The procedural deficit hypothesis (PDH) posits that a range of language, cognitive and motor impairments associated with specific language impairment (SLI) and developmental dyslexia (DD) may be explained by an underlying domain-general dysfunction of the procedural memory system. In contrast, declarative memory is hypothesized to remain intact and to play a compensatory role in the two disorders. The studies in the present thesis were designed to test this hypothesis. Study I examined non-language procedural memory, specifically implicit sequence learning, in children with SLI. It was shown that children with poor performance on tests of grammar were impaired at consolidation of procedural memory compared to children with normal grammar. These findings support the PDH and are line with previous studies suggesting a link between grammar processing and procedural memory. In Study II, the same implicit sequence learning paradigm was used to test procedural memory in children with DD. The DD group showed a learning profile that was similar to that of children with SLI in Study I, with a significant impairment emerging late in learning, after extended practice and including an overnight interval. Further analyses suggested that the DD impairment may not be related to overnight consolidation but to the effects of further practice beyond the initial practice session. In contrast to the predictions of the PDH, the sequence learning deficit was unrelated to phonological processing skills as assessed with a nonword repetition task. Study III examined declarative memory in DD. The performance of the DD group was found to be not only intact, but even enhanced, compared to that of the control children. The results encourage further studies on the potential of declarative memory to compensate for the reading problems in DD. In sum, the results lend partial support for the PDH and suggest further refinements to the theory. Collectively, the studies emphasize the importance of going beyond a narrow focus on language learning and memory functions in the characterization of the two disorders. Such a broader cognitive, motor and language approach may inform the development of future clinical and pedagogical assessment and intervention practices for SLI and DD.

A hierarchical symbiotic framework for policy search with genetic programming (GP) is evaluated in two control-style temporal sequence learning domains. The symbiotic formulation assumes each policy takes the form of a cooperative team between multiple symbiont programs. An initial cycle of evolution establishes a diverse range of host behaviours with limited capability. The second cycle uses these initial policies as meta actions for reuse by symbiont programs. The relationship between development and ecology is explored by explicitly altering the interaction between learning agent and environment at fixed points throughout evolution. In both task domains, this developmental diversity significantly improves performance. Specifically, ecologies designed to promote good specialists in the first developmental phase and then good generalists result in much stronger organisms from the perspective of generalization ability and efficiency. Conversely, when there is no diversity in the interaction between task environment and policy learner, the resulting hierarchy is not as robust or general. The relative contribution from each cycle of evolution in the resulting hierarchical policies is measured from the perspective of multi-level selection. These multi-level policies are shown to be significantly better than the sum of contributing meta actions.

For nearly two decades researchers have been interested in identifying what specifically is learned when individuals learn a sequence (e.g., sequence of stimuli, sequence of motor movements, etc.). Despite extensive research in the area, considerable controversy remains surrounding the locus of learning. There are three main theories concerning the nature of spatial sequence learning: sequence learning is purely perceptual, sequence learning includes a motor component and sequence learning is based on stimulus-response (S-R) rules. The present studies sought to disentangle these theories by demonstrating that sequence learning has both a perceptual and motor component and that altering S-R rules alone disrupts sequence learning. Experiment 1 results fully supported this S-R rule theory of sequence learning. Experiment 2 results provided only partial support for this theory, though the data were also inconsistent with both of the other accounts.

Procedural learning abilities have been shown to be deficient in children who meet criteria for Developmental Dyslexia (DD) and those who meet criteria for Specific Language Impairment (SLI; Lum et al., 2010; Menghini et al., 2006). Further, grammatical understanding has been linked to implicit sequence learning abilities across SLI and typically developing children (Lum, 2012). The present study examined implicit sequence learning, measured by the Serial Reaction Time Task (SRTT), in children who met criteria for DD with or without SLI. Implicit sequence learning was modeled using multi-level growth models of initial reaction time and learning slope across the repeated sequences of the SRTT. We further examined the predictive contributions of grammatical understanding, vocabulary abilities, phonological awareness, and diagnostic groups on implicit learning performance on the SRTT. Results showed language abilities and diagnostic group did not relate strongly to rates of implicit learning.

Chez les humains et les animaux, l’exposition répétée à une séquence de stimulus conduit à la création de représentations mentales isomorphes aux régularités statistiques de cette séquence. Cette thèse étudie la dynamique avec laquelle les unités et sous-unités sont extraites, s’influencent et s’organisent lors de l’apprentissage séquentiel chez l’humain et le primate non-humain (babouin Papio papio). Nos résultats montrent que lors de l’apprentissage d’une séquence régulière ABC, la sous-unité finale BC est apprise plus rapidement que la sous-unité initiale AB chez les humains comme les primates non- humains. Avec une séquence plus longue ABCD, cet effet ne s’étend pas au-delà des deux termes précédents celui à prédire lors des 2000 essais d’apprentissage. Une seconde étude montre que les humains sont capables d’utiliser la prédictibilité des éléments à différents niveaux de complexité, alors que les singes sont limités aux dépendances locales. Enfin notre dernière étude montre que les humains comme les singes ne bénéficient pas d’un contexte d’homogénéité de longueurs des unités par rapport à un contexte d’hétérogénéité de longueurs des unités. Ensemble, ces études montrent des continuités et discontinuités entre les primates humains et non-humains dans l’extraction, l’influence et l’organisation des unités de différents niveaux. Elles donnent des informations d’une finesse nécessaires pour contraindre les différents modèles computationnels qui ont été proposés pour décrire l’apprentissage de régularités. / In humans and animals, the repeated exposure to a sequence of stimulus creates mental representations which have the same statistical regularities as that sequence. This thesis investigates the dynamics, influence and organization of units and sub-units during sequence learning, in humans and non-human primates (Guinea baboons, Papio papio).Our results show an advantage of the final sub- unit BC over the initial sub-unit AB during learning of sequences with the form ABC in both humans and baboons. We interpret this effect as resulting from the richer contextual information AB predicting C compared to the single term A predicting B. This effect is limited to the 2 previous terms before the one term to be predicted, across 2000 learning trials. A second finding shows that in learning sequences of 3 units (i.e. ABC, DEF, HGI), humans are able to use the predictability of elements at both local (within units) and global (within sequence) levels, whereas monkeys are limited to local dependencies. A third section investigates the extraction of regularities in homogeneous (i.e., units of the same length) and heterogeneous contexts (units of different lengths). Results in humans and monkeys showed no advantage for either of these contexts.Considered together, our studies show continuities and discontinuities between human and non-human primates in the extraction, influence and organization of units of different levels. This fine-grained level of information is necessary to constrain the computational models proposed to describe the mechanisms underlying regularity extraction.

This project focuses on semi-supervised human activity recognition (HAR) tasks, in which the inputs are partly labeled time series data acquired from sensors such as accelerometer data, and the outputs are predefined human activities. Most state-of-the-art existing work in HAR area is supervised now, which relies on fully labeled datasets. Since the cost to label the collective instances increases fast with the increasing scale of data, semi-supervised methods are now widely required. This report proposed two semi-supervised methods and then investigated how well they perform on a partly labeled dataset, comparing to the state-of-the-art supervised method. One of these methods is designed based on the state-of-the-art supervised method, Deep-ConvLSTM, together with the semi-supervised learning concepts, self-training. Another one is modified based on a semi-supervised deep learning method, LSTM initialized by seq2seq autoencoder, which is firstly introduced for natural language processing. According to the experiments on a published dataset (Opportunity Activity Recognition dataset), both of these semi-supervised methods have better performance than the state-of-the-art supervised methods. / Detta projekt fokuserar på halvövervakad Human Activity Recognition (HAR), där indata delvis är märkta tidsseriedata från sensorer som t.ex. accelerometrar, och utdata är fördefinierade mänskliga aktiviteter. De främsta arbetena inom HAR-området använder numera övervakade metoder, vilka bygger på fullt märkta dataset. Eftersom kostnaden för att märka de samlade instanserna ökar snabbt med den ökade omfattningen av data, föredras numera ofta halvövervakade metoder. I denna rapport föreslås två halvövervakade metoder och det undersöks hur bra de presterar på ett delvis märkt dataset jämfört med den moderna övervakade metoden. En av dessa metoder utformas baserat på en högkvalitativ övervakad metod, DeepConvLSTM, kombinerad med självutbildning. En annan metod baseras på en halvövervakad djupinlärningsmetod, LSTM, initierad av seq2seq autoencoder, som först införs för behandling av naturligt språk. Enligt experimenten på ett publicerat dataset (Opportunity Activity Recognition dataset) har båda dessa metoder bättre prestanda än de toppmoderna övervakade metoderna.

yes / We investigated the role of the human right Supra-Marginal Gyrus (SMG) in the generation of learned eye movement sequences. Using MRI-guided transcranial magnetic stimulation (TMS) we disrupted neural activity in the SMG whilst human observers performed saccadic eye movements to multiple presentations of either predictable or random target sequences. For the predictable sequences we observed shorter saccadic latencies from the second presentation of the sequence. However, these anticipatory improvements in performance were significantly reduced when TMS was delivered to the right SMG during the inter-trial retention periods. No deficits were induced when TMS was delivered concurrently with the onset of the target visual stimuli. For the random version of the task, neither delivery of TMS to the SMG during the inter-trial period nor during the presentation of the target visual stimuli produced any deficit in performance that was significantly different from the no-TMS or control conditions. These findings demonstrate that neural activity within the right SMG is causally linked to the ability to perform short latency predictive saccades resulting from sequence learning. We conclude that neural activity in rSMG constitutes an instruction set with spatial and temporal directives that are retained and subsequently released for predictive motor planning and responses.

The mammalian visual system is an incredibly complex computation device, capable of performing the various tasks of seeing: navigation, pattern and object recognition, motor coordination, trajectory extrapolation, among others. Decades of research has shown that experience-dependent plasticity of cortical circuitry underlies the impressive ability to rapidly learn many of these tasks and to adjust as required. One particular thread of investigation has focused on unsupervised learning, wherein changes to the visual environment lead to corresponding changes in cortical circuits. The most prominent example of unsupervised learning is ocular dominance plasticity, caused by visual deprivation to one eye and leading to a dramatic re-wiring of cortex. Other examples tend to make more subtle changes to the visual environment through passive exposure to novel visual stimuli. Here, we use one such unsupervised paradigm, sequence learning, to study experience-dependent plasticity in the mouse visual system. Through a combination of theory and experiment, we argue that the mammalian visual system is an unsupervised learning device. Beginning with a mathematical exploration of unsupervised learning in biology, engineering, and machine learning, we seek a more precise expression of our fundamental hypothesis. We draw connections between information theory, efficient coding, and common unsupervised learning algorithms such as Hebbian plasticity and principal component analysis. Efficient coding suggests a simple rule for transmitting information in the nervous system: use more spikes to encode unexpected information, and fewer spikes to encode expected information. Therefore, expectation violations ought to produce prediction errors, or brief periods of heightened firing when an unexpected event occurs. Meanwhile, modern unsupervised learning algorithms show how such expectations can be learned. Next, we review data from decades of visual neuroscience research, highlighting the computational principles and synaptic plasticity processes that support biological learning and seeing. By tracking the flow of visual information from the retina to thalamus and primary visual cortex, we discuss how the principle of efficient coding is evident in neural activity. One common example is predictive coding in the retina, where ganglion cells with canonical center-surround receptive fields compute a prediction error, sending spikes to the central nervous system only in response to locally-unpredictable visual stimuli. This behavior can be learned through simple Hebbian plasticity mechanisms. Similar models explain much of the activity of neurons in primary visual cortex, but we also discuss ways in which the theory fails to capture the rich biological complexity. Finally, we present novel experimental results from physiological investigations of the mouse primary visual cortex. We trained mice by passively exposing them to complex spatiotemporal patterns of light: rapidly-flashed sequences of images. We find evidence that visual cortex learns these sequences in a manner consistent with efficient coding, such that unexpected stimuli tend to elicit more firing than expected ones. Overall, we observe dramatic changes in evoked neural activity across days of passive exposure. Neural responses to the first, unexpected sequence element increase with days of training while responses at other, expected time points either decrease or stay the same. Furthermore, substituting an unexpected element for an expected one or omitting an expected element both cause brief bursts of increased firing. Our results therefore provide evidence for unsupervised learning and efficient coding in the mouse visual system, especially because unexpected events drive prediction errors. Overall, our analysis suggests novel experiments, which could be performed in the near future, and provides a useful framework to understand visual perception and learning.

No description available.

No description available.