Behavioral and cellular analysis of classical conditioning of feeding behavior in Aplysia

In classical conditioning, an associative form of learning, animals learn to associate two stimuli. Cellular and molecular mechanisms for the induction and consolidation of associative learning and memory at the level of single cells and synaptic connections have been studied in both vertebrate and invertebrate animals. The majority of studies, however, relied on aversive stimuli to induce learning. This bias may limit the extent to which identified mechanisms generalize to other forms of associative learning and memory, such as appetitive forms. The goal of the present study was to develop a classical conditioning procedure for the marine mollusk Aplysia californica using appetitive reinforcement, and to analyze associative learning using behavioral and electrophysiological techniques. ^ Using tactile stimulation of the lips as the conditional stimulus (CS) and food as the unconditional stimulus (US) a training protocol was developed that reliably induced classical conditioning of feeding behavior. Memory persisted for at least 24 hours. The gross organization of reinforcement-mediating pathways was analyzed in additional behavioral experiments. Moreover, neurophysiological correlates of classical conditioning were identified and characterized in an in vitro preparation containing the circuitry for feeding behavior. In vitro stimulation of a nerve (AT4) that may mediate the CS during training, resulted in a greater number of buccal motor patterns (BMPs) in brains from conditioned animals, as compared to control animals. The majority of these BMPs were ingestion-like, consistent with the increased number of bites in response to the CS after classical conditioning. Moreover, classical conditioning correlated with increased excitatory synaptic input to BMP-initiating neuron B31/32, in response to stimulation of AT 4, as compared to controls. The expression of the correlates of classical conditioning identified in this study was specific to stimulation of AT 4, which is consistent the stimulus specificity that is characteristic for classical conditioning. ^ The identification of cellular correlates of classical conditioning documented here provides the basis for future, more detailed analyses of an appetitive form of associative learning and memory, that may extend the working knowledge of the cellular and molecular mechanisms for associative plasticity in general. ^

Mechanisms of TGF-beta1-induced neuronal plasticity in Aplysia

Transforming growth factor beta-1 (TGF-β1) is a cytokine and neurotrophic factor whose neuromodulatory effects in Aplysia californica were recently described. Previous results demonstrated that TGF-β1 induces long-term increases in the efficacy of sensorimotor synapses, a neural correlate of sensitization of the defensive tail withdrawal reflex. These results provided the first evidence that a neurotrophic factor regulates neuronal plasticity associated with a simple form of learning in Aplysia, and raised many questions regarding the nature of the modulation. No homologs of TGF-β had previously been identified in Aplysia, and thus, it was not known whether components of TGF-β1 signaling pathways were present in Aplysia. Furthermore, the signaling mechanisms engaged by TGF-β1 had not been identified, and it was not known whether TGF-β1 regulated other aspects of neuronal function.^ The present investigation into the actions of TGF-β1 was initiated by examining the distribution of the type II TGF-β1 receptor, the ligand binding receptor. The receptor was widely distributed in the CNS and most neurons exhibited somatic and neuritic immunoreactivity. In addition, the ability of TGF-β1 to activate the cAMP/PKA and MAPK pathways, known to regulate several important aspects of neuronal function, was examined. TGF-β1 acutely decreased cAMP levels in sensory neurons, activated MAPK and triggered translocation of MAPK to the nucleus. MAPK activation was critical for both short- and long-term regulation of neuronal function by TGF-β1. TGF-β1 acutely decreased synaptic depression induced by low frequency stimuli in a MAPK-dependent manner. This regulation may result, at least in part, from the modulation of synapsin, a major peripheral synaptic vesicle protein. TGF-β1 stimulated MAPK-dependent phosphorylation of synapsin, a process believed to regulate synaptic vesicle mobilization from reserve to readily-releasable pools of neurotransmitter. In addition to its acute effect on synaptic efficacy, TGF-β1 also induced long-term increases in sensory neuron excitability. Whereas transient exposure to TGF-β1 was not sufficient to drive short-or long-term changes in excitability, prolonged exposure to TGF-β1 induced long-term changes in excitability that depended on MAPK. The results of these studies represent significant progress toward an understanding of the role of TGF-β1 in neuronal plasticity. ^

Ionotropic Glutamate Receptors Show Unique Distribution and Localization in the Dorsal Cochlear Nucleus of the Rhesus Monkey

The dorsal cochlear nucleus (DCN) receives auditory information via the auditory nerve coming from the cochlea. It is responsible for much of the integration of auditory information, and it projects this auditory information to higher auditory brain centers for further processing. This study focuses on the DCN of adult Rhesus monkeys to characterize two specific cell types, the fusiform and cartwheel cell, based on morphometric parameters and type of glutamate receptor they express. The fusiform cell is the main projection neuron, while the cartwheel cell is the main inhibitory interneuron. Expression of AMPA glutamate receptor subunits is localized to certain cell types. The activity of the CN depends on the AMPA receptor subunit composition and expression. Immunocytochemistry, using specific antibodies for AMPA glutamate receptor subunits GluR1, GluR2/3 and GluR4, was used in conjunction with morphometry to determine the location, morphological characteristics and expression of AMPA receptor subunits in fusiform and cartwheel cells in the primate DCN. Qualitative as well as quantitative data indicates that there are important morphological differences in cell location and expression of AMPA glutamate receptor subunits between the rodent DCN and that of primates. GluR2/3 is widely expressed in the primate DCN. GluR1 is also widely expressed in the primate DCN. GluR4 is diffusely expressed. Expression of GluR2/3 and GluR4 in the primate is similar to that of the rodent. However, expression of GluR1 is different. GluR1 is only expressed by cartwheel cells in the rodent DCN, but is expressed by a variety of cells, including fusiform cells, in the DCN of the primate.

POU domain transcription factor Brn3b is critical for the normal development and survival of retinal ganglion cells

The POU domain transcription factor Brn3b/POU4F2 plays a critical role regulating gene expression in mouse retinal ganglion cells (RGCs). Previous investigations have shown that Brn3b is not required for initial cell fate specification or migration; however, it is essential for normal RGC differentiation. In contrast to wild type axons, the mutant neurites were phenotypically different: shorter, rougher, disorganized, and poorly fasciculated. Wild type axons stained intensely with axon specific marker tau-1, while mutant projections were weakly stained and the mutant projections showed strong labeling with dendrite specific marker MAP2. Brn-3b mutant axonal projections contained more microtubules and fewer neurofilaments, a dendritic characteristic, than the wild type. The mutant neurites also exhibited significantly weaker staining of neurofilament low-molecular-weight (NF-L) in the axon when compared to the wild type, and NF-L accumulation in the neuron cell body. The absence of Brn-3b results in an inability to form normal axons and enhanced apoptosis in RGCs, suggesting that Brn-3b may control a set of genes involved in axon formation. ^ Brn3b contains several distinct sequence motifs: a glycine/serine rich region, two histidine rich regions, and a fifteen amino acid conserved sequence shared by all Brn3 family members in the N-terminus and a POU specific and POU homeodomain in the C-terminus. Brn3b activates a Luciferase reporter over 25 fold in cell culture when binding to native brn3 binding sites upstream of a minimal promoter. When fused to the Gal4 DNA Binding domain (DBD) and driven by either a strong (CMV) or weaker (pAHD) promoter, the N-terminal of Brn3b is capable of similar activation when binding to Gal4 UAS sites, indicating a presumptive activator of transcription. Both full length Brn3b or the C-terminus fused to the Gal4DBD and driven by pCMV repressed a Luciferase reporter downstream of UAS binding sites. Lower levels of expression of the fusion protein driven by pADH resulted in an alleviation of repression. This repression appears to be a limitation of this system of transcriptional analysis and a potential pitfall in conventional pCMV based transfection assays. ^

Cellular mechanisms of operant and classical conditioning

The ability to associate a predictive stimulus with a subsequent salient event (i.e., classical conditioning) and the ability to associate an expressed behavior with the consequences (i.e., operant conditioning) allow for a predictive understanding of a changing environment. Although they are operationally distinct, there has been considerable debate whether at some fundamental level classical and operant conditioning are mechanistically distinct or similar. Feeding behavior of Aplysia (i.e., biting) was chosen as the model system and was successfully conditioned with appetitive forms of both operant and classical conditioning. The neuronal circuitry responsible for feeding is well understood and is suitable for cellular analyses, thus providing for a mechanistic comparison between these two forms of associative learning. ^ Neuron B51 is part of the feeding circuitry of Aplysia and is critical for the expression of ingestive behaviors. B51 also is a locus of plasticity following both operant and classical conditioning. Both in vivo and in vitro operant conditioning increased the input resistance and the excitability of B51. No pairing-specific changes in the input resistance were observed following both in vivo and in vitro classical conditioning. However, classical conditioning decreased the excitability of B51. Thus, both operant and classical conditioning modified the threshold level for activation of neuron B51, but in opposite directions, revealing key differences in the cellular mechanisms underlying these two forms of associative learning. ^ Next, the cellular mechanisms underlying operant conditioning were investigated in more detail using a single-cell analogue. The single-cell analogue successfully recapitulated the previous in vivo and in vitro operant conditioning results by increasing the input resistance and the excitability of B51. Both PKA and PKC were necessary for operant conditioning. Dopamine appears to be the transmitter mediating the reinforcement signal in this form of conditioning. A D1 dopamine receptor antibody revealed that the D1receptor localizes to the axon hillock, which is also the region that gives the strongest response when iontophoresing dopamine. ^ The studies presented herein, thus, provide for a greater understanding of the mechanisms underlying both of these forms of associative learning and demonstrate that they likely operate through distinct cellular mechanisms. ^

Depolarization-dependent synaptic potentiation and hyperexcitability induced by a Ca2+-independent trigger

Despite vast research efforts since Cajal's seminal thoughts on the adaptation of the nervous system, researchers have only recently begun to understand the diversity of forms of neuronal plasticity and its mechanisms. All known forms of activity-dependent neuronal plasticity utilize alterations in [Ca 2+]i as a signal of changes in the membrane voltage. Ca 2+ sensors trigger modifications in excitability or synaptic strength that last from seconds to weeks and presumably years. Intriguingly, Kunjilwar et al., (unpublished observations) discovered in peripheral sensory axons of Aplysia that the induction of depolarization-dependent long-term axonal hyperexcitability does not require Ca2+ transients. Here we show that induction of depolarization-dependent intermediate-term and long-term synaptic potentiation in Aplysia occurs in conditions that prevent Ca2+ entry through voltage-gated channels and elevation of [Ca2+]i. We found that the intermediate-term synaptic potentiation induced under conditions expected to prevent Ca 2+ transients is associated with increased excitability of sensory neuron axons near presynaptic terminals, suggesting that the synaptic potentiation involves a presynaptic locus. The Ca2+-independent intermediate- and long-term synaptic potentiation appeared similar to previously reported Ca2+-dependent modifications in Aplysia. ^

Chronic stress promotes tumor growth through increased BDNF production and neo-innervation

Background: Activation of the sympathetic nervous system (SNS) in response to chronic biobehavioral stress results in high levels of catecholamines and persistent activation of adrenergic signaling, which promotes tumor growth and progression. However it is unknown how catecholamine levels within the tumor exceed systemic levels in circulation. I hypothesized that neo-innervation of tumors is required for stress-mediated effects on tumor growth. Results: First, I examined whether sympathetic nerves are present in human ovarian cancer samples as well as orthotopic ovarian cancer models. Immunohistochemical (IHC) staining for neurofilament revealed that catecholaminergic neurons are present within tumor tissue. In order to determine whether chronic stress affects the density of nerves in the tumor, I utilized an orthotopic mouse model of ovarian cancer that was exposed to daily restraint stress. IHC analysis revealed that nerve density in tumors increased by more than three-fold in stressed animals versus non-stressed controls. IHC analysis suggested that this results from both recruitment of existing neurons (axonogenesis) as well as new neuron formation (neurogenesis) within the tumor. To determine how tumors are recruiting nerve growth, I utilized a PCR array analysis of 84 nerve growth related genes and their receptors, which showed that stimulation of the SKOV3 ovarian cancer cell line with norepinephrine (NE) leads to increased expression of several neurotrophins, including brain-derived neurotrophic factor (BDNF). Neurite extension assays showed that media conditioned by ovarian cancer cell lines is capable of inducing neurite outgrowth in differentiated neuron-like PC12 cells, and NE treatment of cancer cells potentiates this effect. Norepinephrine-induced neurite extension was abolished after BDNF silencing by siRNA, suggesting that BDNF is critical to tumor cell-induced nerve growth. in vivo BDNF inhibition resulted in complete abrogation of stress-induced increases in tumor weight and nerve density, as well as downstream markers of stress. Discussion: These studies indicate that adrenergic signalling induced by chronic stress promotes neo-innervation in the tumor microenvironment. This results in a mutually beneficial relationship between the tumor cells and neurons. This work is crucial for providing a link between chronic stress and its effects on the tumor and its microenvironment. The data shown here aims to open new venues that can be used in development of therapies designed to block the stress effects on tumor growth.

STUDYING AGGREGATE FORMATION BY AMYOTROPHIC LATERAL SCLEROSIS-ASSOCIATED MUTANT SOD1 PROTEIN IN DROSOPHILA MODEL

A common pathological hallmark of most neurodegenerative disorders is the presence of protein aggregates in the brain. Understanding the regulation of aggregate formation is thus important for elucidating disease pathogenic mechanisms and finding effective preventive avenues and cures. Amyotrophic Lateral Sclerosis (ALS), also known as Lou Gehrig’s disease, is a selective neurodegenerative disorder predominantly affecting motor neurons. The majority of ALS cases are sporadic, however, mutations in superoxide dismutase 1 (SOD1) are responsible for about 20% of familial ALS (fALS). Mutated SOD1 proteins are prone to misfold and form protein aggregates, thus representing a good candidate for studying aggregate formation. The long-term goal of this project is to identify regulators of aggregate formation by mutant SOD1 and other ALS-associated disease proteins. The specific aim of this thesis project is to assess the possibility of using the well-established Drosophila model system to study aggregation by human SOD1 (hSOD1) mutants. To this end, using wild type and the three mutant hSOD1 (A4V, G85R and G93A) most commonly found among fALS, I have generated 16 different SOD1 constructs containing either eGFP or mCherry in-frame fluorescent reporters, established and tested both cell- and animal-based Drosophila hSOD1 models. The experimental strategy allows for clear visualization of ectopic hSOD1 expression as well as versatile co-expression schemes to fully investigate protein aggregation specifically by mutant hSOD1. I have performed pilot cell-transfection experiments and verified induced expression of hSOD1 proteins. Using several tissue- or cell type-specific Gal4 lines, I have confirmed the proper expression of hSOD1 from established transgenic fly lines. Interestingly, in both Drosophila S2 cells and different fly tissues including the eye and motor neurons, robust aggregate formation by either wild type or mutant hSOD1 proteins was not observed. These preliminary observations suggest that Drosophila might not be a good experimental organism to study aggregation and toxicity of mutant hSOD1 protein. Nevertheless this preliminary conclusion implies the potential existence of a potent protective mechanism against mutant hSOD1 aggregation and toxicity in Drosophila. Thus, results from my SOD1-ALS project in Drosophila will help future studies on how to best employ this classic model organism to study ALS and other human brain degenerative diseases.

The role of the intracellular second messenger cAMP in the induction of long-term morphological changes in sensory neurons of {\it Aplysia\/}

The present work examines the role of cAMP in the induction of the type of long-term morphological changes that have been shown to be correlated with long-term sensitization in Aplysia.^ To examine this issue, cAMP was injected into individual tail sensory neurons in the pleural ganglion to mimic, at the single cell level, the effects of behavioral training. After a 22 hr incubation period, the same cells were filled with horseradish peroxidase and 2 hours later the tissue was fixed and processed. Morphological analysis revealed that cAMP induced an increase in two morphological features of the neurons, varicosities and branch points. These structural alterations, which are similar to those seen in siphon sensory neurons of the abdominal ganglion following long-term sensitization training of the siphon-gill withdrawal reflex, could subserve the altered behavioral response of the animal. These results expose another role played by cAMP in the induction of learning, the initiation of a structural substrate, which, in concert with other correlates, underlies learning.^ cAMP was injected into sensory neurons in the presence of the reversible protein synthesis inhibitor, anisomycin. The presence of anisomycin during and immediately following the nucleotide injection completely blocked the structural remodeling. These results indicate that the induction of morphological changes by cAMP is a process dependent on protein synthesis.^ To further examine the temporal requirement for protein synthesis in the induction of these changes, the time of anisomycin exposure was varied. The results indicate that the cellular processes triggered by cAMP are sensitive to the inhibition of protein synthesis for at least 7 hours after the nucleotide injection. This is a longer period of sensitivity than that for the induction of another correlate of long-term sensitization, facilitation of the sensory to motor neuron synaptic connection. Thus, these findings demonstrate that the period of sensitivity to protein synthesis inhibition is not identical for all correlates of learning. In addition, since the induction of the morphological changes can be blocked by anisomycin pulses administered at different times during and following the cAMP injection, this suggests that cAMP is triggering a cascade of protein synthesis, with successive rounds of synthesis being dependent on successful completion of preceding rounds. Inhibition at any time during this cascade can block the entire process and so prevent the development of the structural changes.^ The extent to which cAMP can mimic the structural remodeling induced by long-term training was also examined. Animals were subjected to unilateral sensitization training and the morphology of the sensory neurons was examined twenty-four hours later. Both cAMP injection and long-term training produced a twofold increase in varicosities and approximately a fifty percent increase in the number of branch points in the sensory neuron arborization within the pleural ganglion. (Abstract shortened by UMI.) ^

Calcium-dependent mechanisms and long-term potentiation: Role of NMDA receptors, voltage -dependent calcium channels and persistent protein kinase activity

Long-term potentiation (LTP) is a rapidly induced and long lasting increase in synaptic strength and is the leading cellular model for learning and memory in the mammalian brain. LTP was first identified in the hippocampus, a structure implicated in memory formation. LTP induction is dependent on postsynaptic Ca2+ increases mediated by N-methyl-D-aspartate (NMDA) receptors. Activation of other postsynaptic routes of Ca2+ entry, such as voltage-dependent Ca2+ channels (VDCCs) have subsequently been shown to induce a long-lasting increase in synaptic strength. However, it is unknown if VDCC-induced LTP utilized similar cellular mechanisms as the classical NMDA receptor-dependent LTP and if these two forms of LTP display similar properties. This dissertation determines the similarities and differences in VDCC and NMDA receptor-dependent LTP in area CA1 of hippocampal slices and demonstrates that VDCCs and NMDA receptors activate similar cellular mechanisms, such as protein kinases, to induce LTP. However, VDCC and NMDA receptor activated LTP induction mechanisms are compartmentalized in the postsynaptic neuron, such that they do not interact. Consistent with activation properties of NMDA receptors and VDCCs, NMDA receptor and VDCC-dependent LTP have different induction properties. In contrast to NMDA-dependent LTP, VDCC-induced potentiation does not require evoked presynaptic stimulation or display input specificity. These results indicate that there are two different routes of postsynaptic Ca2+ which can induce LTP and the compartmentation of VDCCs and NMDA receptors and/or their resulting Ca2+ increases may account for the distinction between these LTP induction mechanisms.^ One of the molecular targets for postsynaptic Ca2+ that is required for the induction of LTP is protein kinases. Evidence for the role of protein kinase activity in LTP expression is either correlational or controversial. We have utilized a broad range and potent inhibitors of protein kinases to systematically examine the temporal requirement for protein kinases in the induction and expression of LTP. Our results indicate that there is a critical period of persistent protein kinase activity required for LTP induction activated by tetanic stimulation and extending until 20 min after HFS. In addition, our results suggest that protein kinase activity during and immediately after HFS is not sufficient for LTP induction. These results provide evidence for persistent and/or Ca2+ independent protein kinase activity involvement in LTP induction. ^

Artificial Neural Nets with Interaction of Afferents

The aim is to obtain computationally more powerful, neuro physiologically founded, artificial neurons and neural nets. Artificial Neural Nets (ANN) of the Perceptron type evolved from the original proposal by McCulloch an Pitts classical paper [1]. Essentially, they keep the computing structure of a linear machine followed by a non linear operation. The McCulloch-Pitts formal neuron (which was never considered by the author’s to be models of real neurons) consists of the simplest case of a linear computation of the inputs followed by a threshold. Networks of one layer cannot compute anylogical function of the inputs, but only those which are linearly separable. Thus, the simple exclusive OR (contrast detector) function of two inputs requires two layers of formal neurons

Pyramidal cells in prefrontal cortex: comparative observations reveal unparalleled specializations in neuronal structure among primate species

The most ubiquitous neuron in the cerebral cortex, the pyramidal cell, is characterized by markedly different dendritic structure among different cortical areas. The complex pyramidal cell phenotype in granular prefrontal cortex (gPFC) of higher primates endows specific biophysical properties and patterns of connectivity, which differ from those in other cortical regions. However, within the gPFC, data have been sampled from only a select few cortical areas. The gPFC of species such as human and macaque monkey includes more than 10 cortical areas. It remains unknown as to what degree pyramidal cell structure may vary among these cortical areas. Here we undertook a survey of pyramidal cells in the dorsolateral, medial, and orbital gPFC of cercopithecid primates. We found marked heterogeneity in pyramidal cell structure within and between these regions. Moreover, trends for gradients in neuronal complexity varied among species. As the structure of neurons determines their computational abilities, memory storage capacity and connectivity, we propose that these specializations in the pyramidal cell phenotype are an important determinant of species-specific executive cortical functions in primates.

Semi-supervised subspace clustering and applications to neuroscience

Machine learning techniques are used for extracting valuable knowledge from data. Nowa¬days, these techniques are becoming even more important due to the evolution in data ac¬quisition and storage, which is leading to data with different characteristics that must be exploited. Therefore, advances in data collection must be accompanied with advances in machine learning techniques to solve new challenges that might arise, on both academic and real applications. There are several machine learning techniques depending on both data characteristics and purpose. Unsupervised classification or clustering is one of the most known techniques when data lack of supervision (unlabeled data) and the aim is to discover data groups (clusters) according to their similarity. On the other hand, supervised classification needs data with supervision (labeled data) and its aim is to make predictions about labels of new data. The presence of data labels is a very important characteristic that guides not only the learning task but also other related tasks such as validation. When only some of the available data are labeled whereas the others remain unlabeled (partially labeled data), neither clustering nor supervised classification can be used. This scenario, which is becoming common nowadays because of labeling process ignorance or cost, is tackled with semi-supervised learning techniques. This thesis focuses on the branch of semi-supervised learning closest to clustering, i.e., to discover clusters using available labels as support to guide and improve the clustering process. Another important data characteristic, different from the presence of data labels, is the relevance or not of data features. Data are characterized by features, but it is possible that not all of them are relevant, or equally relevant, for the learning process. A recent clustering tendency, related to data relevance and called subspace clustering, claims that different clusters might be described by different feature subsets. This differs from traditional solutions to data relevance problem, where a single feature subset (usually the complete set of original features) is found and used to perform the clustering process. The proximity of this work to clustering leads to the first goal of this thesis. As commented above, clustering validation is a difficult task due to the absence of data labels. Although there are many indices that can be used to assess the quality of clustering solutions, these validations depend on clustering algorithms and data characteristics. Hence, in the first goal three known clustering algorithms are used to cluster data with outliers and noise, to critically study how some of the most known validation indices behave. The main goal of this work is however to combine semi-supervised clustering with subspace clustering to obtain clustering solutions that can be correctly validated by using either known indices or expert opinions. Two different algorithms are proposed from different points of view to discover clusters characterized by different subspaces. For the first algorithm, available data labels are used for searching for subspaces firstly, before searching for clusters. This algorithm assigns each instance to only one cluster (hard clustering) and is based on mapping known labels to subspaces using supervised classification techniques. Subspaces are then used to find clusters using traditional clustering techniques. The second algorithm uses available data labels to search for subspaces and clusters at the same time in an iterative process. This algorithm assigns each instance to each cluster based on a membership probability (soft clustering) and is based on integrating known labels and the search for subspaces into a model-based clustering approach. The different proposals are tested using different real and synthetic databases, and comparisons to other methods are also included when appropriate. Finally, as an example of real and current application, different machine learning tech¬niques, including one of the proposals of this work (the most sophisticated one) are applied to a task of one of the most challenging biological problems nowadays, the human brain model¬ing. Specifically, expert neuroscientists do not agree with a neuron classification for the brain cortex, which makes impossible not only any modeling attempt but also the day-to-day work without a common way to name neurons. Therefore, machine learning techniques may help to get an accepted solution to this problem, which can be an important milestone for future research in neuroscience. Resumen Las técnicas de aprendizaje automático se usan para extraer información valiosa de datos. Hoy en día, la importancia de estas técnicas está siendo incluso mayor, debido a que la evolución en la adquisición y almacenamiento de datos está llevando a datos con diferentes características que deben ser explotadas. Por lo tanto, los avances en la recolección de datos deben ir ligados a avances en las técnicas de aprendizaje automático para resolver nuevos retos que pueden aparecer, tanto en aplicaciones académicas como reales. Existen varias técnicas de aprendizaje automático dependiendo de las características de los datos y del propósito. La clasificación no supervisada o clustering es una de las técnicas más conocidas cuando los datos carecen de supervisión (datos sin etiqueta), siendo el objetivo descubrir nuevos grupos (agrupaciones) dependiendo de la similitud de los datos. Por otra parte, la clasificación supervisada necesita datos con supervisión (datos etiquetados) y su objetivo es realizar predicciones sobre las etiquetas de nuevos datos. La presencia de las etiquetas es una característica muy importante que guía no solo el aprendizaje sino también otras tareas relacionadas como la validación. Cuando solo algunos de los datos disponibles están etiquetados, mientras que el resto permanece sin etiqueta (datos parcialmente etiquetados), ni el clustering ni la clasificación supervisada se pueden utilizar. Este escenario, que está llegando a ser común hoy en día debido a la ignorancia o el coste del proceso de etiquetado, es abordado utilizando técnicas de aprendizaje semi-supervisadas. Esta tesis trata la rama del aprendizaje semi-supervisado más cercana al clustering, es decir, descubrir agrupaciones utilizando las etiquetas disponibles como apoyo para guiar y mejorar el proceso de clustering. Otra característica importante de los datos, distinta de la presencia de etiquetas, es la relevancia o no de los atributos de los datos. Los datos se caracterizan por atributos, pero es posible que no todos ellos sean relevantes, o igualmente relevantes, para el proceso de aprendizaje. Una tendencia reciente en clustering, relacionada con la relevancia de los datos y llamada clustering en subespacios, afirma que agrupaciones diferentes pueden estar descritas por subconjuntos de atributos diferentes. Esto difiere de las soluciones tradicionales para el problema de la relevancia de los datos, en las que se busca un único subconjunto de atributos (normalmente el conjunto original de atributos) y se utiliza para realizar el proceso de clustering. La cercanía de este trabajo con el clustering lleva al primer objetivo de la tesis. Como se ha comentado previamente, la validación en clustering es una tarea difícil debido a la ausencia de etiquetas. Aunque existen muchos índices que pueden usarse para evaluar la calidad de las soluciones de clustering, estas validaciones dependen de los algoritmos de clustering utilizados y de las características de los datos. Por lo tanto, en el primer objetivo tres conocidos algoritmos se usan para agrupar datos con valores atípicos y ruido para estudiar de forma crítica cómo se comportan algunos de los índices de validación más conocidos. El objetivo principal de este trabajo sin embargo es combinar clustering semi-supervisado con clustering en subespacios para obtener soluciones de clustering que puedan ser validadas de forma correcta utilizando índices conocidos u opiniones expertas. Se proponen dos algoritmos desde dos puntos de vista diferentes para descubrir agrupaciones caracterizadas por diferentes subespacios. Para el primer algoritmo, las etiquetas disponibles se usan para bus¬car en primer lugar los subespacios antes de buscar las agrupaciones. Este algoritmo asigna cada instancia a un único cluster (hard clustering) y se basa en mapear las etiquetas cono-cidas a subespacios utilizando técnicas de clasificación supervisada. El segundo algoritmo utiliza las etiquetas disponibles para buscar de forma simultánea los subespacios y las agru¬paciones en un proceso iterativo. Este algoritmo asigna cada instancia a cada cluster con una probabilidad de pertenencia (soft clustering) y se basa en integrar las etiquetas conocidas y la búsqueda en subespacios dentro de clustering basado en modelos. Las propuestas son probadas utilizando diferentes bases de datos reales y sintéticas, incluyendo comparaciones con otros métodos cuando resulten apropiadas. Finalmente, a modo de ejemplo de una aplicación real y actual, se aplican diferentes técnicas de aprendizaje automático, incluyendo una de las propuestas de este trabajo (la más sofisticada) a una tarea de uno de los problemas biológicos más desafiantes hoy en día, el modelado del cerebro humano. Específicamente, expertos neurocientíficos no se ponen de acuerdo en una clasificación de neuronas para la corteza cerebral, lo que imposibilita no sólo cualquier intento de modelado sino también el trabajo del día a día al no tener una forma estándar de llamar a las neuronas. Por lo tanto, las técnicas de aprendizaje automático pueden ayudar a conseguir una solución aceptada para este problema, lo cual puede ser un importante hito para investigaciones futuras en neurociencia.

A new class of symbolic abstract neural nets

Starting from the way the inter-cellular communication takes place by means of protein channels and also from the standard knowledge about neuron functioning, we propose a computing model called a tissue P system, which processes symbols in a multiset rewriting sense, in a net of cells similar to a neural net. Each cell has a finite state memory, processes multisets of symbol-impulses, and can send impulses (?excitations?) to the neighboring cells. Such cell nets are shown to be rather powerful: they can simulate a Turing machine even when using a small number of cells, each of them having a small number of states. Moreover, in the case when each cell works in the maximal manner and it can excite all the cells to which it can send impulses, then one can easily solve the Hamiltonian Path Problem in linear time. A new characterization of the Parikh images of ET0L languages are also obtained in this framework.

Polynomial approximation using particle swarm optimization of lineal enhanced neural networks with no hidden layers.

This paper presents some ideas about a new neural network architecture that can be compared to a Taylor analysis when dealing with patterns. Such architecture is based on lineal activation functions with an axo-axonic architecture. A biological axo-axonic connection between two neurons is defined as the weight in a connection in given by the output of another third neuron. This idea can be implemented in the so called Enhanced Neural Networks in which two Multilayer Perceptrons are used; the first one will output the weights that the second MLP uses to computed the desired output. This kind of neural network has universal approximation properties even with lineal activation functions. There exists a clear difference between cooperative and competitive strategies. The former ones are based on the swarm colonies, in which all individuals share its knowledge about the goal in order to pass such information to other individuals to get optimum solution. The latter ones are based on genetic models, that is, individuals can die and new individuals are created combining information of alive one; or are based on molecular/celular behaviour passing information from one structure to another. A swarm-based model is applied to obtain the Neural Network, training the net with a Particle Swarm algorithm.