Dataset Information

Comparing feedforward and recurrent neural network architectures with human behavior in artificial grammar learning

ABSTRACT: In recent years artificial neural networks achieved performance close to or better than humans in several domains: tasks that were previously human prerogatives, such as language processing, have witnessed remarkable improvements in state of the art models. One advantage of this technological boost is to facilitate comparison between different neural networks and human performance, in order to deepen our understanding of human cognition. Here, we investigate which neural network architecture (feedforward vs. recurrent) matches human behavior in artificial grammar learning, a crucial aspect of language acquisition. Prior experimental studies proved that artificial grammars can be learnt by human subjects after little exposure and often without explicit knowledge of the underlying rules. We tested four grammars with different complexity levels both in humans and in feedforward and recurrent networks. Our results show that both architectures can “learn” (via error back-propagation) the grammars after the same number of training sequences as humans do, but recurrent networks perform closer to humans than feedforward ones, irrespective of the grammar complexity level. Moreover, similar to visual processing, in which feedforward and recurrent architectures have been related to unconscious and conscious processes, the difference in performance between architectures over ten regular grammars shows that simpler and more explicit grammars are better learnt by recurrent architectures, supporting the hypothesis that explicit learning is best modeled by recurrent networks, whereas feedforward networks supposedly capture the dynamics involved in implicit learning.

SUBMITTER: Alamia A

PROVIDER: S-EPMC7747619 | biostudies-literature | 2020 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundDeep learning has made tremendous successes in numerous artificial intelligence applications and is unsurprisingly penetrating into various biomedical domains. High-throughput omics data in the form of molecular profile matrices, such as transcriptomes and metabolomes, have long existed as a valuable resource for facilitating diagnosis of patient statuses/stages. It is timely imperative to compare deep learning neural networks against classical machine learning methods in the setting of matrix-formed omics data in terms of classification accuracy and robustness.ResultsUsing 37 high throughput omics datasets, covering transcriptomes and metabolomes, we evaluated the classification power of deep learning compared to traditional machine learning methods. Representative deep learning methods, Multi-Layer Perceptrons (MLP) and Convolutional Neural Networks (CNN), were deployed and explored in seeking optimal architectures for the best classification performance. Together with five classical supervised classification methods (Linear Discriminant Analysis, Multinomial Logistic Regression, Naïve Bayes, Random Forest, Support Vector Machine), MLP and CNN were comparatively tested on the 37 datasets to predict disease stages or to discriminate diseased samples from normal samples. MLPs achieved the highest overall accuracy among all methods tested. More thorough analyses revealed that single hidden layer MLPs with ample hidden units outperformed deeper MLPs. Furthermore, MLP was one of the most robust methods against imbalanced class composition and inaccurate class labels.ConclusionOur results concluded that shallow MLPs (of one or two hidden layers) with ample hidden neurons are sufficient to achieve superior and robust classification performance in exploiting numerical matrix-formed omics data for diagnosis purpose. Specific observations regarding optimal network width, class imbalance tolerance, and inaccurate labeling tolerance will inform future improvement of neural network applications on functional genomics data.

Project description:Patients with non-fluent aphasias display impairments of expressive and receptive grammar. This has been attributed to deficits in processing configurational and hierarchical sequencing relationships. This hypothesis had not been formally tested. It was also controversial whether impairments are specific to language, or reflect domain general deficits in processing structured auditory sequences. Here we used an artificial grammar learning paradigm to compare the abilities of controls to participants with agrammatic aphasia of two different aetiologies: stroke and frontotemporal dementia. Ten patients with non-fluent variant primary progressive aphasia (nfvPPA), 12 with non-fluent aphasia due to stroke, and 11 controls implicitly learned a novel mixed-complexity artificial grammar designed to assess processing of increasingly complex sequencing relationships. We compared response profiles for otherwise identical sequences of speech tokens (nonsense words) and tone sweeps. In all three groups the ability to detect grammatical violations varied with sequence complexity, with performance improving over time and being better for adjacent than non-adjacent relationships. Patients performed less well than controls overall, and this was related more strongly to aphasia severity than to aetiology. All groups improved with practice and performed well at a control task of detecting oddball nonwords. Crucially, group differences did not interact with sequence complexity, demonstrating that aphasic patients were not disproportionately impaired on complex structures. Hierarchical cluster analysis revealed that response patterns were very similar across all three groups, but very different between the nonsense word and tone tasks, despite identical artificial grammar structures. Overall, we demonstrate that agrammatic aphasics of two different aetiologies are not disproportionately impaired on complex sequencing relationships, and that the learning of phonological and non-linguistic sequences occurs independently. The similarity of profiles of discriminatory abilities and rule learning across groups suggests that insights from previous studies of implicit sequence learning in vascular aphasia are likely to prove applicable in nfvPPA.

Dataset Information

Comparing feedforward and recurrent neural network architectures with human behavior in artificial grammar learning

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets