Learning Mechanism Research Articles

Robust Multi-Agent Communication With Graph Information Bottleneck Optimization.

Recent research on multi-agent reinforcement learning (MARL) has shown that action coordination of multi-agents can be significantly enhanced by introducing communication learning mechanisms. Meanwhile, graph neural network (GNN) provides a promising paradigm for communication learning of MARL. Under this paradigm, agents and communication channels can be regarded as nodes and edges in the graph, and agents can aggregate information from neighboring agents through GNN. However, this GNN-based communication paradigm is susceptible to adversarial attacks and noise perturbations, and how to achieve robust communication learning under perturbations has been largely neglected. To this end, this paper explores this problem and introduces a robust communication learning mechanism with graph information bottleneck optimization, which can optimally realize the robustness and effectiveness of communication learning. We introduce two information-theoretic regularizers to learn the minimal sufficient message representation for multi-agent communication. The regularizers aim at maximizing the mutual information (MI) between the message representation and action selection while minimizing the MI between the agent feature and message representation. Besides, we present a MARL framework that can integrate the proposed communication mechanism with existing value decomposition methods. Experimental results demonstrate that the proposed method is more robust and efficient than state-of-the-art GNN-based MARL methods.

Self-adaptive enhanced learning differential evolution with surprisingly efficient decomposition approach for parameter identification of photovoltaic models

In order for the photovoltaic power generation system to convert energy with the highest efficiency after installation, the unknown parameters of the photovoltaic model need to be accurately set in advance. Therefore, in order to identify unknown parameters of photovoltaic models more reliably, effectively, and accurately, this paper first uses the decomposition approach that is extremely effective for different types of photovoltaic models, called matrix rotation decomposition. The unknown parameters of different photovoltaic models are decomposed into linear parameters and nonlinear parameters through matrix rotation decomposition approach. Then a simple and effective algorithm—Self-adaptive enhanced learning differential evolution (SaELDE)—is proposed to identify the decomposed nonlinear parameters. In SaELDE, there are three improvements: (1) the classification mutation learning mechanism is designed to make classified individuals evolve towards the promising direction; (2) the introduced crossover rate sorting approach effectively makes up for the neglected relationship between individuals and crossover rates; (3) combined with the population reduction strategy, it avoids the impact of the initial population setting on algorithm performance and balances the transition between the exploration and exploitation phases. Finally, the linear parameters are calculated by calculating the derived matrix equations according to the nonlinear parameters. Experimental results from a variety of common photovoltaic models, and even photovoltaic module models that need to face different temperatures and irradiance, show that the proposed SaELDE performs more accurately and has better robustness. In addition, the correlation coefficient of SaELDE is 1 under all conditions, which further proves the accuracy of the parameters extracted by SaELDE.

The developmental cognitive mechanism of learning algebraic rules from the dual-process theory perspective.

Rule learning is an important ability that enables human beings to adapt to nature and develop civilizations. There have been many discussions on the mechanism and characteristics of algebraic rule learning, but there are still controversies due to the lack of theoretical guidance. Based on the dual-process theory, this study discussed the following arguments for algebraic rule learning across human and animal studies: whether algebraic rule learning is simply Type 1 processing, whether algebraic rule learning is a domain-general ability, whether algebraic rule learning is shared by humans and animals, and whether an algebraic rule is learned consciously. Moreover, we propose that algebraic rule learning is possibly a cognitive process that combines both Type 1 and Type 2 processing. Further exploration is required to establish the essence and neural basis of algebraic rule learning.

Postnatal Development of Synaptic Plasticity at Hippocampal CA1 Synapses: Correlation of Learning Performance with Pathway-Specific Plasticity

To determine the critical timing for learning and the associated synaptic plasticity, we analyzed developmental changes in learning together with training-induced plasticity. Rats were subjected to an inhibitory avoidance (IA) task prior to weaning. While IA training did not alter latency at postnatal day (PN) 16, there was a significant increase in latency from PN 17, indicating a critical day for IA learning between PN 16 and 17. One hour after training, acute hippocampal slices were prepared for whole-cell patch clamp analysis following the retrieval test. In the presence of tetrodotoxin (0.5 µM), miniature excitatory postsynaptic currents (mEPSCs) and inhibitory postsynaptic currents (mIPSCs) were sequentially recorded from the same CA1 neuron. Although no changes in the amplitude of mEPSCs or mIPSCs were observed at PN 16 and 21, significant increases in both excitatory and inhibitory currents were observed at PN 23, suggesting a specific critical day for training-induced plasticity between PN 21 and 23. Training also increased the diversity of postsynaptic currents at PN 23 but not at PN 16 and 21, demonstrating a critical day for training-induced increase in the information entropy of CA1 neurons. Finally, we analyzed the plasticity at entorhinal cortex layer III (ECIII)-CA1 or CA3-CA1 synapses for each individual rat. At either ECIII-CA1 or CA3-CA1 synapses, a significant correlation between mean α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid/N-methyl-D-aspartic acid (AMPA/NMDA) ratio and learning outcomes emerged at PN 23 at both synapses, demonstrating a critical timing for the direct link between AMPA receptor-mediated synaptic plasticity and learning efficacy. Here, we identified multiple critical periods with respect to training-induced synaptic plasticity and delineated developmental trajectories of learning mechanisms at hippocampal CA1 synapses.

Effects of three prophylactic interventions on French middle-schoolers’ mental health: protocol for a randomized controlled trial

Adolescence is a strategic developmental stage in terms of preventing later difficulties and ensuring good mental health. Prophylactic interventions, which are conducted before the onset, prolongation, or worsening of difficulties, and aim to prevent or reduce symptoms or to promote wellbeing, therefore appear particularly appropriate for adolescents. However, existing prophylactic interventions conducted with adolescents have several weaknesses, including sparse theoretical frameworks, ambivalent evidence of their efficacy, and implementation and dissemination difficulties. In addition, no data are currently available on the effectiveness of such interventions in France. To fill this gap, a four-arm randomized controlled trial will be performed to assess the effectiveness of three prophylactic interventions targeting reactive, proactive and interpersonal adaptation in fourth-grade middle-school students, together with participants’ experience and perception of the interventions. Based on existing knowledge about adolescents, their learning mechanisms, and field constraints, these three interventions have been designed to promote their learning and receptiveness to interventions. Compared with baseline (i.e., before the intervention), we expect to observe a significant decrease in the level of distress (anxiety and depressive symptoms, functional impairment, and psychosocial difficulties) and a significant increase in the level of wellbeing after the intervention, across the three intervention groups, but not in the control group. In addition, we expect to observe post-intervention improvements in the processes targeted by the reactive adaptation intervention (operationalized as coping strategy use and flexibility), those targeted by the proactive adaptation intervention (operationalized as the tendency to engage in committed actions and general self-efficacy), and those targeted by the interpersonal adaptation intervention (operationalized as assertiveness in interactions), but only in the corresponding groups, with no change in any of these processes in the control group. The results of this research will not only enrich our knowledge of the processes involved in adolescents’ distress and wellbeing, but also provide clues as to the best targets for intervention. Moreover, the material for these interventions will be freely available in French on request to the corresponding author, providing access to innovative and fully assessed interventions aimed at promoting adolescents’ mental health in France.This clinical trial is currently being registered under no. 2023-A01973-42 on https://ansm.sante.fr/. This is the first version of the protocol.

Reduced learning rates but successful learning of a coordinated rhythmic movement by older adults.

Previous work has investigated the information-based mechanism for learning and transfer of learning in coordinated rhythmic movements. In those papers, we trained young adults to produce either 90° or 60° and showed in both cases that learning entailed learning to use relative position as information for the relative phase. This variable then supported transfer of learning to untrained coordinations +/30° on either side. In this article, we replicate the 90° study with younger adults and extend it by training older adults (aged between 55 and 65 years). Other work has revealed a steep decline in learning rate around this age, and no follow-up study has been able to successfully train older adults to perform a novel coordination. We used a more intensive training paradigm and showed that while older adult learning rates remain about half that of younger adults, given time they are able to acquire the new coordination. They also learn to use relative position, and consequently show the same pattern of transfer. We discuss implications for attempts to model the process of learning in this task.

Perceptions of medical students on narrow learning objectives and structured debriefing in medical escape rooms: a qualitative study

BackgroundEscape rooms are increasingly used in medical education as a complementary learning technique or even alternative to traditional educational approaches. Few studies focus on debriefing following medical escape rooms and how escape rooms can be used to achieve pre-defined learning objectives. Evaluating the use of narrow learning objectives may increase the depth of reflections and transform an engaging team event into an effective learning opportunity. This study aimed to explore participants’ experiences and perceived learning outcomes of narrow learning objectives in a medical escape room with debriefing.MethodsIn this explorative, qualitative study, participants saw a video lecture, participated in an escape room experience, and in a following debriefing. Throughout this learning session, the learning objectives concerned “exchange of information” and are therefore relatively narrow. Participants then participated in a semi-structured focus group interview and completed a demographic questionnaire. Participants were volunteer final-year medical students. Focus group interview recordings were transcribed and analysed using systematic text condensation.ResultsThirty-two students in eight groups completed the study. Five themes were described in the analysis of the focus group interviews: Experience with the narrow learning objectives, topics discussed in the debriefing, learning mechanisms, learning outcomes concerning exchange of information and influences of the learning approach.ConclusionsNarrow learning objectives and structured debriefing seem to increase perceived learning depth of medical escape room sessions. Using semi-structured debriefing still allows for discussions of other elements relevant to the students.Clinical trialsClinical.trials ID NCT04783259.

Deep Learning for Predicting Progression of Patellofemoral Osteoarthritis Based on Lateral Knee Radiographs, Demographic Data and Symptomatic Assessments.

In this study, we propose a novel framework that utilizes deep learning and attention mechanisms to predict the radiographic progression of patellofemoral osteoarthritis (PFOA) over a period of seven years. This study included subjects (1832 subjects, 3276 knees) from the baseline of the Multicenter Osteoarthritis Study (MOST). Patellofemoral joint regions-of interest were identified using an automated landmark detection tool (BoneFinder) on lateral knee X-rays. An end-to-end deep learning method was developed for predicting PFOA progression based on imaging data in a 5-fold cross-validation setting. To evaluate the performance of the models, a set of baselines based on known risk factors were developed and analyzed using gradient boosting machine (GBM). Risk factors included age, sex, BMI and WOMAC score, and the radiographic osteoarthritis stage of the tibiofemoral joint (KL score). Finally, to increase predictive power, we trained an ensemble model using both imaging and clinical data. Among the individual models, the performance of our deep convolutional neural network attention model achieved the best performance with an AUC of 0.856 and AP of 0.431; slightly outperforming the deep learning approach without attention (AUC=0.832, AP= 0.4) and the best performing reference GBM model (AUC=0.767, AP= 0.334). The inclusion of imaging data and clinical variables in an ensemble model allowed statistically more powerful prediction of PFOA progression (AUC = 0.865, AP=0.447), although the clinical significance of this minor performance gain remains unknown. The spatial attention module improved the predictive performance of the backbone model, and the visual interpretation of attention maps focused on the joint space and the regions where osteophytes typically occur. This study demonstrated the potential of machine learning models to predict the progression of PFOA using imaging and clinical variables. These models could be used to identify patients who are at high risk of progression and prioritize them for new treatments. However, even though the accuracy of the models were excellent in this study using the MOST dataset, they should be still validated using external patient cohorts in the future.

The recent research progress in neurobiological characteristics and pain regulation of the cerebrospinal fluid-contacting nucleus

AbstractThe ependymal epithelium forms the cerebrospinal fluid barrier, separating the brain and spinal cord from the cerebrospinal fluid. However, in specific regions of the central nervous system, there are neurons that directly interface with the cerebrospinal fluid, including neuronal bodies, dendrites, or axons, This constitutes what is referred to as the "cerebrospinal fluid contacting neurons system (CSF-CNS)". The research team led by Professor Zhang has successfully utilized cholera toxin subunit B coupled horseradish peroxidase complex (CB-HRP) to selectively label the specialized neuron system that interfaces with cerebrospinal fluid, pioneeringly designating it as the "cerebrospinal fluid-contacting nucleus", commonly referred to as the "CSF-contacting nucleus". For the first time, the discovery of the CSF-contacting nucleus provides compelling morphological evidence for the existence of a distinct neural structure within the brain parenchyma that establishes a connection with the cerebrospinal fluid, thereby suggesting its potential significance in facilitating material and information exchange between the brain parenchyma and cerebrospinal fluid. After conducting a comprehensive series of studies on the morphological structure, material expression, gene analysis and functional aspects of the CSF-contacting nucleus in rodents and non-human primates, it has been revealed that there are fibrous connections between the CSF-contacting nucleus and the cerebral cortex and subcortical nuclei being involved in the regulatory mechanisms of pain, cognition, learning and memory, emotion, addiction, stress and anxiety responses, visceral activity, olfaction, vision processing and perception, auditory processing, perception, motor control and coordination, homeostasis regulation including maintenance of body energy and fluid balance, as well as the control of sleep–wake cycles and synchronization of biological rhythms. Current experiments have confirmed that the CSF-contacting nucleus is related to pain, morphine dependence and withdrawal, learning and memory, as well as stress. This present article offers a comprehensive review of the neurobiological characteristics and recent advancements in pain regulation of the CSF-contacting nucleus. The aim is to provide novel insights into the investigation of pain regulation within bidirectional regulatory pathway between the brain and cerebrospinal fluid, with a specific focus on elucidating the role of the CSF-contacting nucleus as a bridge structure. Additionally, the objective of this research is to propose innovative strategies for pain management and associated disorders in the future.

Deep learning-based correlation analysis for probabilistic power flow considering renewable energy and energy storage

Developing photovoltaic (PV) and wind power is one of the most efficient approaches to reduce carbon emissions. Accumulating the PV and wind energy resources at different geographical locations can minimize total power output variance as injected into the power systems. To some extent, a low degree of the variance amplitude of the renewable resources can reduce the requirement of in-depth regulation and dispatch for the fossil fuel-based thermal power plants. Such an issue can alternatively reduce carbon emissions. Thus, the correlation problem by minimizing the variance of total PV and wind power plays a vital role in power system planning and operation. However, the synergistic effect of power output correlation is mainly considered on the generation side, and it is often neglected for the correlation relationship between the power grid components. To address this problem, this paper proposes a correlation coefficient analysis method for the power grid, which can quantify the relationship between energy storage and the probabilistic power flow (PPF) of the grid. Subsequently, to accelerate the mapping efficiency of power correlation coefficients, a novel deep neural network (DNN) optimized by multi-task learning and attention mechanism (MA-DNN) is developed to predict power flow fluctuations. Finally, the simulation results show that in IEEE 9-bus and IEEE14-bus systems, the strong correlation grouping percentage between the power correlation coefficients and power flow fluctuations reached 92% and 51%, respectively. The percentages of groups indicating weak correlation are 4% and 38%. In the modified IEEE 23-bus system, the computational accuracy of MA-DNN is improved by 37.35% compared to the PPF based on Latin hypercube sampling. Additionally, the MA-DNN regression prediction model exhibits a substantial improvement in assessing power flow fluctuations in the power grid, achieving a speed enhancement of 758.85 times compared to the conventional probability power flow algorithms. These findings provide the rapid selection of the grid access point with the minimum power flow fluctuations.

An Adaptive Temporal Convolutional Network Autoencoder for Malicious Data Detection in Mobile Crowd Sensing.

Mobile crowdsensing (MCS) systems rely on the collective contribution of sensor data from numerous mobile devices carried by participants. However, the open and participatory nature of MCS renders these systems vulnerable to adversarial attacks or data poisoning attempts where threat actors can inject malicious data into the system. There is a need for a detection system that mitigates malicious sensor data to maintain the integrity and reliability of the collected information. This paper addresses this issue by proposing an adaptive and robust model for detecting malicious data in MCS scenarios involving sensor data from mobile devices. The proposed model incorporates an adaptive learning mechanism that enables the TCN-based model to continually evolve and adapt to new patterns, enhancing its capability to detect novel malicious data as threats evolve. We also present a comprehensive evaluation of the proposed model's performance using the SherLock datasets, demonstrating its effectiveness in accurately detecting malicious sensor data and mitigating potential threats to the integrity of MCS systems. Comparative analysis with existing models highlights the performance of the proposed TCN-based model in terms of detection accuracy, with an accuracy score of 98%. Through these contributions, the paper aims to advance the state of the art in ensuring the trustworthiness and security of MCS systems, paving the way for the development of more reliable and robust crowdsensing applications.

CDEST: Class Distinguishability-Enhanced Self-Training Method for Adopting Pre-Trained Models to Downstream Remote Sensing Image Semantic Segmentation

The self-supervised learning (SSL) technique, driven by massive unlabeled data, is expected to be a promising solution for semantic segmentation of remote sensing images (RSIs) with limited labeled data, revolutionizing transfer learning. Traditional ‘local-to-local’ transfer from small, local datasets to another target dataset plays an ever-shrinking role due to RSIs’ diverse distribution shifts. Instead, SSL promotes a ‘global-to-local’ transfer paradigm, in which generalized models pre-trained on arbitrarily large unlabeled datasets are fine-tuned to the target dataset to overcome data distribution shifts. However, the SSL pre-trained models may contain both useful and useless features for the downstream semantic segmentation task, due to the gap between the SSL tasks and the downstream task. To adapt such pre-trained models to semantic segmentation tasks, traditional supervised fine-tuning methods that use only a small number of labeled samples may drop out useful features due to overfitting. The main reason behind this is that supervised fine-tuning aims to map a few training samples from the high-dimensional, sparse image space to the low-dimensional, compact semantic space defined by the downstream labels, resulting in a degradation of the distinguishability. To address the above issues, we propose a class distinguishability-enhanced self-training (CDEST) method to support global-to-local transfer. First, the self-training module in CDEST introduces a semi-supervised learning mechanism to fully utilize the large amount of unlabeled data in the downstream task to increase the size and diversity of the training data, thus alleviating the problem of biased overfitting of the model. Second, the supervised and semi-supervised contrastive learning modules of CDEST can explicitly enhance the class distinguishability of features, helping to preserve the useful features learned from pre-training while adapting to downstream tasks. We evaluate the proposed CDEST method on four RSI semantic segmentation datasets, and our method achieves optimal experimental results on all four datasets compared to supervised fine-tuning as well as three semi-supervised fine-tuning methods.

Underlying interactive neural mechanism of motor learning governed by the cerebellum, the basal ganglia, and motor/sensory cortex: a review from theoretical perspective

Underlying interactive neural mechanism of motor learning governed by the cerebellum, the basal ganglia, and motor/sensory cortex: a review from theoretical perspective

Selective Neural Entrainment Reveals Hierarchical Tuning to Linguistic Regularities in Reading

Abstract Reading is both a visual and a linguistic task, and as such it relies on both general-purpose, visual mechanisms and more abstract, meaning-oriented processes. Disentangling the roles of these resources is of paramount importance in reading research. The present study capitalizes on the coupling of Fast Periodic Visual Stimulation (FPVS; Rossion, 2014) and MEG recordings to address this issue and investigate the role of dierent kinds of visual and linguistic units in the visual word identification system. We compared strings of pseudo-characters (BACS; C. Vidal & Chetail, 2017); strings of consonants (e.g,. sfcl); readable, but unattested strings (e.g., amsi); frequent, but non-meaningful chunks (e.g., idge); suffixes (e.g., ment); and words (e.g., vibe); and looked for discrimination responses with a particular focus on the ventral, occipito-temporal regions. The results revealed sensitivity to alphabetic, readable, familiar and lexical stimuli. Interestingly, there was no discrimination between suffixes and equally frequent, but meaningless endings, thus highlighting a lack of sensitivity to semantics. Taken together, the data suggest that the visual word identification system, at least in its early processing stages, is particularly tuned to form-based regularities, most likely reflecting its reliance on general-purpose, statistical learning mechanisms that are a core feature of the visual system as implemented in the ventral stream.

Efficient parameter extraction of photovoltaic models with a novel enhanced prairie dog optimization algorithm

The growing demand for solar energy conversion underscores the need for precise parameter extraction methods in photovoltaic (PV) plants. This study focuses on enhancing accuracy in PV system parameter extraction, essential for optimizing PV models under diverse environmental conditions. Utilizing primary PV models (single diode, double diode, and three diode) and PV module models, the research emphasizes the importance of accurate parameter identification. In response to the limitations of existing metaheuristic algorithms, the study introduces the enhanced prairie dog optimizer (En-PDO). This novel algorithm integrates the strengths of the prairie dog optimizer (PDO) with random learning and logarithmic spiral search mechanisms. Evaluation against the PDO, and a comprehensive comparison with eighteen recent algorithms, spanning diverse optimization techniques, highlight En-PDO’s exceptional performance across different solar cell models and CEC2020 functions. Application of En-PDO to single diode, double diode, three diode, and PV module models, using experimental datasets (R.T.C. France silicon and Photowatt-PWP201 solar cells) and CEC2020 test functions, demonstrates its consistent superiority. En-PDO achieves competitive or superior root mean square error values, showcasing its efficacy in accurately modeling the behavior of diverse solar cells and performing optimally on CEC2020 test functions. These findings position En-PDO as a robust and reliable approach for precise parameter estimation in solar cell models, emphasizing its potential and advancements compared to existing algorithms.

An Organic Optoelectronic Synapse with Multilevel Memory Enabled by Gate Modulation.

Artificial synaptic devices are emerging as contenders for next-generation computing systems due to their combined advantages of self-adaptive learning mechanisms, high parallel computation capabilities, adjustable memory level, and energy efficiency. Optoelectronic devices are particularly notable for their responsiveness to both voltage inputs and light exposure, making them attractive for dynamic modulation. However, engineering devices with reconfigurable synaptic plasticity and multilevel memory within a singular configuration present a fundamental challenge. Here, we have established an organic transistor-based synaptic device that exhibits both volatile and nonvolatile memory characteristics, modulated through gate voltage together with light stimuli. Our device demonstrates a range of synaptic behaviors, including both short/long-term plasticity (STP and LTP) as well as STP-LTP transitions. Further, as an encoding unit, it delivers exceptional read current levels, achieving a program/erase current ratio exceeding 105, with excellent repeatability. Additionally, a prototype 4 × 4 matrix demonstrates potential in practical neuromorphic systems, showing capabilities in the perception, processing, and memory retention of image inputs.

Deep unsupervised learning using Spike-Timing-Dependent Plasticity

Abstract Spike-Timing-Dependent Plasticity (STDP) is an unsupervised learning mechanism for Spiking Neural Networks (SNNs) that has received significant attention from the neuromorphic hardware community. However, scaling such local learning techniques to deeper networks and large-scale tasks has remained elusive. In this work, we investigate a Deep-STDP framework where a rate-based convolutional network, that can be deployed in a neuromorphic setting, is trained in tandem with pseudo-labels generated by the STDP clustering process on the network outputs. We achieve $24.56\%$ higher accuracy and $3.5\times$ faster convergence speed at iso-accuracy on a 10-class subset of the Tiny ImageNet dataset in contrast to a $k$-means clustering approach.

One Size Does Not Fit All: Idiographic Computational Models Reveal Individual Differences in Learning and Meta-Learning Strategies.

Complex skill learning depends on the joint contribution of multiple interacting systems: working memory (WM), declarative long-term memory (LTM) and reinforcement learning (RL). The present study aims to understand individual differences in the relative contributions of these systems during learning. We built four idiographic, ACT-R models of performance on the stimulus-response learning, Reinforcement Learning Working Memory task. The task consisted of short 3-image, and long 6-image, feedback-based learning blocks. A no-feedback test phase was administered after learning, with an interfering task inserted between learning and test. Our four models included two single-mechanism RL and LTM models, and two integrated RL-LTM models: (a) RL-based meta-learning, which selects RL or LTM to learn based on recent success, and (b) a parameterized RL-LTM selection model at fixed proportions independent of learning success. Each model was the best fit for some proportion of our learners (LTM: 68.7%, RL: 4.8%, Meta-RL: 13.25%, bias-RL:13.25% of participants), suggesting fundamental differences in the way individuals deploy basic learning mechanisms, even for a simple stimulus-response task. Finally, long-term declarative memory seems to be the preferred learning strategy for this task regardless of block length (3- vs 6-image blocks), as determined by the large number of subjects whose learning characteristics were best captured by the LTM only model, and a preference for LTM over RL in both of our integrated-models, owing to the strength of our idiographic approach.

Two Separate Brain Networks for Predicting Trainability and Tracking Training-Related Plasticity in Working Dogs.

Functional brain connectivity based on resting-state functional magnetic resonance imaging (fMRI) has been shown to be correlated with human personality and behavior. In this study, we sought to know whether capabilities and traits in dogs can be predicted from their resting-state connectivity, as in humans. We trained awake dogs to keep their head still inside a 3T MRI scanner while resting-state fMRI data was acquired. Canine behavior was characterized by an integrated behavioral score capturing their hunting, retrieving, and environmental soundness. Functional scans and behavioral measures were acquired at three different time points across detector dog training. The first time point (TP1) was prior to the dogs entering formal working detector dog training. The second time point (TP2) was soon after formal detector dog training. The third time point (TP3) was three months' post detector dog training while the dogs were engaged in a program of maintenance training for detection work. We hypothesized that the correlation between resting-state FC in the dog brain and behavior measures would significantly change during their detection training process (from TP1 to TP2) and would maintain for the subsequent several months of detection work (from TP2 to TP3). To further study the resting-state FC features that can predict the success of training, dogs at TP1 were divided into a successful group and a non-successful group. We observed a core brain network which showed relatively stable (with respect to time) patterns of interaction that were significantly stronger in successful detector dogs compared to failures and whose connectivity strength at the first time point predicted whether a given dog was eventually successful in becoming a detector dog. A second ontologically based flexible peripheral network was observed whose changes in connectivity strength with detection training tracked corresponding changes in behavior over the training program. Comparing dog and human brains, the functional connectivity between the brain stem and the frontal cortex in dogs corresponded to that between the locus coeruleus and left middle frontal gyrus in humans, suggestive of a shared mechanism for learning and retrieval of odors. Overall, the findings point toward the influence of phylogeny and ontogeny in dogs producing two dissociable functional neural networks.

Phonetic perceptual reorganization across the first year of life: Looking back

This paper provides a selective overview of some of the research that has followed from the publication of Werker and Tees (1984a) “Cross-language speech perception: Evidence for Perceptual Reorganization During the First Year of Life.” Specifically, I briefly present the original finding, our interpretation of its meaning, and some key replications and extensions. I then review some of the work that has followed, including work with different kinds of populations, different kinds of speech sound contrasts, as well as attunement (perceptual reorganization) to additional properties of language beyond phonetic contrasts. Included is the body of work that queries whether perceptual attunement is a critical period phenomenon. Potential learning mechanisms for how experience functions to guide phonetic perceptual development are also presented, as is work on the relation between speech perception and word learning.