长时记忆中图形不同特征的加工与提取

Information can be represented both conceptually and imaginarily in long-term memory. However, it seems that only conceptual representation appears, neglecting imaginary information, in most of the long-term memory (LTM) models. In the matter of fact, picture can be stored in LTM directly and conceptually. There is no evidence for what specific type of information, conceptual or imaginary, for the color, shape, or texture to be represented. However, it is evident that the shape and color can be represented separately in LMT. Further research is needed on whether features are represented separately or not, such as color and texture, texture and shape etc. Rehearsal plays important role in picture memory besides the types of storage and representation. Memory of picture is indeed enhanced by rehearsal. There are two types of rehearsal. One is for creating image, another is articulatory loop. Which one will be taken during picture memory process depends on the characteristics of stimuli, subjects' encoding preferences and/or task requirements. Nevertheless, the relation between two types of rehearsal is not very clear yet up to now. Different features could be activated at different time course or possibilities since they can be represented separately. Six experiments were conducted dealing with the characteristics of representation, rehearsal and retrieval of picture in LTM. From these experiments, further understanding of picture information processing was expected. It would add more evidence to the LTM models, and make practical sense to the computer visual identification. The first two experiments were based on the paradigm from Hanna et al.(1996) to investigate separable representation of texture and shape, texture and color. The results indicated that texture could be represented separately with color and shape respectively. It suggested that different features might be processed in different way during remembering. Another interest finding is that recognition performance for shape, color and texture are quite different. What for shape is highest, for color is lowest, and for texture is between of them. Three features of picture can be represented separately. How about the roles of rehearsal when they enter the LTM from short-term memory(STM)? The second three experiments assigned three different types of rehearsal, i. e. visual, verbal, and subject-run(might be both of visual and verbal). The findings are that performances of picture memory were affected significantly by different types of rehearsal. Both visual and verbal rehearsal played important role during remembering process. It seems that verbal rehearsal, which might enhance the relative strength of memory trace, was much more effective than visual one. In addition, subjects tended to choose those difficult-to-name, features to rehearse, to improve the memory performance. Only two features were changed in each of the first two experiments. They might interact (facilitate or disturb) each other when they were retrieved. So it was difficult to identify the retrieval difference between them. In the last experiment, easy-to-name pictures were studied, and only one feature could be recognized. The results indicated that the retrieval performances of three features(shape, color, and texture) were quite different. They were different on the relative strength of memory trace, with the shape was strongest, color was lightest, and texture was in between. No difference was found on the absolute strength of them.

Variable Rate Working Memories for Phonetic Categorization and Invariant Speech Perception

Speech can be understood at widely varying production rates. A working memory is described for short-term storage of temporal lists of input items. The working memory is a cooperative-competitive neural network that automatically adjusts its integration rate, or gain, to generate a short-term memory code for a list that is independent of item presentation rate. Such an invariant working memory model is used to simulate data of Repp (1980) concerning the changes of phonetic category boundaries as a function of their presentation rate. Thus the variability of categorical boundaries can be traced to the temporal in variance of the working memory code.

Self-Organizing Grammar Induction Using a Neural Network Model

This paper presents a self-organizing, real-time, hierarchical neural network model of sequential processing, and shows how it can be used to induce recognition codes corresponding to word categories and elementary grammatical structures. The model, first introduced in Mannes (1992), learns to recognize, store, and recall sequences of unitized patterns in a stable manner, either using short-term memory alone, or using long-term memory weights. Memory capacity is only limited by the number of nodes provided. Sequences are mapped to unitized patterns, making the model suitable for hierarchical operation. By using multiple modules arranged in a hierarchy and a simple mapping between output of lower levels and the input of higher levels, the induction of codes representing word category and simple phrase structures is an emergent property of the model. Simulation results are reported to illustrate this behavior.

An Integrated Neural Network-Event-Related Potentials Model of Temporal and Probability Context Effects on Event Categorization

We present a neural network that adapts and integrates several preexisting or new modules to categorize events in short term memory (STM), encode temporal order in working memory, evaluate timing and probability context in medium and long term memory. The model shows how processed contextual information modulates event recognition and categorization, focal attention and incentive motivation. The model is based on a compendium of Event Related Potentials (ERPs) and behavioral results either collected by the authors or compiled from the classical ERP literature. Its hallmark is, at the functional level, the interplay of memory registers endowed with widely different dynamical ranges, and at the structural level, the attempt to relate the different modules to known anatomical structures.

A Neural Network Model of Spatio-Temporal Pattern Recognition, Recall and Timing

This paper describes the design of a self~organizing, hierarchical neural network model of unsupervised serial learning. The model learns to recognize, store, and recall sequences of unitized patterns, using either short-term memory (STM) or both STM and long-term memory (LTM) mechanisms. Timing information is learned and recall {both from STM and from LTM) is performed with a learned rhythmical structure. The network, bearing similarities with ART (Carpenter & Grossberg 1987a), learns to map temporal sequences to unitized patterns, which makes it suitable for hierarchical operation. It is therefore capable of self-organizing codes for sequences of sequences. The capacity is only limited by the number of nodes provided. Selected simulation results are reported to illustrate system properties.

On-line Emotion Recognition in a 3-D Activation-Valence-Time Continuum using Acoustic and Linguistic Cues

For many applications of emotion recognition, such as virtual agents, the system must select responses while the user is speaking. This requires reliable on-line recognition of the user’s affect. However most emotion recognition systems are based on turnwise processing. We present a novel approach to on-line emotion recognition from speech using Long Short-Term Memory Recurrent Neural Networks. Emotion is recognised frame-wise in a two-dimensional valence-activation continuum. In contrast to current state-of-the-art approaches, recognition is performed on low-level signal frames, similar to those used for speech recognition. No statistical functionals are applied to low-level feature contours. Framing at a higher level is therefore unnecessary and regression outputs can be produced in real-time for every low-level input frame. We also investigate the benefits of including linguistic features on the signal frame level obtained by a keyword spotter.

Altered Long-Range Phase Synchronization and Cortical Activation in Children Born Very Preterm

Children born very preterm, even with broadly normal IQ, commonly show selective difficulties in visuospatial processing and executive functioning. Very little, however, is known what alterations in cortical processing underlie these deficits. We recorded MEG while eight children born very preterm (=32 weeks gestational age) and eight full-term controls performed a visual short-term memory task at mean age 7.5 years (range 6.4 - 8.4). Previously, we demonstrated increased long-range alpha and beta band phase synchronization between MEG sensors during STM retention in a group of 17 full-term children age 6-10 years. Here we present preliminary evidence that long-range phase synchronization in very preterm children, relative to controls, is reduced in the alpha-band but increased in the theta-band. In addition, we investigated cortical activation during STM retention employing synthetic aperture magnetometry (SAM) beamformer to localize changes in gamma-band power. Preliminary results indicate sequential activation of occipital, parietal and frontal cortex in control children, as well as reduced activation in very preterm children relative to controls. These preliminary results suggest that children born very preterm exhibit altered inter-regional functional connectivity and cortical activation during cognitive processing.

Applying gaze-contingent training within community settings to infants from diverse SES backgrounds

Even in infancy children from low-SES backgrounds differ in frontal cortex functioning and, by the start of preschool, they frequently show poor performance on executive functions including attention control. These differences may causally mediate later difficulties in academic learning. Here, we present a study to assess the feasibility of using computerized paradigms to train attention control in infants, delivered weekly over five sessions in early intervention centres for low-SES families. Thirty-three 12-month-old infants were recruited, of whom 23 completed the training. Our results showed the feasibility of repeat-visit cognitive training within community settings. Training-related improvements were found, relative to active controls, on tasks assessing visual sustained attention, saccadic reaction time, and rule learning, whereas trend improvements were found on assessments of short-term memory. No significant improvements were found in task switching. These results warrant further investigation into the potential of this method for targeting ‘at-risk’ infants in community settings.

Are Deep Learning Approaches Suitable for Natural Language Processing?

In recent years, Deep Learning (DL) techniques have gained much at-tention from Artificial Intelligence (AI) and Natural Language Processing (NLP) research communities because these approaches can often learn features from data without the need for human design or engineering interventions. In addition, DL approaches have achieved some remarkable results. In this paper, we have surveyed major recent contributions that use DL techniques for NLP tasks. All these reviewed topics have been limited to show contributions to text understand-ing, such as sentence modelling, sentiment classification, semantic role labelling, question answering, etc. We provide an overview of deep learning architectures based on Artificial Neural Networks (ANNs), Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM), and Recursive Neural Networks (RNNs).

The four- vs. alternative six-factor structure of the French WISC-IV: Comparison using confirmatory factor analyses

Exploratory and confirmatory factor analyses reported in the French technical manual of the WISC-IV provides evidence supporting a structure with four indices: Verbal Comprehension (VCI), Perceptual Reasoning (PRI), Working Memory (WMI), and Processing Speed (PSI). Although the WISC-IV is more attuned to contemporary theory, it is still not in total accordance with the dominant theory: the Cattell-Horn-Carroll (CHC) theory of cognitive ability. This study was designed to determine whether the French WISC-IV is better described with the four-factor solution or whether an alternative model based on the CHC theory is more appropriate. The intercorrelations matrix reported in the French technical manual was submitted to confirmatory factor analysis. A comparison of competing models suggests that a model based on the CHC theory fits the data better than the current WISC-IV structure. It appears that the French WISC-IV in fact measures six factors: crystallized intelligence (Gc), fluid intelligence (Gf), short-term memory (Gsm), processing speed (Gs), quantitative knowledge (Gq), and visual processing (Gv). We recommend that clinicians interpret the subtests of the French WISC-IV in relation to this CHC model in addition to the four indices.

A comparison of cognitive processes and components of neurological maturation in 3-and-11 year-old children

The relationship between the child's cogni tive development and neurological maturation has been of theoretical interest for many year s. Due to diff iculties such as the lack of sophisticated techniques for measur ing neurolog ical changes and a paucity of normative data, few studies exist that have attempted to correlate the two factors. Recent theory on intellectual development has proposed that neurological maturation may be a factor in the increase of short-term memory storage space. Improved technology has allowed reliable recordings of neurolog ical maturation.. In an attempt to correlate cogni tive development and neurological maturation, this study tested 3-and II-year old children. Fine motor and gross motor short-term memory tests were used to index cogni tive development. Somatosensory evoked potentials elici ted by median nerve stimulation were used to measure the time required for the sensation to pass along the nerve to specific points on the somatosensory pathway. Times were recorded for N14, N20, and P22 interpeak latencies. Maturation of the central nervous system (brain and spinal cord) and the peripheral nervous system (outside the brain and spinal cord) was indi~ated by the recorded times. Signif icant developmental di fferences occurred between 3-and ll-year-olds in memory levels, per ipheral conduction velocity and central conduction times. Linear regression analyses showed that as age increased, memory levels increased and central conduction times decreased. Between the ll-year-old groups, there were no significant differences in central or peripheral nervous system maturation between subjects who achieved a 12 plus score on the digit span test of the WISC-R and those who scored 7 or lower on the same test. Levels achieved on the experimental gross and fine motor short-term memory tests differed significantly within the ll-year-old group.

The performance and strategies of adolescents in learning to read artificial words /

The present study explored processing strategies used by individuals when they begin to read c;l script. Stimuli were artificial words created from symbols and based on an alphabetic system. The words were.presented to Grade Nine and Ten students, with variations included in the difficulty of orthography and word familiarity, and then scores were recorded on the mean number of trials for defined learning variables. Qualitative findings revealed that subjects 1 earned parts of the visual a'nd auditory features of words prior to hooking up the visual stimulus to the word's name. Performance measures-which appear to affect the rate of learning were as follows: auditory short-term memory, auditory delayed short-term memory, visual delayed short- term memory, and word attack or decod~ng skills. Qualitative data emerging in verbal reports by the subjects revealed that strategies they pefceived to use were, graphic, phonetic decoding and word .reading.

L’approche psycholinguistique de la mémoire à court terme verbale : études neuropsychologiques

L’approche psycholinguistique suggère que la rétention à court terme verbale et le langage dépendent de mécanismes communs. Elle prédit que les caractéristiques linguistiques des items verbaux (e.g. phonologiques, lexicales, sémantiques) influencent le rappel immédiat (1) et que la contribution des niveaux de représentations linguistiques dépend du contexte de rappel, certaines conditions expérimentales (e.g. format des stimuli) favorisant l’utilisation de codes spécifiques (2). Ces prédictions sont évaluées par le biais de deux études empiriques réalisées auprès d’une patiente cérébrolésée qui présente une atteinte du traitement phonologique (I.R.) et de participants contrôles. Une première étude (Article 1) teste l’impact des modes de présentation et de rappel sur les effets de similarité phonologique et de catégorie sémantique de listes de mots. Une seconde étude (Article 2) évalue la contribution du code orthographique en mémoire à court terme (MCT) verbale en testant l’effet de la densité du voisinage orthographique des mots sur le rappel sériel immédiat de mots présentés visuellement. Compte tenu du rôle déterminant du code phonologique en MCT et du type d’atteinte de I.R., des effets linguistiques distincts étaient attendus chez elle et chez les contrôles. Selon le contexte de rappel, des effets sémantiques (Article 1) et orthographiques (Article 2) plus importants étaient prédits chez I.R. et des effets phonologiques plus marqués étaient attendus chez les participants contrôles. Chez I.R., le rappel est influencé par les caractéristiques sémantiques et orthographiques des mots, mais peu par leurs caractéristiques phonologiques et le contexte de rappel module l’utilisation de différents niveaux de représentations linguistiques. Chez les contrôles, une contribution relativement plus stable des représentations phonologiques est observée. Les données appuient une approche psycholinguistique qui postule que des mécanismes communs régissent la rétention à court terme verbale et le langage. Les implications théoriques et cliniques des résultats sont discutées en regard de modèles psycholinguistiques actuels.

Représentation corticale de la mémoire à court-terme tactile chez l'humain pour une stimulation de la main : étude par magnétoencéphalographie

L'activité cérébrale, reliée spécifiquement à la rétention d'information en mémoire à court-terme tactile, a été investiguée à l'aide de l'enregistrement des champs magnétiques produits par l'activité neuronale générée durant la période de rétention par une tâche de mémoire tactile. Une, deux, trois ou quatre positions, sur une possibilité de huit sur les phalangines et les phalangettes, de la main gauche ou droite, lors de blocs d'essai différents, ont été stimulées simultanément. Le patron de stimulation tactile devait être retenu pendant 1800 ms avant d'être comparé avec un patron test qui était, soit identique, soit différent par une seule position. Nos analyses se sont concentrées sur les régions du cerveau qui montraient une augmentation monotone du niveau d'activité soutenu durant la période de rétention pour un nombre croissant de positions à retenir dans le patron de stimulation. Ces régions ont plus de chance de participer à la rétention active de l'information à maintenir en mémoire à court-terme tactile. Le gyrus cingulaire (BA32), le gyrus frontal supérieur droit (BA 8), le precuneus gauche (BA 7, 19), le gyrus postcentral gauche (BA 7), le gyrus precentral droit (BA 6), le gyrus frontal supérieur gauche (BA 6) et le lobule pariétal inférieur droit (BA 40) semblent tous impliqués dans un réseau mnésique qui maintient les informations sensorielles tactiles dans un système de mémoire à court-terme spécialisé pour l'information tactile.

Approche psycholinguistique des relations entre le traitement sémantique des mots et la rétention à court terme chez l’individu sain et dans la démence de type Alzheimer

Depuis ces deux dernières décennies, des efforts considérables en psychologie cognitive et neuropsychologie ont été déployés pour mieux comprendre les dynamiques entre les différents systèmes cognitifs (fonctions exécutives, mémoire, langage, etc.). L’observation d’une co-existence entre des atteintes en rétention à court terme verbale (RCTv) et en langage dans plusieurs pathologies a conduit certains chercheurs à concevoir une alternative à la conceptualisation classique de la RCTv, telle que proposée par le modèle initial de la mémoire de travail de Baddeley (1986). Particulièrement, les travaux sur l’aphasie ont conduit à l’émergence d’une approche psycholinguistique de la RCTv, postulant que le traitement et le maintien des mots font appel à des processus communs. Cette approche, et particulièrement le modèle d’activation interactive (N. Martin & Gupta, 2004; N. Martin & Saffran, 1997), prédit que les capacités en RCTv sont notamment modulées par les caractéristiques linguistiques engagées durant l’épreuve, variant selon la nature des items verbaux et la modalité de rappel, ainsi que par les compétences linguistiques des individus. L’objectif de la présente thèse était de tester ces prédictions à partir d’une exploration des relations entre le traitement sémantique des mots et la RCTv chez l’adulte sain (article 1) et dans la démence de type Alzheimer (DTA) (article 2). Dans le premier article, deux expériences mettent en évidence l’influence des liens associatifs entre les mots sur les capacités à maintenir temporairement des informations verbales. Les participants ont tendance à faussement reconnaître des mots associés aux mots de la liste, reflet d’une activation du réseau sémantique durant la rétention temporaire. Cette contribution sémantique est accentuée en situation de suppression articulatoire, une condition qui empêche le participant de répéter les listes durant la tâche. Les résultats du second article indiquent que la modalité de réponse module différemment la performance en RCTv chez les participants âgés sans atteinte neurologique et ceux atteints de la DTA. Ces données en RCTv sont compatibles avec les atteintes spécifiques du traitement du mot, également trouvées chez le groupe avec DTA. Les implications théoriques et cliniques de ces résultats sont discutées. Les limites et perspectives futures sont également abordées.