Convolutional neural networks can decode eye movement data: A black box approach to predicting task from eye movements.

Previous attempts to classify task from eye movement data have relied on model architectures designed to emulate theoretically defined cognitive processes and/or data that have been processed into aggregate (e.g., fixations, saccades) or statistical (e.g., fixation density) features. Black box convolutional neural networks (CNNs) are capable of identifying relevant features in raw and minimally processed data and images, but difficulty interpreting these model architectures has contributed to challenges in generalizing lab-trained CNNs to applied contexts. In the current study, a CNN classifier was used to classify task from two eye movement datasets (Exploratory and Confirmatory) in which participants searched, memorized, or rated indoor and outdoor scene images. The Exploratory dataset was used to tune the hyperparameters of the model, and the resulting model architecture was retrained, validated, and tested on the Confirmatory dataset. The data were formatted into timelines (i.e., x-coordinate, y-coordinate, pupil size) and minimally processed images. To further understand the informational value of each component of the eye movement data, the timeline and image datasets were broken down into subsets with one or more components systematically removed. Classification of the timeline data consistently outperformed the image data. The Memorize condition was most often confused with Search and Rate. Pupil size was the least uniquely informative component when compared with the x- and y-coordinates. The general pattern of results for the Exploratory dataset was replicated in the Confirmatory dataset. Overall, the present study provides a practical and reliable black box solution to classifying task from eye movement data.

A practical comparison of algorithms for the measurement of multiscale entropy in neural time series data.

There is a broad family of statistical methods for capturing time series regularity, with increasingly widespread adoption by the neuroscientific community. A common feature of these methods is that they permit investigators to quantify the entropy of brain signals - an index of unpredictability/complexity. Despite the proliferation of algorithms for computing entropy from neural time series data there is scant evidence concerning their relative stability and efficiency. Here we evaluated several different algorithmic implementations (sample, fuzzy, dispersion and permutation) of multiscale entropy in terms of their stability across sessions, internal consistency and computational speed, accuracy and precision using a combination of electroencephalogram (EEG) and synthetic 1/ƒ noise signals. Overall, we report fair to excellent internal consistency and longitudinal stability over a one-week period for the majority of entropy estimates, with several caveats. Computational timing estimates suggest distinct advantages for dispersion and permutation entropy over other entropy estimates. Considered alongside the psychometric evidence, we suggest several ways in which researchers can maximize computational resources (without sacrificing reliability), especially when working with high-density M/EEG data or multivoxel BOLD time series signals.

Entrainment of visual steady-state responses is modulated by global spatial statistics.

The rhythmic delivery of visual stimuli evokes large-scale neuronal entrainment in the form of steady-state oscillatory field potentials. The spatiotemporal properties of stimulus drive appear to constrain the relative degrees of neuronal entrainment. Specific frequency ranges, for example, are uniquely suited for enhancing the strength of stimulus-driven brain oscillations. When it comes to the nature of the visual stimulus itself, studies have used a plethora of inputs ranging from spatially unstructured empty fields to simple contrast patterns (checkerboards, gratings, stripes) and complex arrays (human faces, houses, natural scenes). At present, little is known about how the global spatial statistics of the input stimulus influence entrainment of scalp-recorded electrophysiological signals. In this study, we used rhythmic entrainment source separation of scalp EEG to compare stimulus-driven phase alignment for distinct classes of visual inputs, including broadband spatial noise ensembles with varying second-order statistics, natural scenes, and narrowband sine-wave gratings delivered at a constant flicker frequency. The relative magnitude of visual entrainment was modulated by the global properties of the driving stimulus. Entrainment was strongest for pseudo-naturalistic broadband visual noise patterns in which luminance contrast is greatest at low spatial frequencies (a power spectrum slope characterized by 1/ƒ-2).NEW & NOTEWORTHY Rhythmically modulated visual stimuli entrain the activity of neuronal populations, but the effect of global stimulus statistics on this entrainment is unknown. We assessed entrainment evoked by 1) visual noise ensembles with different spectral slopes, 2) complex natural scenes, and 3) narrowband sinusoidal gratings. Entrainment was most effective for broadband noise with naturalistic luminance contrast. This reveals some global properties shaping stimulus-driven brain oscillations in the human visual system.

Representation of affect in sensory cortex.

Contemporary neuroscience suggests that perception is perhaps best understood as a dynamically iterative process that does not honor cleanly segregated "bottom-up" or "top-down" streams. We argue that there is substantial empirical support for the idea that affective influences infiltrate the earliest reaches of sensory processing and even that primitive internal affective dimensions (e.g., goodness-to-badness) are represented alongside physical dimensions of the external world.

Deep-Learning-Based Multivariate Pattern Analysis (dMVPA): A Tutorial and a Toolbox.

In recent years, multivariate pattern analysis (MVPA) has been hugely beneficial for cognitive neuroscience by making new experiment designs possible and by increasing the inferential power of functional magnetic resonance imaging (fMRI), electroencephalography (EEG), and other neuroimaging methodologies. In a similar time frame, "deep learning" (a term for the use of artificial neural networks with convolutional, recurrent, or similarly sophisticated architectures) has produced a parallel revolution in the field of machine learning and has been employed across a wide variety of applications. Traditional MVPA also uses a form of machine learning, but most commonly with much simpler techniques based on linear calculations; a number of studies have applied deep learning techniques to neuroimaging data, but we believe that those have barely scratched the surface of the potential deep learning holds for the field. In this paper, we provide a brief introduction to deep learning for those new to the technique, explore the logistical pros and cons of using deep learning to analyze neuroimaging data - which we term "deep MVPA," or dMVPA - and introduce a new software toolbox (the "Deep Learning In Neuroimaging: Exploration, Analysis, Tools, and Education" package, DeLINEATE for short) intended to facilitate dMVPA for neuroscientists (and indeed, scientists more broadly) everywhere.

Parallel effects of retrieval ease on attributions about the past and the future.

The present study explored the role of task difficulty in judgments about the past and the future. Participants recalled events from childhood and imagined future events. The difficulty of the task was manipulated by asking participants to generate either four or twelve events. Participants then rated how well they could generally remember events from their childhood or how well planned their futures were. Consistent with past research (e.g., Winkielman, Schwarz, & Belli, 1998), participants in the difficult recall group rated their childhood memories as less complete than participants in the easy recall group. A parallel effect was found in participants' judgments of their futures. Participants who were asked to imagine twelve future events rated their future plans as less complete than those who imagined four events. Moreover, there was a negative correlation between the rated difficulty of the task and the degree to which participants found their memories and plans to be complete. We also examined the valence of the generated events. These results showed a strong positivity bias for both types of judgments, and the bias was particularly strong when thinking of future events. The results suggest that similar attributional processes mediate beliefs about the past and the future.

Reliability of graph metrics derived from resting-state human EEG.

It is increasingly appreciated that a complete description of brain functioning will necessarily involve the characterization of large-scale interregional temporal synchronization of neuronal assemblies. The need to capture the dynamic formation of such large-scale networks has yielded a renewed interest in the human EEG in combination with a suite of methods for estimating functional connectivity along with the graph theoretical approaches for characterizing network structure. While initial work has established generally good reproducibility for a limited selection of these graph theoretical measures, there remains an obvious need to document the reproducibility of a more extensive array of commonly used graph metrics. We sought to evaluate the test-retest reliability of a much richer suite of graph theoretic measures as applied to weighted networks derived from high-density resting-state human EEG. Our findings were promising overall, with some important qualifications when considering the frequency bands of interest and the method used to calculate functional connectivity as well as some substantial variance between individual graph metrics. In general, the reliability of networks in the α and ß frequency bands was improved when functional connectivity was defined solely on the basis of relative phase distributions. In the δ and θ bands, reliability was substantially better when functional connectivity was based on coherence, which incorporates both phase and amplitude information.

Charting moment-to-moment brain signal variability from early to late childhood.

Large-scale brain signals exhibit rich intermittent patterning, reflecting the fact that the cortex actively eschews fixed points in favor of itinerant wandering with frequent state transitions. Fluctuations in endogenous cortical activity occur at multiple time scales and index a dynamic repertoire of network states that are continuously explored, even in the absence of external sensory inputs. Here, we quantified such moment-to-moment brain signal variability at rest in a large, cross-sectional sample of children ranging in age from seven to eleven years. Our findings revealed a monotonic rise in the complexity of electroencephalogram (EEG) signals as measured by sample entropy, from the youngest to the oldest age cohort, across a range of time scales and spatial regions. From year to year, the greatest changes in intraindividual brain signal variability were recorded at electrodes covering the anterior cortical zones. These results provide converging evidence concerning the age-dependent expansion of functional cortical network states during a critical developmental period ranging from early to late childhood.