Search

In-context learning refers to the ability of a neural network to learn from information presented in its context. While traditional learning in neural networks requires adjusting network weights for every new task, in-context learning operates purely by updating internal activations without needing any updates to network weights. The emergence of this ability in large language models has led to a paradigm shift in machine learning and has forced researchers to reconceptualize how they think about learning in neural networks. Looking beyond language models, we can find in-context learning in many computational models relevant to cognitive science, including those that emerge from meta-learning.

Electroencephalogram (EEG) signals are physiological indicators of brain activity, offering the advantage of high temporal resolution for capturing subtle emotional changes and providing rich information for emotion recognition. However, extracting effective features from EEG data with a low signal-to-noise ratio poses a significant challenge that hinders progress in this research field. To address this issue, we propose a multi-view time-frequency contrastive learning framework called MV-TFCL to enhance the information representation capability of EEG signals from multiple perspectives. Firstly, we introduce a recursive neural network based on multi-scale time-frequency consistency, which integrates global semantic information across different scales through gated units. To our knowledge, this is the first proposal of the theory of multi-scale time-frequency consistency applied in emotion recognition research. Subsequently, we design a tree-structured time-frequency encoder to capture local semantic information within the time-frequency domain. Finally, we incorporate semantic consistency constraints from both global and local perspectives to learn more generalizable and robust features. Extensive experimental results on two publicly available datasets demonstrate the effectiveness and superiority of our proposed method.

This paper introduces a novel perspective to significantly mitigate catastrophic forgetting in continuous learning (CL), which emphasizes models' capacity to preserve existing knowledge and assimilate new information. Current replay-based methods treat every task and data sample equally and thus can not fully exploit the potential of the replay buffer.In response, we propose COgnitive REplay (CORE), which draws inspiration from human cognitive review processes. CORE includes two key strategies: Adaptive Quantity Allocation and Quality-Focused Data Selection. The former adaptively modulates the replay buffer allocation for each task based on its forgetting rate, while the latter guarantees the inclusion of representative data that best encapsulates the characteristics of each task within the buffer. Our approach achieves an average accuracy of 37.95\% on split-CIFAR10, surpassing the best baseline method by 6.52\%. Additionally, it significantly enhances the accuracy of the poorest-performing task by 6.30\% compared to the top baseline. Code is available at https://github.com/sterzhang/CORE.

Temporal language grounding (TLG) aims to localize query-related events in videos, which explores how to cognize relationships of video content with language descriptions. According to selective visual attention mechanism in cognitive science, people's cognition and understanding of what happens often rely on dynamic foreground information in the video. Nonetheless, background usually predominates the scenes so that query-related visual features and irrelevant ones are confused. Thus, we propose a Foreground Enhanced Network (FEN) to diminish the background effect from two aspects.FEN at first in spatial dimension explicitly models the evolving foreground in video features by removing relatively unchanged background content. Besides, we propose a progressive contrastive sample generation module to gradually learn the differences between the predicted proposal and its elongated proposals that include the former as a portion, thereby distinguishing similar neighborhood frames. Experiments on two common-used datasets show the efficacy of our model.

Children learn powerful internal models of the world around them from a few years of egocentric visual experience. Can such internal models be learned from a child's visual experience with highly generic learning algorithms or do they require strong inductive biases? Recent advances in collecting large-scale, longitudinal, developmentally realistic video datasets and generic self-supervised learning (SSL) algorithms are allowing us to begin to tackle this nature vs. nurture question. However, existing work typically focuses on image-based SSL algorithms and visual capabilities that can be learned from static images (e.g. object recognition), thus ignoring temporal aspects of the world. To close this gap, here we train self-supervised video models on longitudinal, egocentric headcam recordings collected from a child over a two year period in their early development (6-31 months). The resulting models are highly effective at facilitating the learning of action concepts from a small number of labeled examples; they have favorable data size scaling properties; and they display emergent video interpolation capabilities. Video models also learn more robust object representations than image-based models trained with the exact same data. These results suggest that important temporal aspects of a child's internal model of the world may be learnable from their visual experience using highly generic learning algorithms and without strong inductive biases.

Social rewards shape human behavior. During development, a caregiver guides a learner's behavior towards culturally aligned goals and values. How do these behaviors persist and generalize when the caregiver is no longer present, and the learner must continue autonomously? Here, we propose a model of value internalization where social feedback trains an internal social reward (ISR) model that generates internal rewards when social rewards are unavailable. Through empirical simulations, we show that an ISR model prevents agents from unlearning socialized behaviors and enables generalization in out-of-distribution tasks. Incomplete internalization, akin to "reward hacking" on the ISR, is observed when the model is undertrained. Finally, we show that our model internalizes prosocial behavior in a multi-agent environment. Our work provides a framework for understanding how humans acquire and generalize values and offers insights for aligning AI with human values.

Code Knowledge Tracing (CKT) aims to model students' programming proficiency from their coding activities. Existing approaches mainly rely on answer records and lack problem descriptions and knowledge concepts, which fail to capture the inherent information. To solve this problem, we propose ECKT, an Enhanced Code Knowledge Tracing framework using Large Language Models (LLMs), which simulate human cognitive process through chain-of-thought prompting and adapts quickly to new tasks with limited data using few-shot learning. Specifically, ECKT generates detailed problem descriptions and knowledge concepts from student code, enhancing the model's understanding of programming concepts and proficiency. Additionally, ECKT incorporates task difficulty information by correlating problems with difficulty levels based on student performance scores. This integration allows for a more accurate assessment of student proficiency across varying levels of difficulty. Also, ECKT can explicitly capture the essential information of code and learn a better representation of them. Experimental results demonstrate that ECKT effectively improves the performance of knowledge tracing in programming education. This advancement not only supports personalized learning but also contributes to a deeper understanding of coding activities.

Emerging folklore in the cognitive sciences suggests that interpretability techniques to reverse-engineer artificial neural networks (ANNs) could speed up discovery and theory-building. For many researchers in psychology, linguistics, neuroscience, and artificial intelligence (AI), the full observability and perturbability of ANNs trained on complex tasks affords a shortcut to domain insights, cognitive theories, neurocognitive models, application improvement, and user safety. Folklore intuitions, however, are typically disconnected from other relevant knowledge. Here we examine these intuitions formally by drawing relevant connections to computational complexity theory. We model interpretability queries computationally and analyze their resource demands for biological/artificial high-level cognition. We prove mathematically that, contrary to folklore, basic circuit-finding queries in classic ANNs are already infeasibly demanding to answer even approximately. We discuss how interdisciplinary integration can mitigate this disconnect and situate the broader implications for the cognitive sciences, the philosophy of AI-fueled discovery, and AI ethics.

Word Sense Disambiguation (WSD) has been a fundamental task for human language understanding. In specific contexts, a word may have different meanings. For rarely seen word senses, the disambiguation becomes challenging with limited examples. Meta-learning, as a widely adopted machine learning method for few-shot learning, addresses this by extracting metacognitive knowledge from training data, aiding models in "learning to learn". Hence, the advancement of meta-learning hinges on leveraging high-quality metacognitive knowledge. In light of this, we propose a Bi-Branch Meta-Learning method for WSD to enrich and accumulate metacognitive insights. Our method employs two branches during training and testing. During training, we use a bi-branch loss with original and augmented data from large language models to compensate for data scarcity. In testing, information from base classes generates bi-branch scores to refine predictions. Experiments show our method achieves a 74.3 F1 score in few-shot scenarios, demonstrating its potential for few-shot WSD.

Theory of mind is an essential ability for complex social interaction and collaboration. Researchers in cognitive science and psychology have previously sought to integrate theory of mind capabilities into artificial intelligence (AI) agents to improve collaborative abilities (Cuzzolin, Morelli, Cirstea, & Sarahakian, 2020). We introduce the Recurrent Conditional Variational Autoencoder (RCVAE), a novel model which leverages the ability of generative models to learn rich abstracted representations of contextual behaviors to predict behavioral intent from human behavioral trajectories. Advancing on current concept learning models, this model allows for the discovery of latent intent in human behavior trajectories, while maintaining the scalability and performance of generative AI models. We show that in the Overcooked-AI environment, the RCVAE outperforms baseline Long Short-Term Memory (LSTM) models in predicting intent, achieving higher prediction accuracy and greater predictive stability. The implications of these results are significant; the RCVAE's proficiency in learning the relationship between basic actions and resulting contextual behaviors represents a significant advancement in concept learning for behavioral intent prediction.

Humans learn that temporarily occluded objects continue to exist within the first months of their lives. Deep learning mod- els, on the other hand, struggle to generalize such concepts from observations, due to missing proper inductive biases. Here, we introduce the first self-supervised interpretable ma- chine learning model that learns about object permanence di- rectly from video data without supervision. We augment a slot- based autoregressive deep learning system with the ability to adaptively and selectively fuse latent imaginations with pixel- based observations into consistent object-specific ‘what' and ‘where' encodings over time. We show that (i) Loci-Looped tracks objects through occlusions and anticipates their reap- pearance while outperforming state-of-the-art baseline models, (ii) Loci-Looped shows signs of surprise when the principle of object permanence is violated, and (iii) Loci-Looped's internal latent loop is key for learning object permanence.

Eye movement detection plays a crucial role in various fields, including eye tracking applications and understanding human perception and cognitive states. Existing detection methods typically rely on gaze positions predicted by gaze estimation algorithms, which may introduce cumulative errors. While certain video-based methods, directly classifying behaviours from videos, have been introduced to address this issue, they often have limitations as they primarily focus on detecting blinks. In this paper, we propose a video-based two-stream framework designed to detect four eye movement behaviours—fixations, saccades, smooth pursuits, and blinks—from infrared near-eye videos. To explicitly capture motion information, we introduce optical flow as the input for one stream. Additionally, we propose a spatio-temporal feature fusion module to combine information from the two streams. The framework is evaluated on a large-scale eye movement dataset and performs excellent results.

Does grammatical gender influence the meaning of inanimate nouns? We examined word embeddings from distributional semantics models, representing meanings in a vector space. In 26 gendered languages and non-gendered English, we measured the meaning similarity of inanimate nouns to gendered anchor nouns like 'male' and 'female.' In gendered languages, noun meanings aligned more with the anchor noun congruent with grammatical gender. This effect persisted when comparing the same nouns across languages (e.g., 'cucchiaio' vs 'cuchara' vs 'spoon'). We propose that grammatical gender introduces a gender bias into lexical semantics through distributional similarities with anchor words, revealing masculine/feminine features even without direct sensorimotor experience. This suggests that embodiment in language processing may become statistically embedded in word usage patterns.

A hallmark of human vision is the ability to rapidly recognize objects in a complex naturalistic scene. However, the exact mechanisms behind the computational invariance of object recognition remain unknown. In this study, we investigate object constancy by estimating how the ventral visual stream processes shading, shadows, textures, and specularities. To accomplish this, we use object meshes from the Objaverse dataset to create distinct multiclass classification tasks. For every task, we render a dataset by excluding exactly one of the previously stated features at a time. Subsequently, we train a ResNet50 model on each dataset. The trained model is evaluated on Brain-Score; deviations in these metrics indicate the importance of a brain region in achieving invariance to a specific feature. A reduced score for a removed feature in a particular region implies its crucial role in processing that feature since the model classifies objects based on remaining scene intrinsics.

Although object detection AI plays an important role in many critical systems, corresponding Explainable AI (XAI) methods remain very limited. Here we first developed FullGrad-CAM and FullGrad-CAM++ by extending traditional gradient-based methods to generate object-specific explanations with higher plausibility. Since human attention may reflect features more in-terpretable to humans, we explored the possibility to use it as guidance to learn how to combine the explanatory information in the detector model to best present as an XAI saliency map that is interpretable (plausible) to humans. Interestingly, we found that human attention maps had higher faithfulness for explaining the detector model than existing saliency-based XAI methods. By using trainable activation functions and smoothing kernels to maximize the XAI saliency map similarity to human attention maps, the generated map had higher faithfulness and plausibility than both existing XAI methods and human atten-tion maps. The learned functions were model-specific, well generalizable to other databases.

For the last decade, convolutional neural networks (CNNs) have vastly superseded their predecessors in nearly all vision tasks in artificial intelligence, including object recognition. However, despite abundant advancements, they continue to pale in comparison to biological vision. This chasm has prompted the development of biologically-inspired models that have attempted to mimic the human visual system, primarily at a neural level, which is evaluated using standard dataset benchmarks. However, more work is needed to understand how these models perceive the visual world. This article proposes a state-of-the-art procedure that generates a new metric, Psychophysical-Score, which is grounded in visual psychophysics and is capable of reliably estimating perceptual responses across numerous models --- representing a large range in complexity and biological inspiration. We perform the procedure on twelve models that vary in degree of biological inspiration and complexity, we compare the results against the aggregated results of 2,390 Amazon Mechanical Turk workers who together provided $\sim2.7$ million perceptual responses. Each model's Psychophysical-Score is compared against the state-of-the-art neural activity-based metric, Brain-Score. Our study indicates that models with a high correlation to human perceptual behavior also have a high correlation with the corresponding neural activity.

There is major potential for using electroencephalography (EEG) in brain decoding that has been untapped due to the need for large amounts of data. Advances in machine learning have mitigated this need through data augmentation techniques, such as Generative Adversarial Networks (GANs). Here, we gauged the extent to which GANs can augment EEG data to enhance classification performance. Our objectives were to determine which classifiers benefit from GAN-augmented EEG and to estimate the impact of sample sizes on GAN-enhancements. We investigated three classifiers---neural networks, support vector machines, and logistic regressions---across seven sample sizes ranging from 5 to 100 participants. GAN-augmented EEG enhanced classification for neural networks and support vector machines, but not logistic regressions. Further, GAN-enhancements diminished as sample sizes increased---suggesting it is most effective with small samples, which may facilitate research that is unable to collect large amounts of data.

Facial recognition Deep Neural Networks (DNNs) and humans both show systematic differences in face recognition accuracy for different demographic groups. Differential accuracy has been shown due to race, age and gender, raising important questions about the impact of facial recognition decisions on the fairness of society. Current methods for measuring bias require curating large databases of face images, which is labor-intensive, producing bespoke tests that cannot be easily shared or standardized due to data privacy. Here we develop a novel solution to this problem inspired by psychological research in face perception. We ask whether face averages can be used to predict face recognition accuracy differentials for DNNs and humans. We generate sets of average images from random samples of exemplar faces from a demographic group to measure the density of these sets in representational space. We find that this approach provides reliable predictions of differential accuracy across demographic groups in both DNNs and human participants. However, we also find evidence that face averages were not represented in the center of face categories. This finding should be addressed in future development of our approach, and also challenges influential cognitive models of face identity representation.

Recent advancements in Large Language Models (LLMs) harness linguistic associations in vast natural language data for practical applications. However, their ability to understand the physical world using only language data remains a question. After reviewing existing protocols, we explore this question using a novel and tightly controlled reasoning test (ART) and compare human norms against versions of GPT-3. Our findings highlight the categories of common-sense relations models that could learn directly from data and areas of weakness. GPT-3 offers evidence for verbal reasoning on a par with human subjects for several relations including Synonymy, Antonymy, and Default inheritance, Without reinforcement learning from human judgements, it appears GPT-3 performs at the lower end of the reference interval for Has-part and Contained-in. Weaknesses were observed also in affordance characteristics through Necessary-quality, Order-of-size and Order-of-intensity. Combining LLMs with symbolic world grounding is a promising direction to address associative learning.

In this paper, we study human sequential behavior by integrating cognitive, evolutionary, and computational approaches. Our work centers around the emergence of shared vocabularies in the Embodied Communication Game (ECG). Here, participant pairs solve a shared task without access to conventional means of communication, enforcing the emergence of a new communication system. This problem is solved typically by negotiating a shared set of sequential signals that acquire meaning through interactions. Individual differences in Personal Need for Structure (PNS) have been found to influence how this process develops. We trained deep neural networks to mimic the emergence of new communicative systems and used hyperparameter optimization to approximate latent human cognitive variables to explain human behavior. We demonstrate that models based on bidirectional LSTM networks are better at capturing human behavior than unidirectional LSTM networks. This suggests that human sequence processing in the ECG is influenced by expected future states. The approximated variables cannot explain the differences in PNS, but we do provide evidence suggesting that random and uncertainty-directed exploration strategies are combined to develop optimal behavior.

Brain function connectivity, derived from functional magnetic resonance imaging (fMRI), has enjoyed high popularity in studies of Autism Spectrum Disorder (ASD) diagnosis. Albeit rapid progress has been made, most studies still suffer from several knotty issues: (1) the hardship of effectively modeling the sophisticated brain neuronal connectivity;(2) the dimensionality explosion resulted from excessive voxels in each fMRI sample; (3) the poor interpretability giving rise to unpersuasive diagnosis. To ameliorate these issues, we propose a GCN-based model, namely ConResNet. Specifically, a convolutional restructuring attention is designed for efficiently ROI-interaction feature extraction. Besides, we embed a salient graph pooling method to sift salient nodes. Extensive experiments conducted on ABIDE dataset demonstrate ConResNet achieves state-of-the-art performance. Moreover, the most salient brain regions predicted by ConResNet are closely consistent with theoretical knowledge in the medical domain, providing potential biomarkers for ASD clinical diagnosis.

Autoencoder frameworks have received attention for graph embedding, particularly those utilizing generative adversarial networks (GAN). However, GAN-based frameworks do not fully utilize the original graph information and lack stable updates in the GAN component. To bridge this gap, we propose a dual-adversarial framework for graph embedding that expands mutual information (MI) in positive and negative samples for adversarial training using GAN. We further improve model performance by incorporating reinforcement learning ideas. Our framework includes two variants: a pessimistic adversarial graph autoencoder (PAGAE), and a pessimistic adversarial graph autoencoder with PO loss (PAGAEPO). Essentially, we present a pessimistic module to negative sample generator to boost original discriminator, thereby reinforcing the generator’s ability. Additionally, we designed a PO loss function on discriminator to stabilize the learning process and it will further improve the ability of model. Experimental results show that our models are competitive with state-of-the-art GAEs on benchmark datasets.

There is an apparent discrepancy between visual perception, which is colorful, complete, and in high resolution, and the saccadic, and spatially heterogeneous retinal input data. In this work, we computationally emulated foveated color maps and intensity channels as well as intra-saccadic motion data using a neuromorphic event camera. We used a convolutional neural network, U-Net, and adversarial optimization to demonstrate how retinal inputs can be used for the reconstruction of colorful images in high resolution. Our model may set the groundwork for the development of biologically plausible neural networks for computational vision perception.

Cross-lingual adaptation has proven effective in spoken language understanding (SLU) systems with limited resources.Existing methods are frequently unsatisfactory for intent detection and slot filling, particularly for distant languages that differ significantly from the source language in scripts, morphology, and syntax. Latent Dialogue Action (LaDA) layer is proposed to optimize decoding strategy in order to address the aforementioned issues. The model consists of an additional layer of latent dialogue action. It enables our model to improve a system's capability of handling conversations with complex multilingual intent and slot values of distant languages.To the best of our knowledge, this is the first exhaustive investigation of the use of latent variables for optimizing cross-lingual SLU policy during the decode stage. LaDA obtains state-of-the-art results on public datasets for both zero-shot and few-shot adaptation.

Dependency tree is efficient for relation extraction model to exploit relations between words.Recent approaches have achieved promising performance while still suffering from inherent limitations, such as the computation efficiency and flexible pruning strategies. In this paper, we propose a novel relation extraction framework called Progressive Graph Learning over pruned dependency tree (PGLNet). PGLNet constructs a set of graphs by progressively adapting to input sentence. Specially, we implement the model to construct progressive weighted adjacency matrices by learning the relations among graph nodes with multi-head self-attention mechanism.Then, the model takes the learned weights as reference to prune dependency tree in order to preserve useful relevant sub-structures for the relation extraction while removing irrelevant words. Next, progressive convolution module is designed to encode the relations of entities and followed by relations classification.We evaluate our proposed model using public real-world datasets, experimental results demonstrate that the proposed model achieves state-of-the-art performance with consistency in all datasets. We conclude that the ability of PGLNet to progressively adapt to input data and enable the model with robustness.

Bayesian models of theory of mind (ToM) have been successful in explaining how humans infer goals from the actions of other agents. However, they have typically been limited to small and fixed sets of possible goals specified in advance by the modeler, leaving open the question of how spontaneous goal inference occurs in rich and complex environments. To address this question, we posit that people are guided by informative, context-specific goal priors and proposals when making inferences about others. As proxies for these informed priors, we make use of context-conditioned large language models (LLMs) and integrate them into a Bayesian inverse planning framework. We find that LLMs can serve as usefully informative priors and proposals over goals compared to a structural baseline prior, allowing them to be used as models of the learned statistical knowledge that humans bring to bear in their inferences about others' goals.

Regular polysemes are sets of ambiguous words that all share the same relationship between their meanings, such as CHICKEN and LOBSTER both referring to an animal or its meat. To probe how a context embedding model, here exemplified by BERT, represents regular polysemy, we analyzed whether its embeddings support answering sense analogy questions similar to “is the mapping be- tween CHICKEN (as an animal) and CHICKEN (as a meat) the same as that which maps between LOBSTER (as an animal) to LOBSTER (as a meat)?” We found that (1) the model was sensitive to the shared structure within a regularity type; (2) the shared structure varies across regularity types, potentially reflective of a “regularity continuum;” (3) some high-order latent structure may be shared across regularity types, suggestive of a similar la- tent structure across types; and (4) there is equivocal ev- idence that the aforementioned effects are explained by meaning overlap.

Given that the current knowledge tracking model still has problems such as insufficient interpretability of the knowledge forgetting process and ignoring the relationship between different topics with the same knowledge concept. we propose an interpretable knowledge tracing model (KVFKT) based on key-value and forgetting. This approach keeps trace of the students' knowledge state matrices in each memory unit and also keeps trace of the forgetting matrices to monitor when the students' knowledge concepts were last reviewed. In addition, this model introduces the IRT two-parameter model from the perspective of educational psychology to further enhance the interpretability of the model. Finally, four real-world datasets are used to test the model, experimental results show that KVFKT can better trace students’ knowledge state and outperforms existing models in terms of ACC and AUC.

The unprecedented spread of smartphone usage and its various boarding sensors have been garnering increasing interest in automatic mental health detection. However, there are two major barriers to reliable mental health detection applications that can be adopted in real-life: (a)The outputs of the complex machine learning model are not explainable, which reduces the trust of users and thus hinders the application in real-life scenarios. (b)The sensor signal distribution discrepancy across individuals is a major barrier to accurate detection since each individual has their own characteristics. We propose an explainable mental health detection model. Spatial and temporal features of multiple sensory sequences are extracted and fused with different weights generated by the attention mechanism so that the discrepancy of contribution to classifiers across different modalities can be considered in the model. Through a series of experiments on real-life datasets, results show the effectiveness of our model compared to the existing approaches.

To develop computational agents that better communicate using their own emergent language, we endow the agents with an ability to focus their attention on particular concepts in the environment.Humans often understand an object or scene as a composite of concepts and those concepts are further mapped onto words.We implement this intuition as cross-modal attention mechanisms in Speaker and Listener agents in a referential game and show attention leads to more compositional and interpretable emergent language.We also demonstrate how attention aids in understanding the learned communication protocol by investigating the attention weights associated with each message symbol and the alignment of attention weights between Speaker and Listener agents.Overall, our results suggest that attention is a promising mechanism for developing more human-like emergent language.

There is growing interest in emotions in textual data. Based on psychological theories, research has been conducted on assigning emotions as labels to text datasets or developing models to detect emotions present in text. However, little research has been done on the appropriateness of using these theories. In this study, we reviewed three commonly used basic emotion theories: Ekman’s Basic Emotions, Plutchik’s Wheel of Emotions, and GoEmotions. By leveraging a research finding that evaluated the emotional values of words, we were able to fine-tune a language model emotionally. Using it, we analyzed the emotional relationship between the names of basic emotions and evaluated the adequacy of the emotional structure each theory presents. Clear patterns of similarity emerged based on the emotional meaning of the words. Ekman’s and GoEmotions were almost in line with our results, while Plutchik’s had some differences. We discussed these matches and mismatches.

Artificial intelligence is one of the most popular, frequently discussed, and, meanwhile, ambiguous and controversial metaphorical concepts, which defines a scientific direction in computer science that studies the techniques for gaining knowledge, their computer representation, transformation, and application. Presently, it is intensively penetrating into many areas of human activities, including hydrometeorological ones. The concept of artificial intelligence, the history of its origin, and its methods and technologies are considered. The author analyzes the studies related to the use of artificial intelligence in short- and medium-range weather forecasting, including the collection and quality control of meteorological information, assimilation of data in order to generate initial conditions for numerical weather prediction models, development of forecast models and parameterization schemes for physical processes, postprocessing and physical-statistical interpretation of the output data of numerical weather prediction models. [ABSTRACT FROM AUTHOR]

Detecting micron-sized particles is an essential task for the analysis of complex plasmas because a large part of the analysis is based on the initially detected positions of the particles. Accordingly, high accuracy in particle detection is desirable. Previous studies have shown that machine learning algorithms have made great progress and outperformed classical approaches. This work presents an approach for tracking micron-sized particles in a dense cloud of particles in a dusty plasma at Plasmakristall-Experiment 4 using a U-Net. The U-net is a convolutional network architecture for the fast and precise segmentation of images that was developed at the Computer Science Department of the University of Freiburg. The U-Net architecture, with its intricate design and skip connections, has been a powerhouse in achieving precise object delineation. However, as experiments are to be conducted in resource-constrained environments, such as parabolic flights, preferably with real-time applications, there is growing interest in exploring less complex U-net architectures that balance efficiency and effectiveness. We compare the full-size neural network, three optimized neural networks, the well-known StarDist and trackpy, in terms of accuracy in artificial data analysis. Finally, we determine which of the compact U-net architectures provides the best balance between efficiency and effectiveness. We also apply the full-size neural network and the the most effective compact network to the data of the PK-4 experiment. The experimental data were generated under laboratory conditions. [ABSTRACT FROM AUTHOR]

Many theories of verbal working memory have claimed that novel word lists are unlike natural language. As a result, recall of the novel word lists is not affected by sequential statistics in natural language and must be governed by short-term memory. We analyzed memory for novel, dative-like word lists (e.g., ADJ-NOUN-GAVE-ADJ-NOUN-NOUN) across three experiments. We used a powerful neural language model (Ng et al., 2019) to compute the surprisal associated with the critical words in each list, because surprisal indexes psycholinguistic processing difficulty (Levy, 2008). Surprisal at critical words strongly predicted the number of words recalled in each list. These results demonstrate lexico-syntactic factors can support syntactic structure even in novel word lists and that continuous measures of processing difficulty may be used to characterize similarity of novel lists to natural language. We discuss implications for theories of verbal working memory.

Formal mathematical reasoning is unique in its precision: any valid conclusion can be justified by a sequence of base axioms. But human-written proofs or solutions rarely operate at that level. Instead, obvious steps are skipped to provide a simple, lucid argument. This is especially important in an educational setting, where too many details in an example solution, or too few, can confuse a student. What are the key steps for humans in a given formal solution? We investigate several computational hypotheses in the context of equation solving. Specifically, we take a reinforcement learning agent that solves equations using low-level axioms, and propose a series of methods for abstracting its solutions by selecting key steps. We consider methods based on the semantic distance between subsequent steps, based on the steps with the highest uncertainty for the agent, and based on transitions between latent "high-level skills" learned from a large number of agent-produced solutions. In a human evaluation we find that skill-base simplifications were judged most useful. These results suggest new directions for understanding human mathematical reasoning.

During speech perception, humans constantly sense multiple sources of information. As so, the McGurk effect has been typically conceived as a prototypical example of language multi-modal integration, specifically, it has been used to study audio-visualintegration during speech. The McGurk effect arises when in-congruent audio-visual stimuli} are paired and perceived as a different syllable ( auditory /ba/ +visual /ga/ = percept “da” ).We developed a hierarchical computer model, based on self-organizing maps and Hebbian learning to study speech multi-modal integration. Our architecture allows studying the McGurk effect purely on bottom-up processing of audio-visual information.We trained several versions of the model and measured the activation similarity betweenMcGurk alike stimuli and congruent ones by means of mutual information. Our results suggest that the illusorypercept arises from the best congruent representation that reduces uncertainty. Furthermore, the reliability of each sensory modality determines the best congruent representation.

English is the dominant language in computer science. In addition to English-based academic papers, English is frequently the only language provided in introduction sections and manuals of command and software libraries, which are essential aspects of computer programming. Hence, English-as-a-second-language (ESL) learners may have difficulty studying computer science because they must learn this field while also learning English. Despite this problem, few studies have assessed the difficulty level of computer science texts for ESL learners.Ideally, the difficulty levels of texts are assessed by having groups of ESL learners read them. However, owing to the excessive time and financial costs involved, such practices can be impractical.Hence, using two highly accurate automatic readability assessors based on natural language processing (NLP) techniques, we assessed the readability of various computer-science-related texts for ESL learners. The first assessor is based on state-of-the-art deep transfer learning, and the second is based on classical machine learning and applied linguistics. For training the assessors, we used a standard corpus employed in NLP, which was annotated by professional English teachers to evaluate the readability of the texts for ESL learners. To conduct the experiments, we built a collection of computer science texts ranging from academic papers to software manuals (READMEs) crawled from a source-code hosting website, namely GitHub. The experimental results showed that intermediate ESL learners were able to read most of the computer science related texts.

Modern Hopfield networks (HNs) exhibit properties of a Content Addressable Memory (CAM) that can store and retrieve a large number of memories. They also provide a basis for modelling associative memory in humans. However, the implementation of these networks is often not biologically plausible as they assume the strengths of synaptic connections are symmetric, and utilize functions that rely on many-body synapses. More biologically realistic versions of Modern HNs have been proposed, although these implementations often still utilize the softmax function. Computing the softmax for a single node requires the knowledge of all other neurons, and thus still poses a degree of biological implausibility. We present a Modern HN that uses a version of softmax that can be computed in a more bio-realistic way, and hence moves us closer to a model of memory that is biologically sound. We also show that our proposed network can learn the connection weights using a local learning rule, derived from gradient descent on the energy function. Finally, we verify that our proposed biological network behaves like a Modern HN and explore some of its other interesting properties.

Over-paramaterized neural models have become dominant in Natural Language Processing. Increasing the size of a neural network seems to result in improved performance across a a broad range of tasks. Despite their size these models have been shown to generalize poorly outside their training data. Seemingly failing to extract the systematic generalizations that humans use to generate and interpret language. Increasingly work has questioned whether these models are learning to generalize or memorize, with larger capacity models potentially just memorizing their data more and more effectively. We suggest the tradeoff between memorization and generalization may be more nuanced; with the capacity of a model shaping the kinds of generalizations they are likely to acquire. Our results on a linguistic task suggest that while all models develop generalization strategies, smaller models may arrive at a smaller distribution of strategies that generalize more robustly to novel data.

Children do not learn language from passively analyzing correlations between language and observations, but from interaction with caregivers or peers. The non-nativist approach claims that the main driver of language learning should be to achieve communicative goals. Imitation, on the other hand, is another natural desire that many argue influences language learning. However, there are still gaps in the research on what roles communicative goals and imitating linguistic input play in language acquisition, due to the difficulty of performing comprehensive experiments with human learners. In this paper, we propose a computational framework using simulated experiments that allows us to compare the roles of the two drivers. Specifically, we simulate a two-way communication game between a speaker, corresponding to a language learner, and a listener, corresponding to a caregiver or teacher. The speaker's communicative goals are modeled as rewards for successful completion of a referential game, and imitation is performed by mimicking feedback from the listener. The listener adaptively chooses to give feedback and makes choices based on the speaker's utterances. With empirical results on naturalistic visual and language data, we find that communicative goals play an important role in driving language learning, whereas imitation accelerates the learning process. We also find that (1) models trained with communicative goals tend to use minimal vocabulary and utterances and overextend them to concepts outside the original word meanings; (2) the strategy with which the listener provides feedback also influences the learning results and speed. Code and data for replicating the experiments are available\footnote{\url{https://bit.ly/interactgym}} to spur future research on models for computational studies of language learning.

Although sentence comprehension and production are increasingly often studied by combining computational modeling and human experiments, this approach remains mostly restricted to studies of monolingual or first-language (L1) processing. There are currently only very few sentence-level computational models of second-language (L2) or bilingual processing (Frank, 2021). This lack of computational specifications can hamper further progress in bilingualism research. Moreover, better understanding of bilingual processing will give more insights into more general mechanisms such as cognitive control processes involved while switching languages (Luk et al., 2012). Our symposium aims to bring together researchers from different labs and with different research traditions, working on the intersection of models and experiments in bilingual sentence processing.The symposium has four talks, by Edith Kaan (associate professor, specializing in psycholinguistics of bilingualism), Yung Han Khoe (PhD student, working on models of bilingual sentence production), Lin Chen (research associate with an expertise in reading processes), and a joint talk by Irene Winther (PhD student working on bilingual sentence processing) and Yevgen Matusevych (research associate in computational cognitive science of language). Finally, we will have a panel discussion to suggest how models could be challenged by experimental data, and provide new explanatory mechanisms. This discussion will be moderated by Xavier Hinaut (research scientist in computational neuroscience) and Stefan Frank (associate professor in computational psycholinguistics).

Complex systems, as interwoven miscellaneous interacting entities that emerge and evolve through self-organization in a myriad of spiraling contexts, exhibit subtleties on global scale besides steering the way to understand complexity which has been under evolutionary processes with unfolding cumulative nature wherein order is viewed as the unifying framework. Indicating the striking feature of non-separability in components, a complex system cannot be understood in terms of the individual isolated constituents' properties per se, it can rather be comprehended as a way to multilevel approach systems behavior with systems whose emergent behavior and pattern transcend the characteristics of ubiquitous units composing the system itself. This observation specifies a change of scientific paradigm, presenting that a reductionist perspective does not by any means imply a constructionist view; and in that vein, complex systems science, associated with multiscale problems, is regarded as ascendancy of emergence over reductionism and level of mechanistic insight evolving into complex system. While evolvability being related to the species and humans owing their existence to their ancestors' capability with regards to adapting, emerging and evolving besides the relation between complexity of models, designs, visualization and optimality, a horizon that can take into account the subtleties making their own means of solutions applicable is to be entailed by complexity. Such views attach their germane importance to the future science of complexity which may probably be best regarded as a minimal history congruent with observable variations, namely the most parallelizable or symmetric process which can turn random inputs into regular outputs. Interestingly enough, chaos and nonlinear systems come into this picture as cousins of complexity which with tons of its components are involved in a hectic interaction with one another in a nonlinear fashion amongst the other related systems and fields. Relation, in mathematics, is a way of connecting two or more things, which is to say numbers, sets or other mathematical objects, and it is a relation that describes the way the things are interrelated to facilitate making sense of complex mathematical systems. Accordingly, mathematical modeling and scientific computing are proven principal tools toward the solution of problems arising in complex systems' exploration with sound, stimulating and innovative aspects attributed to data science as a tailored-made discipline to enable making sense out of voluminous (-big) data. Regarding the computation of the complexity of any mathematical model, conducting the analyses over the run time is related to the sort of data determined and employed along with the methods. This enables the possibility of examining the data applied in the study, which is dependent on the capacity of the computer at work. Besides these, varying capacities of the computers have impact on the results; nevertheless, the application of the method on the code step by step must be taken into consideration. In this sense, the definition of complexity evaluated over different data lends a broader applicability range with more realism and convenience since the process is dependent on concrete mathematical foundations. All of these indicate that the methods need to be investigated based on their mathematical foundation together with the methods. In that way, it can become foreseeable what level of complexity will emerge for any data desired to be employed. With relation to fractals, fractal theory and analysis are geared toward assessing the fractal characteristics of data, several methods being at stake to assign fractal dimensions to the datasets, and within that perspective, fractal analysis provides expansion of knowledge regarding the functions and structures of complex systems while acting as a potential means to evaluate the novel areas of research and to capture the roughness of objects, their nonlinearity, randomness, and so on. The idea of fractional-order integration and differentiation as well as the inverse relationship between them lends fractional calculus applications in various fields spanning across science, medicine and engineering, amongst the others. The approach of fractional calculus, within mathematics-informed frameworks employed to enable reliable comprehension into complex processes which encompass an array of temporal and spatial scales notably provides the novel applicable models through fractional-order calculus to optimization methods. Computational science and modeling, notwithstanding, are oriented toward the simulation and investigation of complex systems through the use of computers by making use of domains ranging from mathematics to physics as well as computer science. A computational model consisting of numerous variables that characterize the system under consideration allows the performing of many simulated experiments via computerized means. Furthermore, Artificial Intelligence (AI) techniques whether combined or not with fractal, fractional analysis as well as mathematical models have enabled various applications including the prediction of mechanisms ranging extensively from living organisms to other interactions across incredible spectra besides providing solutions to real-world complex problems both on local and global scale. While enabling model accuracy maximization, AI can also ensure the minimization of functions such as computational burden. Relatedly, level of complexity, often employed in computer science for decision-making and problem-solving processes, aims to evaluate the difficulty of algorithms, and by so doing, it helps to determine the number of required resources and time for task completion. Computational (-algorithmic) complexity, referring to the measure of the amount of computing resources (memory and storage) which a specific algorithm consumes when it is run, essentially signifies the complexity of an algorithm, yielding an approximate sense of the volume of computing resources and seeking to prove the input data with different values and sizes. Computational complexity, with search algorithms and solution landscapes, eventually points toward reductions vis à vis universality to explore varying degrees of problems with different ranges of predictability. Taken together, this line of sophisticated and computer-assisted proof approach can fulfill the requirements of accuracy, interpretability, predictability and reliance on mathematical sciences with the assistance of AI and machine learning being at the plinth of and at the intersection with different domains among many other related points in line with the concurrent technical analyses, computing processes, computational foundations and mathematical modeling. Consequently, as distinctive from the other ones, our special issue series provides a novel direction for stimulating, refreshing and innovative interdisciplinary, multidisciplinary and transdisciplinary understanding and research in model-based, data-driven modes to be able to obtain feasible accurate solutions, designed simulations, optimization processes, among many more. Hence, we address the theoretical reflections on how all these processes are modeled, merging all together the advanced methods, mathematical analyses, computational technologies, quantum means elaborating and exhibiting the implications of applicable approaches in real-world systems and other related domains. [ABSTRACT FROM AUTHOR]

The recognition of human activities (HAR) using wearable device data, such as smartwatches, has gained significant attention in the field of computer science due to its potential to provide insights into individuals' daily activities. This article aims to conduct a comparative study of deep learning techniques for recognizing activities of daily living (ADL). A mapping of HAR techniques was performed, and three techniques were selected for evaluation, along with a dataset. Experiments were conducted using the selected techniques to assess their performance in ADL recognition, employing standardized evaluation metrics, such as accuracy, precision, recall, and F1-score. Among the evaluated techniques, the DeepConvLSTM architecture, consisting of recurrent convolutional layers and a single LSTM layer, achieved the most promising results. These findings suggest that software applications utilizing this architecture can assist smartwatch users in understanding their movement routines more quickly and accurately. [ABSTRACT FROM AUTHOR]