Brain Activity and Functional Connectivity Patterns Associated With Fast and Slow Motor Sequence Learning in Late Middle Adulthood

The human brain undergoes structural and functional changes across the lifespan. The study of motor sequence learning in elderly subjects is of particularly interest since previous findings in young adults might not replicate during later stages of adulthood. The present functional magnetic resonance imaging (fMRI) study assessed the performance, brain activity and functional connectivity patterns associated with motor sequence learning in late middle adulthood. For this purpose, a total of 25 subjects were evaluated during early stages of learning [i.e., fast learning (FL)]. A subset of these subjects (n = 11) was evaluated after extensive practice of a motor sequence [i.e., slow learning (SL) phase]. As expected, late middle adults improved motor performance from FL to SL. Learning-related brain activity patterns replicated most of the findings reported previously in young subjects except for the lack of hippocampal activity during FL and the involvement of cerebellum during SL. Regarding functional connectivity, precuneus and sensorimotor lobule VI of the cerebellum showed a central role during improvement of novel motor performance. In the sample of subjects evaluated, connectivity between the posterior putamen and parietal and frontal regions was significantly decreased with aging during SL. This age-related connectivity pattern may reflect losses in network efficiency when approaching late adulthood. Altogether, these results may have important applications, for instance, in motor rehabilitation programs.

Disease Progression Detection via Deep Sequence Learning of Successive Radiographic Scans

The highly rapid spread of the current pandemic has quickly overwhelmed hospitals all over the world and motivated extensive research to address a wide range of emerging problems. The unforeseen influx of COVID-19 patients to hospitals has made it inevitable to deploy a rapid and accurate triage system, monitor progression, and predict patients at higher risk of deterioration in order to make informed decisions regarding hospital resource management. Disease detection in radiographic scans, severity estimation, and progression and prognosis prediction have been extensively studied with the help of end-to-end methods based on deep learning. The majority of recent works have utilized a single scan to determine severity or predict progression of the disease. In this paper, we present a method based on deep sequence learning to predict improvement or deterioration in successive chest X-ray scans and build a mathematical model to determine individual patient disease progression profile using successive scans. A deep convolutional neural network pretrained on a diverse lung disease dataset was used as a feature extractor to generate the sequences. We devised three strategies for sequence modeling in order to obtain both fine-grained and coarse-grained features and construct sequences of different lengths. We also devised a strategy to quantify positive or negative change in successive scans, which was then combined with age-related risk factors to construct disease progression profile for COVID-19 patients. The age-related risk factors allowed us to model rapid deterioration and slower recovery in older patients. Experiments conducted on two large datasets showed that the proposed method could accurately predict disease progression. With the best feature extractor, the proposed method was able to achieve AUC of 0.98 with the features obtained from radiographs. Furthermore, the proposed patient profiling method accurately estimated the health profile of patients.

Expectation violations produce error signals in mouse V1

Repeated exposure to visual sequences changes the form of evoked activity in the primary visual cortex (V1). Predictive coding theory provides a potential explanation for this, namely that plasticity shapes cortical circuits to encode spatiotemporal predictions and that subsequent responses are modulated by the degree to which actual inputs match these expectations. Here we use a recently developed statistical modeling technique called Model-Based Targeted Dimensionality Reduction (MbTDR) to study visually-evoked dynamics in mouse V1 in context of a previously described experimental paradigm called "sequence learning". We report that evoked spiking activity changed significantly with training, in a manner generally consistent with the predictive coding framework. Neural responses to expected stimuli were suppressed in a late window (100-150ms) after stimulus onset following training, while responses to novel stimuli were not. Omitting predictable stimuli led to increased firing at the expected time of stimulus onset, but only in trained mice. Substituting a novel stimulus for a familiar one led to changes in firing that persisted for at least 300ms. In addition, we show that spiking data can be used to accurately decode time within the sequence. Our findings are consistent with the idea that plasticity in early visual circuits is involved in coding spatiotemporal information.

POLLA: Enhancing the Local Structure Awareness in Long Sequence Spatial-temporal Modeling

The spatial-temporal modeling on long sequences is of great importance in many real-world applications. Recent studies have shown the potential of applying the self-attention mechanism to improve capturing the complex spatial-temporal dependencies. However, the lack of underlying structure information weakens its general performance on long sequence spatial-temporal problem. To overcome this limitation, we proposed a novel method, named the Proximity-aware Long Sequence Learning framework, and apply it to the spatial-temporal forecasting task. The model substitutes the canonical self-attention by leveraging the proximity-aware attention, which enhances local structure clues in building long-range dependencies with a linear approximation of attention scores. The relief adjacency matrix technique can utilize the historical global graph information for consistent proximity learning. Meanwhile, the reduced decoder allows for fast inference in a non-autoregressive manner. Extensive experiments are conducted on five large-scale datasets, which demonstrate that our method achieves state-of-the-art performance and validates the effectiveness brought by local structure information.

Sequence-in-Sequence Learning for SOH Estimation of Lithium-Ion Battery

State-of-Health (SOH) prediction of a Lithium-ion battery is essential for preventing malfunction and maintaining efficient working behaviors for the battery. In practice, this task is difficult due to the high level of noise and complexity. There are many machine learning methods, especially deep learning approaches, that have been proposed to address this problem recently. However, there is much room for improvement because the nature of the battery data is highly non-linear and exhibits higher dependence on multidisciplinary parameters such as resistance, voltage and external conditions the battery is subjected to. In this paper, we propose an approach known as bidirectional sequence-in-sequence, which exploits the dependency of nested cycle-wise and channel-wise battery data. Experimented with real dataset acquired from NASA, our method results in significant reduction of error of approximately up to 32.5%.

Mental Number Representations are Spatially Mapped both by Their Magnitudes and Ordinal Positions

One of the most fundamental effects used to investigate number representations is the Spatial-Numerical Association of Response Codes (SNARC) effect showing that responses to small/large numbers are faster with the left/right hand, respectively. However, in recent years, it is hotly debated whether the SNARC effect is based upon cardinal representation of number magnitude or ordinal representation of number sequence in working memory. However, one problem is that evidence comes from different paradigms, e.g., evidence for ordinal sequences comes usually from experiments, where ordinal sequences have to be learnt and it has been ar-gued that this secondary task triggers the effect. Therefore, in this preregistered study we em-ployed a SNARC task, without secondary ordinal sequence learning, in which we can dissociate ordinal and magnitude accounts by careful manipulation of experimental stimulus sets and com-pare magnitude and ordinal models. The results indicate that even though the observed data is better accounted for by the magnitude model, the ordinal position seems to matter as well. Thus, it appears that the mechanisms described by both accounts play a significant role when mental numbers are temporarily mapped onto space even when no ordinal learning is involved.