Logic-based machine learning has the crucial advantage of transparency. However, despite significant recent progress, further research is needed to close the accuracy gap between logic-based architectures and deep neural network ones. This paper introduces a novel variant of the Tsetlin machine (TM) that randomly drops clauses, the logical learning element of TMs. In effect, TM with Drop Clause ignores a random selection of the clauses in each epoch, selected according to a predefined probability. In this way, the TM learning phase becomes more diverse. To explore the effects that Drop Clause has on accuracy, training time and robustness, we conduct extensive experiments on nine benchmark datasets in natural language processing (IMDb, R8, R52, MR, and TREC) and image classification (MNIST, Fashion MNIST, CIFAR-10, and CIFAR-100). Our proposed model outperforms baseline machine learning algorithms by a wide margin and achieves competitive performance compared with recent deep learning models, such as BERT-Large and AlexNet-DFA. In brief, we observe up to +10% increase in accuracy and 2x to 4x faster learning than for the standard TM. We visualize the patterns learnt by Drop Clause TM in the form of heatmaps and show evidence of the ability of drop clause to learn more unique and discriminative patterns. We finally evaluate how Drop Clause affects learning robustness by introducing corruptions and alterations in the image/language test data, which exposes increased learning robustness.

The state-of-the-art natural language processing models have raised the bar for excellent performance on a variety of tasks in recent years. However, concerns are rising over their primitive sensitivity to distribution biases that reside in the training and testing data. This issue hugely impacts the performance of the models when exposed to out-of-distribution and counterfactual data. The root cause seems to be that many machine learning models are prone to learn the shortcuts, modelling simple correlations rather than more fundamental and general relationships. As a result, such text classifiers tend to perform poorly when a human makes minor modifications to the data, which raises questions regarding their robustness. In this paper, we employ a rule-based architecture called Tsetlin Machine (TM) that learns both simple and complex correlations by ANDing features and their negations. As such, it generates explainable AND-rules using negated and non-negated reasoning. Here, we explore how non-negated reasoning can be more prone to distribution biases than negated reasoning. We further leverage this finding by adapting the TM architecture to mainly perform negated reasoning using the specificity parameter s. As a result, the AND-rules becomes robust to spurious correlations and can also correctly predict counterfactual data. Our empirical investigation of the model's robustness uses the specificity s to control the degree of negated reasoning. Experiments on publicly available Counterfactually-Augmented Data demonstrate that the negated clauses are robust to spurious correlations and outperform Naive Bayes, SVM, and Bi-LSTM by up to 20 %, and ELMo by almost 6 % on counterfactual test data.

This paper proposes human-interpretable learning of aspect-based sentiment analysis (ABSA), employing the recently introduced Tsetlin Machines (TMs). We attain interpretability by converting the intricate position-dependent textual semantics into binary form, mapping all the features into bag-of-words (BOWs). The binary form BOWs are encoded so that the information on the aspect and context words are nearly lossless for sentiment classification. We further adapt the BOWs as input to the TM, enabling learning of aspect-based sentiment patterns in propositional logic. To evaluate interpretability and accuracy, we conducted experiments on two widely used ABSA datasets of SemEval 2014: Restaurant 14 and Laptop 14. The experiments show how each relevant feature takes part in conjunctive clauses that contain the context information for the corresponding aspect word, demonstrating human-level interpretability. At the same time, the obtained accuracy is competitive with existing neural network models, reaching 78.02% on Restaurant 14 and 73.51% on Laptop 14.

Using logical clauses to represent patterns, Tsetlin Machine (TM) have recently obtained competitive performance in terms of accuracy, memory footprint, energy, and learning speed on several benchmarks. Each TM clause votes for or against a particular class, with classification resolved using a majority vote. While the evaluation of clauses is fast, being based on binary operators, the voting makes it necessary to synchronize the clause evaluation, impeding parallelization. In this paper, we propose a novel scheme for desynchronizing the evaluation of clauses, eliminating the voting bottleneck. In brief, every clause runs in its own thread for massive native parallelism. For each training example, we keep track of the class votes obtained from the clauses in local voting tallies. The local voting tallies allow us to detach the processing of each clause from the rest of the clauses, supporting decentralized learning. This means that the TM most of the time will operate on outdated voting tallies. We evaluated the proposed parallelization across diverse learning tasks and it turns out that our decentralized TM learning algorithm copes well with working on outdated data, resulting in no significant loss in learning accuracy. Furthermore, we show that the approach provides up to 50 times faster learning. Finally, learning time is almost constant for reasonable clause amounts (employing from 20 to 7,000 clauses on a Tesla V100 GPU). For sufficiently large clause numbers, computation time increases approximately proportionally. Our parallel and asynchronous architecture thus allows processing of more massive datasets and operating with more clauses for higher accuracy.

Aspect-based sentiment analysis (ABSA) aims at identifying fine-grained polarity of opinion associated with a given aspect word. Several existing articles demonstrated promising ABSA accuracy using positional embedding to show the relationship between an aspect word and its context. In most cases, the positional embedding depends on the distance between the aspect word and the remaining words in the context, known as the position index sequence. However, these techniques usually employ both complex preprocessing approaches with additional trainable positional embedding and complex architectures to obtain the state-of-the-art performance. In this paper, we simplify preprocessing by including polarity lexicon replacement and masking techniques that carry the information of the aspect word’s position and eliminate the positional embedding. We then adopt a novel and concise architecture using two Bidirectional GRU along with an attention layer to classify the aspect based on its context words. Experiment results show that the simplified preprocessing and the concise architecture significantly improve the accuracy of the publicly available ABSA datasets, obtaining 81.37%, 75.39%, 80.88%, and 89.30% in restaurant 14, laptop 14, restaurant 15, and restaurant 16 respectively.