Mental workload is known to alter cardiovascular function leading to increased cardiovascular risk. Nevertheless, there is no clear autonomic nervous system unbalance to be quantified during mental stress. We aimed to characterize the mental workload impact on the cardiovascular function with a focus on heart rate variability (HRV) non-linear indexes. A 1-h computerized switching task (letter recognition) was performed by 24 subjects while monitoring their performance (accuracy, response time), electrocardiogram and blood pressure waveform (finger volume clamp method). The HRV was evaluated from the beat-to-beat RR intervals (RRI) in time-, frequency-, and informational- domains, before (Control) and during the task. The task induced a significant mental workload (visual analog scale of fatigue from 27 ± 26 to 50 ± 31 mm, p < 0.001, and NASA-TLX score of 56 ± 17). The heart rate, blood pressure and baroreflex function were unchanged, whereas most of the HRV parameters markedly decreased. The maximum decrease occurred during the first 15 min of the task (P1), before starting to return to the baseline values reached at the end of the task (P4). The RRI dimension correlation (D2) decrease was the most significant (P1 vs. Control: 1.42 ± 0.85 vs. 2.21 ± 0.8, p < 0.001) and only D2 lasted until the task ended (P4 vs. Control: 1.96 ± 0.9 vs. 2.21 ± 0.9, p < 0.05). D2 was identified as the most robust cardiovascular variable impacted by the mental workload as determined by posterior predictive simulations ( p = 0.9). The Spearman correlation matrix highlighted that D2 could be a marker of the generated frustration ( R = –0.61, p < 0.01) induced by a mental task, as well as the myocardial oxygen consumption changes assessed by the double product ( R = –0.53, p < 0.05). In conclusion, we showed that mental workload sharply lowered the non-linear RRI dynamics, particularly the RRI correlation dimension.

No abstract

While neural networks (NN) have been successfully applied to many NLP tasks, the way they function is often difficult to interpret. In this article, we focus on binary text classification via NNs and propose a new tool, which includes a visualization of the decision boundary and the distances of data elements to this boundary. This tool increases the interpretability of NN. Our approach uses two innovative views: (1) an overview of the text representation space and (2) a local view allowing data exploration around the decision boundary for various localities of this representation space. These views are integrated into a visual platform, EBBE-Text, which also contains state-of-the-art visualizations of NN representation spaces and several kinds of information obtained from the classification process. The various views are linked through numerous interactive functionalities that enable easy exploration of texts and classification results via the various complementary views. A user study shows the effectiveness of the visual encoding and a case study illustrates the benefits of using our tool for the analysis of the classifications obtained with several recent NNs and two datasets.