Spike-timing dependent plasticity is a learning mechanism used extensively within neural modelling. The learning rule has been shown to allow a neuron to find the onset of a spatio-temporal pattern repeated among its afferents. In this thesis, the first question addressed is ‘what does this neuron learn?’ With a spiking neuron model and linear prediction, evidence is adduced that the neuron learns two components: (1) the level of average background activity and (2) specific spike times of a pattern. Taking advantage of these findings, a network is developed that can train recognisers for longer spatio-temporal input signals using spike-timing dependent plasticity. Using a number of neurons that are mutually connected by plastic synapses and subject to a global winner-takes-all mechanism, chains of neurons can form where each neuron is selective to a different segment of a repeating input pattern, and the neurons are feedforwardly connected in such a way that both the correct stimulus and the firing of the previous neurons are required in order to activate the next neuron in the chain. This is akin to a simple class of finite state automata. Following this, a novel resource-based STDP learning rule is introduced. The learning rule has several advantages over typical implementations of STDP and results in synaptic statistics which match favourably with those observed experimentally. For example, synaptic weight distributions and the presence of silent synapses match experimental data.

Les preuves par récurrence sont parfaitement adaptées au raisonnement sur des structures de données non-bornées, comme par exemple les entiers et les listes, ou, de manière plus générale, sur des ensembles d’éléments non vides munis d’ordres noethériens. Leur domaine d’application est très vaste, une utilité particulière portant sur la validation des propriétés d’applications industrielles dans des domaines critiques tels que les télécommunications et les cartes à puces. Le principe de récurrence noethérienne est à la base d’un ensemble de techniques de preuve par récurrence modernes, dont celles basées sur la récurrence implicite. Dans cette thèse, nous nous intéresserons à l’intégration du raisonnement par récurrence implicite tel qu’il est implémenté dans le démonstrateur Spike en utilisant l’environnement de preuve certifié Coq. Basé sur la récurrence implicite, Spike est capable de raisonner automatiquement sur des théories conditionnelles de premier ordre. L’implémentation de Spike n’est pas encore certifiée, même si les fondements théoriques sous-jacents ont été approuvés à plusieurs reprises par la communauté scientifique. Une alternative convenable serait de certifier seulement les preuves générées par Spike. Dans ce cas, le processus de certification doit être automatique car les scripts de preuves de Spike sont souvent longs. Des travaux précédents ont montré la possibilité de certifier automatiquement des preuves par récurrence implicite générées par Spike à l’aide de l’environnement certifié de l’assistant de preuve Coq. Nous proposerons des nouvelles tactiques Coq qui seront capables de prouver automatiquement des théorèmes par récurrence implicite. Deux approches seront étudiées. La première approche consiste à utiliser Spike comme un outil externe. Elle est limitée au traitement des spécifications Coq qui peuvent être traduites dans des spécifications conditionnelles, ainsi qu’à des théorèmes convertibles dans des équations conditionnelles. Les traces de preuves générées par Spike sont ensuite traduites dans des scripts Coq qui sont finalement validés par son noyau. Une autre limitation est due à la traduction des applications d’un sous-ensemble de règles d’inférence de Spike. La deuxième approche est l’utilisation des stratégies à la Spike pour construire automatiquement des preuves par récurrence implicite dans Coq. Cette approche se base sur des tactiques Coq qui simulent des règles d’inférence de Spike pour générer de nouveaux sous-buts. Par rapport à la première approche, ces tactiques peuvent utiliser des techniques de raisonnement de Coq qui ne sont pas présentes dans Spike et ouvre la possibilité de mélanger des étapes de preuves automatiques et manuelles. Ces deux approches ont été mises en œuvre et testées sur différents exemples dont des lemmes utilisés dans la preuve de validité de l’algorithme de conformité du protocole de télécommunication ABR / Proofs by induction are perfectly adequate to reasoning on unbounded data structures, for example naturals, lists and more generally on non-empty sets of elements provided with noetherian orders. They are largely used on different fields, particularly for the validation of properties of industrial applications in critical areas such as telecommunications and smart cards. The principle of noetherian induction is the basis of a set of modern techniques of proof by induction, including those based on implicit induction. In this thesis, we will focus on the integration of implicit induction reasoning like it is implemented by spike using the certified proof environnement Coq. Spike is an automatic theorem prover based on implicit induction that is capable of reasoning on conditional first-order theories. The implementation of Spike is not yet certified, even if the underlying theoretical foundations have been approved repeatedly by the scientific community. A suitable alternative is to certify only the proofs produced by Spike. In this case, the certification process must be automatic because scripts of Spike’s proofs are often long. Previous work has shown the possibility of certifying automatically some proofs by implicit induction generated by Spike using the certified environment provided by the Coq proof-assistant. We will propose new Coq tactics that are able to prove automatically theorems by implicit induction. Two approaches will be studied. The first approach consists on using Spike as an external tool. It is limited to process Coq specifications which can be translated in conditional specifications, as well as theorems convertible in conditional equations. Proofs generated by Spike are then translated into Coq scripts finally validated by its kernel. Another limitation is due to the translation of the application of a subset of the Spike inference rules. The second approche is to use strategies à la Spike to automatically build implicit induction proofs in Coq. This approach consists on creating tactics that perform like Spike inference rules to generate new subgoals in Coq. Comparing to the first approach, these tactics permit the use of Coq reasoning techniques which are not present in Spike and opens the possibility of mixing automatic and manual proof steps. Both approaches have been implemented and tested on several examples including lemmas used in the proof of validity of the conformity algorithm for the ABR telecommunications protocol

Spiking neural networks have, in recent years, become a popular tool for investigating the properties and computational performance of large massively connected networks of neurons. Equally as interesting is the investigation of the potential computational power of individual spiking neurons. An overview is provided of current and relevant research into the Liquid Sate Machine, biologically inspired artificial STDP learning mechanisms and the investigation of aspects of the computational power of artificial, recurrent networks of spiking neurons. First, it is shown that, using simple structures of spiking Leaky Integrate and Fire (LIF) neurons, a network n(P), can be built to perform any program P that can be performed by a general parallel programming language. Next, a form of STDP learning with normalisation is developed, referred to as STDP + N learning. The effects of applying this STDP + N learning within recurrently connected networks of neurons is then investigated. It is shown experimentally that, in very specific circumstances Anti-Hebbian and Hebbian STDP learning may be considered to be approximately equivalent processes. A metric is then developed that can be used to measure the distance between any two spike trains. The metric is then used, along with the STDP + N learning, in an experiment to examine the capacity of a single spiking neuron that receives multiple input spike trains, to simultaneously learn many temporally precise Input/Output spike train associations. The STDP +N learning is further modified for use in recurrent networks of spiking neurons, to give the STDP + NType2 learning methodology. An experiment is devised which demonstrates that the Type 2 method of applying learning to the synapses of a recurrent network — effectively a randomly shifting locality of learning — can enable the network to learn firing patterns that the typical application of learning is unable to learn. The resulting networks could, in theory, be used to create to simple structures discussed in the first chapter of original work.

In hippocampal CA1 and CA3 regions, various properties of neuronal activity follow skewed, lognormal-like distributions, including average firing rates, rate and magnitude of spike bursts, magnitude of population synchrony, and correlations between pre- and postsynaptic spikes. In recent studies, the lognormal features of hippocampal activities were well replicated by a multi-timescale adaptive threshold (MAT) neuron network of lognormally distributed excitatory-to-excitatory synaptic weights, though it remains unknown whether and how other neuronal and network properties can be replicated in this model. Here we implement two additional studies of the same network: first, we further analyze its burstiness properties by identifying and clustering neurons with exceptionally bursty features, once again demonstrating the importance of the lognormal synaptic weight distribution. Second, we characterize dynamical patterns of activity termed neuronal avalanches in in vivo CA3 recordings of behaving rats and in the model network, revealing the similarities and differences between experimental and model avalanche size distributions across the sleep-wake cycle. These results show the comparison between the MAT neuron network and hippocampal readings in a different approach than shown before, providing more insight into the mechanisms behind activity in hippocampal subregions. / Nas regiões CA1 e CA3 do hipocampo, várias propriedades da atividade neuronal seguem distribuições assimétricas com características lognormais, incluindo frequência de disparo média, frequência e magnitude de rajadas de disparo (bursts), magnitude da sincronia populacional e correlações entre disparos pré- e pós-sinápticos. Em estudos recentes, as características lognormais das atividades hipocampais foram bem reproduzidas por uma rede de neurônios de limiar adaptativo (multi-timescale adaptive threshold, MAT) com pesos sinápticos entre neurônios excitatórios seguindo uma distribuição lognormal, embora ainda não se saiba se e como outras propriedades neuronais e da rede podem ser replicadas nesse modelo. Nesse trabalho implementamos dois estudos adicionais da mesma rede: primeiramente, analisamos mais a fundo as propriedades dos bursts identificando e agrupando neurônios com capacidade de burst excepcional, mostrando mais uma vez a importância da distribuição lognormal de pesos sinápticos. Em seguida, caracterizamos padrões dinâmicos de atividade chamados avalanches neuronais no modelo e em aquisições in vivo do CA3 de roedores em atividades comportamentais, revelando as semelhanças e diferenças entre as distribuições de tamanho de avalanche através do ciclo sono-vigília. Esses resultados mostram a comparação entre a rede de neurônios MAT e medições hipocampais em uma abordagem diferente da apresentada anteriormente, fornecendo mais percepção acerca dos mecanismos por trás da atividade em subregiões hipocampais.

In hippocampal CA1 and CA3 regions, various properties of neuronal activity follow skewed, lognormal-like distributions, including average firing rates, rate and magnitude of spike bursts, magnitude of population synchrony, and correlations between pre- and postsynaptic spikes. In recent studies, the lognormal features of hippocampal activities were well replicated by a multi-timescale adaptive threshold (MAT) neuron network of lognormally distributed excitatory-to-excitatory synaptic weights, though it remains unknown whether and how other neuronal and network properties can be replicated in this model. Here we implement two additional studies of the same network: first, we further analyze its burstiness properties by identifying and clustering neurons with exceptionally bursty features, once again demonstrating the importance of the lognormal synaptic weight distribution. Second, we characterize dynamical patterns of activity termed neuronal avalanches in in vivo CA3 recordings of behaving rats and in the model network, revealing the similarities and differences between experimental and model avalanche size distributions across the sleep-wake cycle. These results show the comparison between the MAT neuron network and hippocampal readings in a different approach than shown before, providing more insight into the mechanisms behind activity in hippocampal subregions. / Nas regiões CA1 e CA3 do hipocampo, várias propriedades da atividade neuronal seguem distribuições assimétricas com características lognormais, incluindo frequência de disparo média, frequência e magnitude de rajadas de disparo (bursts), magnitude da sincronia populacional e correlações entre disparos pré- e pós-sinápticos. Em estudos recentes, as características lognormais das atividades hipocampais foram bem reproduzidas por uma rede de neurônios de limiar adaptativo (multi-timescale adaptive threshold, MAT) com pesos sinápticos entre neurônios excitatórios seguindo uma distribuição lognormal, embora ainda não se saiba se e como outras propriedades neuronais e da rede podem ser replicadas nesse modelo. Nesse trabalho implementamos dois estudos adicionais da mesma rede: primeiramente, analisamos mais a fundo as propriedades dos bursts identificando e agrupando neurônios com capacidade de burst excepcional, mostrando mais uma vez a importância da distribuição lognormal de pesos sinápticos. Em seguida, caracterizamos padrões dinâmicos de atividade chamados avalanches neuronais no modelo e em aquisições in vivo do CA3 de roedores em atividades comportamentais, revelando as semelhanças e diferenças entre as distribuições de tamanho de avalanche através do ciclo sono-vigília. Esses resultados mostram a comparação entre a rede de neurônios MAT e medições hipocampais em uma abordagem diferente da apresentada anteriormente, fornecendo mais percepção acerca dos mecanismos por trás da atividade em subregiões hipocampais.

Excerpt: Besides the lead character herself, the leather-clad vampire Spike -- introduced as the "Big Bad" in Buffy the Vampire Slayer (BtVS) Season 2 -- the most analyzed character in the Buffyverse.

In spite of the many discoveries made in neuroscience, the mechanism by which memories are formed is still unclear. To better understand how some disorders of the brain arise, it is necessary to improve our knowledge of memory formation in the brain. With the aid of a biological experiment, an artificial neural network is developed to provide insight into how information is stored and recalled. In particular, the bi-conditional association of distinct spatial and non-spatial information is examined using computational techniques. The thesis defines three versions of a computational model based on a combination of feedforward and recurrent neural networks and a biologically-inspired spike time dependent plasticity learning rule. The ability of the computational model to store and recall the bi-conditional object-space association task through reward-modulated plastic synapses is numerically investigated. Further, the network's response to variation of certain parameter values is numerically addressed. A parallel algorithm is introduced to reduce the running time necessary to test the robustness of this artificial neural network. The numerical results produced with this algorithm are then analyzed by a statistical approach, and the network's ability for learning is assessed.

Fusarium head blight (FHB) is a serious disease of wheat (Triticum aestivum) and other small grains; disease severity is affected by temperature and rainfall. This research comprised three studies: an artificially warmed experiment during 2016-2017, a morphology study and an FHB resistance screening study in 2015-2016, using approximately 250 wheat cultivars and breeding lines from programs in the eastern US. The location was the University of Kentucky Spindletop Research Farm in Lexington, KY. Higher levels of Fusarium damaged kernels and the toxin deoxynivalenol (DON) were observed in the warmed treatment compared to the control, and plant development was accelerated. In the FHB resistance screen, significant (p < 0.05) genotype differences for all traits were observed. A GWAS identified 16 SNPs associated with resistance and susceptibility, ranging from -2.14 to 4.01%. Three DON-associated SNPs reduced toxin levels by 3.2, 2.1, and 1.5 ppm. In the morphology study, negative correlations were observed among morphological and disease traits. Small effect SNPs were identified for all morphological traits, which might be useful in genomic selection; traits like spike length, spikelet number and inclination could be used in phenotyping. Response to warming indicates that existing resistance sources may be less effective in a warming climate.

This paper is devoted principally to the presentation and discussion of the results obtained when certain contrasting characters were brought together in a wheat cross between Ridit and Sevier 59. this is one of the crosses made in an attempt to develop a wheat adapted to this region with the following desirable qualities: Bunt resistance, strong straw, hard kernels, and heavy yield. How near this ideal is approached in succeeding generations can be told only by extensive tests. The genetic study herein presented was made to hasten the time when some of the progeny of this cross may become of economic value.

<p>In this thesis five different protection harnesses for hunting dogs against canidae attacks were assessed. Hunting dogs can be attacked and severely injured or killed by wolves (<em>Canis lupus</em>) when released during hunting. So far there is no effective protection method. Similar problems are reported with African wild dogs (<em>Lycaon pictus</em>) and dholes (<em>Cuon alpinus</em>) with other domestic animals. In this study the experimental harnesses were presented on a dummy to lure the animals to attack them. The harnesses with physical (screws or spikes on the back) and ultrasound (immediate bite controlled and 19 second continuous ultrasound) deterrents were only assessed during wolf attacks, whereas the harness with electric shocks was also tested on the other two species. Neither physical nor ultrasound deterrents showed a large enough aversive response in the wolves. Electric shocks, given to the animals when biting the dummy, triggered an immediate release of the dummy in all three species. Long term effects differed between species and individuals. The exposed wolf did not touch the dummy again after a second exposure, whereas all except one African wild dog bit the dummy again in consecutive trials. Some individuals returned to bite a second time even in the same trial. An assessment of the long term effect on dholes was not possible, as the individuals were undistinguishable. Based on the data obtained in this study a harness with electric deterrent seems the most promising.</p>