpret and verify RNNs' behaviors directly. To this end, many efforts have been made to extract finite automata from RNNs. Existing approaches such as exact learning are effective in extracting finite-state models to characterize the state dynamics of RNNs for formal languages, but are limited in the scalability to process natural languages. Compositional approaches that are scablable to natural languages fall short in extraction precision. In this paper, we i…
Uncertainty quantification is crucial for building reliable and trustable machine learning systems. We propose to estimate uncertainty in recurrent neural networks (RNNs) via stochastic discrete state transitions over recurrent timesteps. The uncertainty of the model can be quantified by running a prediction several times, each time sampling from the recurrent state transition distribution, leading to potentially different results if the model is uncertain. Alongside uncertainty quantification, our proposed method offers several advantages in different settings. The proposed method can (1) lea…
Recurrent neural networks are a widely used class of neural architectures. They have, however, two shortcomings. First, it is difficult to understand what exactly they learn. Second, they tend to work poorly on sequences requiring long-term memorization, despite having this capacity in principle. We aim to address both shortcomings with a class of recurrent networks that use a stochastic state transition mechanism between cell applications. This mechanism, which we term state-regularization, makes RNNs transition between a finite set of learnable states. We evaluate state-regularized RNNs on (…
University of Luxembourg（卢森堡大学）；Delft Technical University（代尔夫特理工大学）
We present an interactive version of an evidence-driven state-merging (EDSM) algorithm for learning variants of finite state automata. Learning these automata often amounts to recovering or reverse engineering the model generating the data despite noisy, incomplete, or imperfectly sampled data sources rather than optimizing a purely numeric target function. Domain expertise and human knowledge about the target domain can guide this process, and typically is captured in parameter settings. Often, domain expertise is subconscious and not expressed explicitly. Directly interacting with the learni…
Rice University（莱斯大学）；Baylor College of Medicine（贝勒医学院）
We investigate the internal representations that a recurrent neural network (RNN) uses while learning to recognize a regular formal language. Specifically, we train a RNN on positive and negative examples from a regular language, and ask if there is a simple decoding function that maps states of this RNN to states of the minimal deterministic finite automaton (MDFA) for the language. Our experiments show that such a decoding function indeed exists, and that it maps states of the RNN not to MDFA states, but to states of an {\em abstraction} obtained by clustering small sets of MDFA states into …
Attractor-merging Crises and Intermittency in Reservoir Computing
Graduate School of Information Science and Technology, University of Tokyo, Bunkyo-ku, Tokyo 113-0033, Japan（信息科学与技术研究生学校，东京大学，东京都，东京113-0033，日本）；International Research Center for Neurointelligence, University of Tokyo Institutes for Advanced Study, University of Tokyo, Bunkyo-ku, Tokyo 113-0033, Japan（神经智能国际研究中心，东京大学先进研究所，东京大学，东京都，东京113-0033，日本）
We present a novel algorithm that uses exact learning and abstraction to extract a deterministic finite automaton describing the state dynamics of a given trained RNN. We do this using Angluin's L* algorithm as a learner and the trained RNN as an oracle. Our technique efficiently extracts accurate automata from trained RNNs, even when the state vectors are large and require fine differentiation.
Penn State University（宾夕法尼亚州立大学）；Rochester Institute of Technology（罗切斯特理工学院）
In order to learn complex grammars, recurrent neural networks (RNNs) require sufficient computational resources to ensure correct grammar recognition. A widely-used approach to expand model capacity would be to couple an RNN to an external memory stack. Here, we introduce a "neural state" pushdown automaton (NSPDA), which consists of a digital stack, instead of an analog one, that is coupled to a neural network state machine. We empirically show its effectiveness in recognizing various context-free grammars (CFGs). First, we develop the underlying mechanics of the proposed higher order recurre…
W. M. Keck Science Department, Claremont McKenna, Scripps, and Pitzer College（克莱蒙特·麦肯纳学院、斯克里普斯学院和皮策学院 W. M. 凯克科学系）；University of California, Davis（加州大学戴维斯分校）
Reservoir computers (RCs) and recurrent neural networks (RNNs) can mimic any finite-state automaton in theory, and some workers demonstrated that this can hold in practice. We test the capability of generalized linear models, RCs, and Long Short-Term Memory (LSTM) RNN architectures to predict the stochastic processes generated by a large suite of probabilistic deterministic finite-state automata (PDFA). PDFAs provide an excellent performance benchmark in that they can be systematically enumerated, the randomness and correlation structure of their generated processes are exactly known, and thei…
University of Washington（华盛顿大学）；Allen Institute for Artificial Intelligence（艾伦人工智能研究所）；Carnegie Mellon University（卡内基梅隆大学）
Recurrent and convolutional neural networks comprise two distinct families of models that have proven to be useful for encoding natural language utterances. In this paper we present SoPa, a new model that aims to bridge these two approaches. SoPa combines neural representation learning with weighted finite-state automata (WFSAs) to learn a soft version of traditional surface patterns. We show that SoPa is an extension of a one-layer CNN, and that such CNNs are equivalent to a restricted version of SoPa, and accordingly, to a restricted form of WFSA. Empirically, on three text classification ta…
Technion（以色列理工学院）；Bar Ilan University（巴伊兰大学）；Allen Institute for AI（艾伦人工智能研究所）
We present an algorithm for extraction of a probabilistic deterministic finite automaton (PDFA) from a given black-box language model, such as a recurrent neural network (RNN). The algorithm is a variant of the exact-learning algorithm L*, adapted to a probabilistic setting with noise. The key insight is the use of conditional probabilities for observations, and the introduction of a local tolerance when comparing them. When applied to RNNs, our algorithm often achieves better word error rate (WER) and normalised distributed cumulative gain (NDCG) than that achieved by spectral extraction of w…
We present an algorithm for extracting a subclass of the context free grammars (CFGs) from a trained recurrent neural network (RNN). We develop a new framework, pattern rule sets (PRSs), which describe sequences of deterministic finite automata (DFAs) that approximate a non-regular language. We present an algorithm for recovering the PRS behind a sequence of such automata, and apply it to the sequences of automata extracted from trained RNNs using the L* algorithm. We then show how the PRS may converted into a CFG, enabling a familiar and useful presentation of the learned language. Extracting…
Carnegie Mellon University（卡内基梅隆大学）；Allen Institute for Artificial Intelligence（艾伦人工智能研究所）；University of Washington（华盛顿大学）
Neural models for NLP typically use large numbers of parameters to reach state-of-the-art performance, which can lead to excessive memory usage and increased runtime. We present a structure learning method for learning sparse, parameter-efficient NLP models. Our method applies group lasso to rational RNNs (Peng et al., 2018), a family of models that is closely connected to weighted finite-state automata (WFSAs). We take advantage of rational RNNs' natural grouping of the weights, so the group lasso penalty directly removes WFSA states, substantially reducing the number of parameters in the mod…
Department of Computer Science and Technology, University of Cambridge（剑桥大学计算机科学与技术系）；Department of Mechanical and Process Engineering, ETH Zurich（苏黎世联邦理工学院机械与过程工程系）
