Searcharxiv⌕ Search

arXiv subjects

Anirudh Natarajan

Publications and source records attributed to Anirudh Natarajan.

4 recordsLinked to original sources

Mind-wandering preserves word-level coupling while disrupting comprehension

Readers can move their eyes through a page but retain little of its meaning when their minds wander. We tested whether word properties still guide their eyes during these lapses. We analyzed public eye-tracking and electroencephalography (EEG) data from 44 adults who read articles. They marked the text read during each lapse they noticed. We linked these spans to the sentences needed for later comprehension questions. Word frequency, length and surprisal still predicted which words readers skipped and how long they looked at them. Compared with on-task reading, these effects met equivalence bounds of $\pm$10% for skipping and $\pm$20% for gaze duration. Readers spent 14.6% longer on words during lapses. Yet readers answered less accurately when lapses covered sentences needed for a later question. The adjusted difference was 25.7 percentage points for lapses covering all versus none of these sentences. The fixation-related EEG response was 42.7% smaller during lapses. At left posterior-lateral sensor P7, the median increase in alpha power was 10.0%. Occipital sensor Oz showed no reliable increase. These results suggest that word properties can guide the eyes while readers fail to form or retain a useful representation of the text.

q-bio.NC↗

Multimodal Physiological Decoding Reveals Individualized Arousal Dynamics in Closed-Loop Neurofeedback

Most arousal-based brain-computer interfaces (BCIs) use EEG to decode cognitive state. But arousal is an autonomic process. We examined if peripheral physiological signals give a better decode of task-related arousal. We used a public dataset from a difficult boundary-avoidance flight task. In this task, participants received EEG-based BCI neurofeedback, sham feedback, or no feedback. We trained a multimodal deep learning decoder on heart rate, heart rate variability (HRV), respiration, electrodermal activity, and pupil diameter. We compared it with an EEG-only decoder and with the original filter-bank common spatial pattern (FBCSP) decoder. All analyses were offline. The peripheral decoder had a within-subject AUC of 93.0%. The EEG-only decoder had 85.2% and the FBCSP decoder had 79.8%. Only the peripheral decoder showed the Yerkes-Dodson inverted-U relation between decoded arousal and performance. The feedback conditions did not change the arousal trajectories. Thus, the BCI benefit possibly comes from regulation at critical moments in the task, not from a continuous shift. We calculated a time-varying optimal arousal trajectory from control trials. Per-trial deviations from this trajectory predicted performance (r = -0.28 to -0.24, p < 0.01). Baseline HRV and gamma power changed the arousal-performance curve of each subject. We used these values to divide subjects into arousal-sensitive and arousal-tolerant groups. Control bands specific to each group increased trial separation (Cohen's d from 1.21 to 1.32 and from 0.85 to 1.10). These results show that peripheral signals are better than EEG for continuous arousal decoding. They also give a specification for individualized closed-loop arousal regulation. A prospective study must test this specification, for example with personalized transcutaneous vagus nerve stimulation.

q-bio.NC↗

Efficient emotion recognition using hyperdimensional computing with combinatorial channel encoding and cellular automata

In this paper, a hardware-optimized approach to emotion recognition based on the efficient brain-inspired hyperdimensional computing (HDC) paradigm is proposed. Emotion recognition provides valuable information for human-computer interactions, however the large number of input channels (>200) and modalities (>3) involved in emotion recognition are significantly expensive from a memory perspective. To address this, methods for memory reduction and optimization are proposed, including a novel approach that takes advantage of the combinatorial nature of the encoding process, and an elementary cellular automaton. HDC with early sensor fusion is implemented alongside the proposed techniques achieving two-class multi-modal classification accuracies of >76% for valence and >73% for arousal on the multi-modal AMIGOS and DEAP datasets, almost always better than state of the art. The required vector storage is seamlessly reduced by 98% and the frequency of vector requests by at least 1/5. The results demonstrate the potential of efficient hyperdimensional computing for low-power, multi-channeled emotion recognition tasks.

cs.ET↗

Scale bridging materials physics: Active learning workflows and integrable deep neural networks for free energy function representations in alloys

The free energy plays a fundamental role in descriptions of many systems in continuum physics. Notably, in multiphysics applications, it encodes thermodynamic coupling between different fields. It thereby gives rise to driving forces on the dynamics of interaction between the constituent phenomena. In mechano-chemically interacting materials systems, even consideration of only compositions, order parameters and strains can render the free energy to be reasonably high-dimensional. In proposing the free energy as a paradigm for scale bridging, we have previously exploited neural networks for their representation of such high-dimensional functions. Specifically, we have developed an integrable deep neural network (IDNN) that can be trained to free energy derivative data obtained from atomic scale models and statistical mechanics, then analytically integrated to recover a free energy density function. The motivation comes from the statistical mechanics formalism, in which certain free energy derivatives are accessible for control of the system, rather than the free energy itself. Our current work combines the IDNN with an active learning workflow to improve sampling of the free energy derivative data in a high-dimensional input space. Treated as input-output maps, machine learning accommodates role reversals between independent and dependent quantities as the mathematical descriptions change with scale bridging. As a prototypical system we focus on Ni-Al. Phase field simulations using the resulting IDNN representation for the free energy density of Ni-Al demonstrate that the appropriate physics of the material have been learned. To the best of our knowledge, this represents the most complete treatment of scale bridging, using the free energy for a practical materials system, that starts with electronic structure calculations and proceeds through statistical mechanics to continuum physics.

cs.LG↗