Anxiety-Related Difficulties With Complex Arithmetic

Human behavior depends on the interplay between cognition and emotion. Negative emotions like anxiety affect performance, particularly in complex tasks, by limiting cognitive resources – known as the anxiety–complexity effect. This study set out to replicate the anxiety–complexity effect in a web-based experiment. We investigated individual differences in math anxiety – a negative emotional response specific to math – and arithmetic performance ( N = 382). The mental arithmetic task consisted of a two-digit addition and subtraction, with/without carrying or borrowing, respectively. As expected and preregistered, higher math anxiety was related to poorer arithmetic performance, especially in complex tasks – indicating the anxiety–complexity effect. Consequently, the negative math anxiety-performance link is especially pronounced for complex arithmetic, which requires calculations across place-values and thus working memory resources. This successful replication of the anxiety–complexity effect suggests that math-anxious individuals have particular difficulties in complex arithmetic.

EXPRESS: Effects of Presentation Modality and Duration on Children’s Strategy Use: A Study in Computational Estimation

Two experiments were run to determine how presentation modality and duration influence children’s arithmetic performance and strategy selection. Third and fourth graders were asked to find estimates for two-digit addition problems (e.g., 52 + 39). Children were tested in three conditions: (1) time-unlimited visual, (2) time-limited visual, or (3) time-limited auditory conditions. Moreover, we assessed children’s working-memory updating and arithmetic fluency. Children were told which strategy to use on each problem to assess arithmetic performance while executing strategies, in Experiment 1, and were asked to choose the best strategy of three available strategies to assess strategy selection, in Experiment 2. Presentation modality influenced strategy execution (i.e., children were faster and more accurate in problems under visual than auditory conditions) but only in children with low updating abilities. In contrast, presentation modality had no effect on children’s strategy selection. Presentation duration had an effect on both strategy execution and strategy selection with time-limited presentation leading to a decline in children’s performance. Interestingly, specifically in children with low updating abilities, time-limited presentation led to poorer performance. Hence, efficient updating seemed to compensate for detrimental effects of auditory in comparison to visual and time-limited in comparison to time-unlimited presentation. These findings have important implications for determining conditions under which children execute strategies most efficiently and select the best strategy on each problem most often, as well as for understanding mechanisms underlying strategic behaviour.

Mathematics-gender stereotype endorsement influences mathematics anxiety, self-concept, and performance differently in men and women

Mathematics anxiety (MA) is negatively associated with mathematics performance. Although some aspects, such as mathematics self-concept (M-self-concept), seem to modulate this association, the underlying mechanism is still unclear. In addition, the false gender-stereotype according to which women are worse than men in mathematics, can have a detrimental effect on women. Nevertheless, the role that endorsement of this stereotype can have might differ between men and women. Therefore, within a structural equational approach, we investigated how MA and mathematics self-concept relate to arithmetic performance when considering one’s mathematics-gender stereotype endorsement and gender in a large sample (N = 923) of university students. Mathematics-gender stereotype endorsement influenced arithmetic performance through different mediation patterns via MA, M-self-concept in men and women. It was linked to higher MA, lower M-self-concept, and arithmetic performance in women, while in men, its effect was generally weaker but more complex (it was linked to higher M-self-concept and slightly higher numerical anxiety component of MA). Moreover, men and women perceived the questions included in the considered instruments differently, implying that their numerical scores may not be directly comparable, which has even broader theoretical and methodological implications for MA research.

Personalized brain stimulation for effective neurointervention across participants

Accumulating evidence from human-based research has highlighted that the prevalent one-size-fits-all approach for neural and behavioral interventions is inefficient. This approach can benefit one individual, but be ineffective or even detrimental for another. Studying the efficacy of the large range of different parameters for different individuals is costly, time-consuming and requires a large sample size that makes such research impractical and hinders effective interventions. Here an active machine learning technique is presented across participants—personalized Bayesian optimization (pBO)—that searches available parameter combinations to optimize an intervention as a function of an individual’s ability. This novel technique was utilized to identify transcranial alternating current stimulation (tACS) frequency and current strength combinations most likely to improve arithmetic performance, based on a subject’s baseline arithmetic abilities. The pBO was performed across all subjects tested, building a model of subject performance, capable of recommending parameters for future subjects based on their baseline arithmetic ability. pBO successfully searches, learns, and recommends parameters for an effective neurointervention as supported by behavioral, simulation, and neural data. The application of pBO in human-based research opens up new avenues for personalized and more effective interventions, as well as discoveries of protocols for treatment and translation to other clinical and non-clinical domains.

Using machine learning to model the training scalability of convolutional neural networks on clusters of GPUs

In this work, we build a general piece-wise model to analyze data-parallel (DP) training costs of convolutional neural networks (CNNs) on clusters of GPUs. This general model is based on i) multi-layer perceptrons (MLPs) in charge of modeling the NVIDIA cuDNN/cuBLAS library kernels involved in the training of some of the state-of-the-art CNNs; and ii) an analytical model in charge of modeling the NVIDIA NCCL Allreduce collective primitive using the Ring algorithm. The CNN training scalability study performed using this model in combination with the Roofline technique on varying batch sizes, node (floating-point) arithmetic performance, node memory bandwidth, network link bandwidth, and cluster dimension unveil some crucial bottlenecks at both GPU and cluster level. To provide evidence of this analysis, we validate the accuracy of the proposed model against a Python library for distributed deep learning training.