Neural network based off-line handwritten text recognition system

This dissertation introduces a new system for handwritten text recognition based on an improved neural network design. Most of the existing neural networks treat mean square error function as the standard error function. The system as proposed in this dissertation utilizes the mean quartic error function, where the third and fourth derivatives are non-zero. Consequently, many improvements on the training methods were achieved. The training results are carefully assessed before and after the update. To evaluate the performance of a training system, there are three essential factors to be considered, and they are from high to low importance priority: (1) error rate on testing set, (2) processing time needed to recognize a segmented character and (3) the total training time and subsequently the total testing time. It is observed that bounded training methods accelerate the training process, while semi-third order training methods, next-minimal training methods, and preprocessing operations reduce the error rate on the testing set. Empirical observations suggest that two combinations of training methods are needed for different case character recognition. Since character segmentation is required for word and sentence recognition, this dissertation provides also an effective rule-based segmentation method, which is different from the conventional adaptive segmentation methods. Dictionary-based correction is utilized to correct mistakes resulting from the recognition and segmentation phases. The integration of the segmentation methods with the handwritten character recognition algorithm yielded an accuracy of 92% for lower case characters and 97% for upper case characters. In the testing phase, the database consists of 20,000 handwritten characters, with 10,000 for each case. The testing phase on the recognition 10,000 handwritten characters required 8.5 seconds in processing time.

Flexible pavement layer moduli determination : an adaptive artificial neural network approach

The estimation of pavement layer moduli through the use of an artificial neural network is a new concept which provides a less strenuous strategy for backcalculation procedures. Artificial Neural Networks are biologically inspired models of the human nervous system. They are specifically designed to carry out a mapping characteristic. This study demonstrates how an artificial neural network uses non-destructive pavement test data in determining flexible pavement layer moduli. The input parameters include plate loadings, corresponding sensor deflections, temperature of pavement surface, pavement layer thicknesses and independently deduced pavement layer moduli.

A Static Traffic Assignment Model Combined with an Artificial Neural Network Delay Model

As traffic congestion continues to worsen in large urban areas, solutions are urgently sought. However, transportation planning models, which estimate traffic volumes on transportation network links, are often unable to realistically consider travel time delays at intersections. Introducing signal controls in models often result in significant and unstable changes in network attributes, which, in turn, leads to instability of models. Ignoring the effect of delays at intersections makes the model output inaccurate and unable to predict travel time. To represent traffic conditions in a network more accurately, planning models should be capable of arriving at a network solution based on travel costs that are consistent with the intersection delays due to signal controls. This research attempts to achieve this goal by optimizing signal controls and estimating intersection delays accordingly, which are then used in traffic assignment. Simultaneous optimization of traffic routing and signal controls has not been accomplished in real-world applications of traffic assignment. To this end, a delay model dealing with five major types of intersections has been developed using artificial neural networks (ANNs). An ANN architecture consists of interconnecting artificial neurons. The architecture may either be used to gain an understanding of biological neural networks, or for solving artificial intelligence problems without necessarily creating a model of a real biological system. The ANN delay model has been trained using extensive simulations based on TRANSYT-7F signal optimizations. The delay estimates by the ANN delay model have percentage root-mean-squared errors (%RMSE) that are less than 25.6%, which is satisfactory for planning purposes. Larger prediction errors are typically associated with severely oversaturated conditions. A combined system has also been developed that includes the artificial neural network (ANN) delay estimating model and a user-equilibrium (UE) traffic assignment model. The combined system employs the Frank-Wolfe method to achieve a convergent solution. Because the ANN delay model provides no derivatives of the delay function, a Mesh Adaptive Direct Search (MADS) method is applied to assist in and expedite the iterative process of the Frank-Wolfe method. The performance of the combined system confirms that the convergence of the solution is achieved, although the global optimum may not be guaranteed.

The Neural Basis of Involuntary Episodic Memories

Involuntary episodic memories are memories that come into consciousness without preceding retrieval effort. These memories are commonplace and are relevant to multiple mental disorders. However, they are vastly understudied. We use a novel paradigm to elicit involuntary memories in the laboratory so that we can study their neural basis. In session one, an encoding session, sounds are presented with picture pairs or alone. In session two, in the scanner, sounds-picture pairs and unpaired sounds are reencoded. Immediately following, participants are split into two groups: a voluntary and an involuntary group. Both groups perform a sound localization task in which they hear the sounds and indicate the side from which they are coming. The voluntary group additionally tries to remember the pictures that were paired with the sounds. Looking at neural activity, we find a main effect of condition (paired vs. unpaired sounds) showing similar activity in both groups for voluntary and involuntary memories in regions typically associated with retrieval. There is also a main effect of group (voluntary vs. involuntary) in the dorsolateral prefrontal cortex, a region typically associated with cognitive control. Turning to connectivity similarities and differences between groups again, there is a main effect of condition showing paired > unpaired sounds are associated with a recollection network. In addition, three group differences were found: (1) increased connectivity between the pulvinar nucleus of the thalamus and the recollection network for the voluntary group, (2) a higher association between the voluntary group and a network that includes regions typically found in frontoparietal and cingulo-opercular networks, and (3) shorter path length for about half of the nodes in these networks for the voluntary group. Finally, we use the same paradigm to compare involuntary memories in people with posttraumatic stress disorder (PTSD) to trauma-controls. This study also included the addition of emotional pictures. There were two main findings. (1) A similar pattern of activity was found for paired > unpaired sounds for both groups but this activity was delayed in the PTSD group. (2) A similar pattern of activity was found for high > low emotion stimuli but it occurred early in the PTSD group compared to the control group. Our results suggest that involuntary and voluntary memories share the same neural representation but that voluntary memories are associated with additional cognitive control processes. They also suggest that disorders associated with cognitive deficits, like PTSD, can affect the processing of involuntary memories.

Tourism demand modelling and forecasting with artificial neural network models: the Mozambique case study

This study is aimed to model and forecast the tourism demand for Mozambique for the period from January 2004 to December 2013 using artificial neural networks models. The number of overnight stays in Hotels was used as representative of the tourism demand. A set of independent variables were experimented in the input of the model, namely: Consumer Price Index, Gross Domestic Product and Exchange Rates, of the outbound tourism markets, South Africa, United State of America, Mozambique, Portugal and the United Kingdom. The best model achieved has 6.5% for Mean Absolute Percentage Error and 0.696 for Pearson correlation coefficient. A model like this with high accuracy of forecast is important for the economic agents to know the future growth of this activity sector, as it is important for stakeholders to provide products, services and infrastructures and for the hotels establishments to adequate its level of capacity to the tourism demand.

Improving drug discovery using hybrid softcomputing methods

Virtual screening (VS) methods can considerably aid clinical research, predicting how ligands interact with drug targets. Most VS methods suppose a unique binding site for the target, but it has been demonstrated that diverse ligands interact with unrelated parts of the target and many VS methods do not take into account this relevant fact. This problem is circumvented by a novel VS methodology named BINDSURF that scans the whole protein surface in order to find new hotspots, where ligands might potentially interact with, and which is implemented in last generation massively parallel GPU hardware, allowing fast processing of large ligand databases. BINDSURF can thus be used in drug discovery, drug design, drug repurposing and therefore helps considerably in clinical research. However, the accuracy of most VS methods and concretely BINDSURF is constrained by limitations in the scoring function that describes biomolecular interactions, and even nowadays these uncertainties are not completely understood. In order to improve accuracy of the scoring functions used in BINDSURF we propose a hybrid novel approach where neural networks (NNET) and support vector machines (SVM) methods are trained with databases of known active (drugs) and inactive compounds, being this information exploited afterwards to improve BINDSURF VS predictions.

Improvement of Virtual Screening Predictions using Computational Intelligence Methods

Virtual Screening (VS) methods can considerably aid clinical research, predicting how ligands interact with drug targets. However, the accuracy of most VS methods is constrained by limitations in the scoring function that describes biomolecular interactions, and even nowadays these uncertainties are not completely understood. In order to improve accuracy of scoring functions used in most VS methods we propose a hybrid novel approach where neural networks (NNET) and support vector machines (SVM) methods are trained with databases of known active (drugs) and inactive compounds, this information being exploited afterwards to improve VS predictions.

Neural fields with sigmoidal firing rates: approximate solutions

Many tissue level models of neural networks are written in the language of nonlinear integro-differential equations. Analytical solutions have only been obtained for the special case that the nonlinearity is a Heaviside function. Thus the pursuit of even approximate solutions to such models is of interest to the broad mathematical neuroscience community. Here we develop one such scheme, for stationary and travelling wave solutions, that can deal with a certain class of smoothed Heaviside functions. The distribution that smoothes the Heaviside is viewed as a fundamental object, and all expressions describing the scheme are constructed in terms of integrals over this distribution. The comparison of our scheme and results from direct numerical simulations is used to highlight the very good levels of approximation that can be achieved by iterating the process only a small number of times.

Self-Organizing Map Neural Architectures Based on Limit Cycle Attractors

Recent efforts to develop large-scale neural architectures have paid relatively little attention to the use of self-organizing maps (SOMs). Part of the reason is that most conventional SOMs use a static encoding representation: Each input is typically represented by the fixed activation of a single node in the map layer. This not only carries information in an inefficient and unreliable way that impedes building robust multi-SOM neural architectures, but it is also inconsistent with rhythmic oscillations in biological neural networks. Here I develop and study an alternative encoding scheme that instead uses limit cycle attractors of multi-focal activity patterns to represent input patterns/sequences. Such a fundamental change in representation raises several questions: Can this be done effectively and reliably? If so, will map formation still occur? What properties would limit cycle SOMs exhibit? Could multiple such SOMs interact effectively? Could robust architectures based on such SOMs be built for practical applications? The principal results of examining these questions are as follows. First, conditions are established for limit cycle attractors to emerge in a SOM through self-organization when encoding both static and temporal sequence inputs. It is found that under appropriate conditions a set of learned limit cycles are stable, unique, and preserve input relationships. In spite of the continually changing activity in a limit cycle SOM, map formation continues to occur reliably. Next, associations between limit cycles in different SOMs are learned. It is shown that limit cycles in one SOM can be successfully retrieved by another SOM’s limit cycle activity. Control timings can be set quite arbitrarily during both training and activation. Importantly, the learned associations generalize to new inputs that have never been seen during training. Finally, a complete neural architecture based on multiple limit cycle SOMs is presented for robotic arm control. This architecture combines open-loop and closed-loop methods to achieve high accuracy and fast movements through smooth trajectories. The architecture is robust in that disrupting or damaging the system in a variety of ways does not completely destroy the system. I conclude that limit cycle SOMs have great potentials for use in constructing robust neural architectures.

Synchronization of Heteroclinic Circuits Through Learning in Chains of Neural Motifs

The synchronization of oscillatory activity in networks of neural networks is usually implemented through coupling the state variables describing neuronal dynamics. In this study we discuss another but complementary mechanism based on a learning process with memory. A driver network motif, acting as a teacher, exhibits winner-less competition (WLC) dynamics, while a driven motif, a learner, tunes its internal couplings according to the oscillations observed in the teacher. We show that under appropriate training the learner motif can dynamically copy the coupling pattern of the teacher and thus synchronize oscillations with the teacher. Then, we demonstrate that the replication of the WLC dynamics occurs for intermediate memory lengths only. In a unidirectional chain of N motifs coupled through teacher-learner paradigm the time interval required for pattern replication grows linearly with the chain size, hence the learning process does not blow up and at the end we observe phase synchronized oscillations along the chain. We also show that in a learning chain closed into a ring the network motifs come to a consensus, i.e. to a state with the same connectivity pattern corresponding to the mean initial pattern averaged over all network motifs.

Functional, clustered, feedforward, and mesoscale brain networks

The brain is a network spanning multiple scales from subcellular to macroscopic. In this thesis I present four projects studying brain networks at different levels of abstraction. The first involves determining a functional connectivity network based on neural spike trains and using a graph theoretical method to cluster groups of neurons into putative cell assemblies. In the second project I model neural networks at a microscopic level. Using diferent clustered wiring schemes, I show that almost identical spatiotemporal activity patterns can be observed, demonstrating that there is a broad neuro-architectural basis to attain structured spatiotemporal dynamics. Remarkably, irrespective of the precise topological mechanism, this behavior can be predicted by examining the spectral properties of the synaptic weight matrix. The third project introduces, via two circuit architectures, a new paradigm for feedforward processing in which inhibitory neurons have the complex and pivotal role in governing information flow in cortical network models. Finally, I analyze axonal projections in sleep deprived mice using data collected as part of the Allen Institute's Mesoscopic Connectivity Atlas. After normalizing for experimental variability, the results indicate there is no single explanatory difference in the mesoscale network between control and sleep deprived mice. Using machine learning techniques, however, animal classification could be done at levels significantly above chance. This reveals that intricate changes in connectivity do occur due to chronic sleep deprivation.

Edge-aware Graph Attention Networks: Joint Reasoning On Text and Knowledge Graphs for Biomedical Question Answering

Much of the real-world dataset, including textual data, can be represented using graph structures. The use of graphs to represent textual data has many advantages, mainly related to maintaining a more significant amount of information, such as the relationships between words and their types. In recent years, many neural network architectures have been proposed to deal with tasks on graphs. Many of them consider only node features, ignoring or not giving the proper relevance to relationships between them. However, in many node classification tasks, they play a fundamental role. This thesis aims to analyze the main GNNs, evaluate their advantages and disadvantages, propose an innovative solution considered as an extension of GAT, and apply them to a case study in the biomedical field. We propose the reference GNNs, implemented with methodologies later analyzed, and then applied to a question answering system in the biomedical field as a replacement for the pre-existing GNN. We attempt to obtain better results by using models that can accept as input both node and edge features. As shown later, our proposed models can beat the original solution and define the state-of-the-art for the task under analysis.

Deploying and processing neural representations of signals

Neural representations (NR) have emerged in the last few years as a powerful tool to represent signals from several domains, such as images, 3D shapes, or audio. Indeed, deep neural networks have been shown capable of approximating continuous functions that describe a given signal with theoretical infinite resolution. This finding allows obtaining representations whose memory footprint is fixed and decoupled from the resolution at which the underlying signal can be sampled, something that is not possible with traditional discrete representations, e.g., grids of pixels for images or voxels for 3D shapes. During the last two years, many techniques have been proposed to improve the capability of NR to approximate high-frequency details and to make the optimization procedures required to obtain NR less demanding both in terms of time and data requirements, motivating many researchers to deploy NR as the main form of data representation for complex pipelines. Following this line of research, we first show that NR can approximate precisely Unsigned Distance Functions, providing an effective way to represent garments that feature open 3D surfaces and unknown topology. Then, we present a pipeline to obtain in a few minutes a compact Neural Twin® for a given object, by exploiting the recent advances in modeling neural radiance fields. Furthermore, we move a step in the direction of adopting NR as a standalone representation, by considering the possibility of performing downstream tasks by processing directly the NR weights. We first show that deep neural networks can be compressed into compact latent codes. Then, we show how this technique can be exploited to perform deep learning on implicit neural representations (INR) of 3D shapes, by only looking at the weights of the networks.

Regularization meets GreenAI: a new framework for image reconstruction in life sciences applications

Ill-conditioned inverse problems frequently arise in life sciences, particularly in the context of image deblurring and medical image reconstruction. These problems have been addressed through iterative variational algorithms, which regularize the reconstruction by adding prior knowledge about the problem's solution. Despite the theoretical reliability of these methods, their practical utility is constrained by the time required to converge. Recently, the advent of neural networks allowed the development of reconstruction algorithms that can compute highly accurate solutions with minimal time demands. Regrettably, it is well-known that neural networks are sensitive to unexpected noise, and the quality of their reconstructions quickly deteriorates when the input is slightly perturbed. Modern efforts to address this challenge have led to the creation of massive neural network architectures, but this approach is unsustainable from both ecological and economic standpoints. The recently introduced GreenAI paradigm argues that developing sustainable neural network models is essential for practical applications. In this thesis, we aim to bridge the gap between theory and practice by introducing a novel framework that combines the reliability of model-based iterative algorithms with the speed and accuracy of end-to-end neural networks. Additionally, we demonstrate that our framework yields results comparable to state-of-the-art methods while using relatively small, sustainable models. In the first part of this thesis, we discuss the proposed framework from a theoretical perspective. We provide an extension of classical regularization theory, applicable in scenarios where neural networks are employed to solve inverse problems, and we show there exists a trade-off between accuracy and stability. Furthermore, we demonstrate the effectiveness of our methods in common life science-related scenarios. In the second part of the thesis, we initiate an exploration extending the proposed method into the probabilistic domain. We analyze some properties of deep generative models, revealing their potential applicability in addressing ill-posed inverse problems.

Uncovering the encoding and transmission of behaviourally relevant information in neural activity

Most cognitive functions require the encoding and routing of information across distributed networks of brain regions. Information propagation is typically attributed to physical connections existing between brain regions, and contributes to the formation of spatially correlated activity patterns, known as functional connectivity. While structural connectivity provides the anatomical foundation for neural interactions, the exact manner in which it shapes functional connectivity is complex and not yet fully understood. Additionally, traditional measures of directed functional connectivity only capture the overall correlation between neural activity, and provide no insight on the content of transmitted information, limiting their ability in understanding neural computations underlying the distributed processing of behaviorally-relevant variables. In this work, we first study the relationship between structural and functional connectivity in simulated recurrent spiking neural networks with spike timing dependent plasticity. We use established measures of time-lagged correlation and overall information propagation to infer the temporal evolution of synaptic weights, showing that measures of dynamic functional connectivity can be used to reliably reconstruct the evolution of structural properties of the network. Then, we extend current methods of directed causal communication between brain areas, by deriving an information-theoretic measure of Feature-specific Information Transfer (FIT) quantifying the amount, content and direction of information flow. We test FIT on simulated data, showing its key properties and advantages over traditional measures of overall propagated information. We show applications of FIT to several neural datasets obtained with different recording methods (magneto and electro-encephalography, spiking activity, local field potentials) during various cognitive functions, ranging from sensory perception to decision making and motor learning. Overall, these analyses demonstrate the ability of FIT to advance the investigation of communication between brain regions, uncovering the previously unaddressed content of directed information flow.