arXiv subjects

Matheus Palhares Viana

Publications and source records attributed to Matheus Palhares Viana.

7 recordsLinked to original sources

arXiv2025

Bright 4B: Scaling Hyperspherical Learning for Segmentation in 3D Brightfield Microscopy

Label-free 3D brightfield microscopy offers a fast and noninvasive way to visualize cellular morphology, yet robust volumetric segmentation still typically depends on fluorescence or heavy post-processing. We address this gap by introducing Bright-4B, a 4 billion parameter foundation model that learns on the unit hypersphere to segment subcellular structures directly from 3D brightfield volumes. Bright-4B combines a hardware-aligned Native Sparse Attention mechanism (capturing local, coarse, and selected global context), depth-width residual HyperConnections that stabilize representation flow, and a soft Mixture-of-Experts for adaptive capacity. A plug-and-play anisotropic patch embed further respects confocal point-spread and axial thinning, enabling geometry-faithful 3D tokenization. The resulting model produces morphology-accurate segmentations of nuclei, mitochondria, and other organelles from brightfield stacks alone--without fluorescence, auxiliary channels, or handcrafted post-processing. Across multiple confocal datasets, Bright-4B preserves fine structural detail across depth and cell types, outperforming contemporary CNN and Transformer baselines. All code, pretrained weights, and models for downstream finetuning will be released to advance large-scale, label-free 3D cell mapping.

cs.CV↗

arXiv2020

An argument in favor of strong scaling for deep neural networks with small datasets

In recent years, with the popularization of deep learning frameworks and large datasets, researchers have started parallelizing their models in order to train faster. This is crucially important, because they typically explore many hyperparameters in order to find the best ones for their applications. This process is time consuming and, consequently, speeding up training improves productivity. One approach to parallelize deep learning models followed by many researchers is based on weak scaling. The minibatches increase in size as new GPUs are added to the system. In addition, new learning rates schedules have been proposed to fix optimization issues that occur with large minibatch sizes. In this paper, however, we show that the recommendations provided by recent work do not apply to models that lack large datasets. In fact, we argument in favor of using strong scaling for achieving reliable performance in such cases. We evaluated our approach with up to 32 GPUs and show that weak scaling not only does not have the same accuracy as the sequential model, it also fails to converge most of time. Meanwhile, strong scaling has good scalability while having exactly the same accuracy of a sequential implementation.

cs.DC↗

arXiv2012

Predicting epidemic outbreak from individual features of the spreaders

Knowing which individuals can be more efficient in spreading a pathogen throughout a determinate environment is a fundamental question in disease control. Indeed, over the last years the spread of epidemic diseases and its relationship with the topology of the involved system have been a recurrent topic in complex network theory, taking into account both network models and real-world data. In this paper we explore possible correlations between the heterogeneous spread of an epidemic disease governed by the susceptible-infected-recovered (SIR) model, and several attributes of the originating vertices, considering Erdös-Rényi (ER), Barabási-Albert (BA) and random geometric graphs (RGG), as well as a real case of study, the US Air Transportation Network that comprises the US 500 busiest airports along with inter-connections. Initially, the heterogeneity of the spreading is achieved considering the RGG networks, in which we analytically derive an expression for the distribution of the spreading rates among the established contacts, by assuming that such rates decay exponentially with the distance that separates the individuals. Such distribution is also considered for the ER and BA models, where we observe topological effects on the correlations. In the case of the airport network, the spreading rates are empirically defined, assumed to be directly proportional to the seat availability. Among both the theoretical and the real networks considered, we observe a high correlation between the total epidemic prevalence and the degree, as well as the strength and the accessibility of the epidemic sources. For attributes such as the betweenness centrality and the $k$-shell index, however, the correlation depends on the topology considered.

physics.soc-ph↗

arXiv2009

On the Efficiency of Underground Systems in Large Cities

We report an analysis of the accessibility between different locations in big cities, which is illustrated with respect to London and Paris. The effects of the respective underground systems in facilitating more uniform access to diverse places are also quantified and investigated. It is shown that London and Paris have markedly different patterns of accessibility, as a consequence of the number of bridges and large parks of London, and that in both cases the respective underground systems imply in general increase of accessibility.

physics.soc-ph↗

arXiv2005

Hierarchy, Fractality, Small-World and Resilience of Haversian Bone Structure: A Complex Network Study

This article describes the application of recently introduced complex networks concepts and methods to the characterization and analysis of cortical bone structure. Three-dimensional reconstructions of the system of channels underlying bone structure are obtained by using histological and computer graphics methods and then represented in terms of complex networks. Conﬂuences of two or more channels are represented as nodes, while the interconnecting channels are expressed as edges. The hierarchical backbone (the tree with maximum depth) of such a network is obtained and understood to correspond to the main structure underlying the channel system. The remainder of the network is shown to correspond to geographical communities, suggesting that the bone channel structure involves a number of regular communities appended along the hierarchical backbone. It is shown that such additional edges play a crucial role in enhancing the network resilience and in reducing the shortest paths in both topology and geometry.The recently introduced concept of fractal dimension of a network (cond-mat/0503078) is then correlated with the resilience of the several components of the bone channel structure to obstruction and failure, with important implications for the understanding of the organization and robustness of cortical bone structure.

q-bio.TO↗

arXiv2005

The Dynamics of Hierarchical Evolution of Complex Networks

Introduced recently, the concept of hierarchical degree allows a more complete characterization of the topological context of a node in a complex network than the traditional node degree. This article presents analytical characterization and studies of the density of hierarchical degrees in random and scale free networks. The obtained results allowed the identification of a hierarchy-dependent power law for the degrees of nodes in random complex networks, with Poisson density for the first hierarchical degree (obtained through master equation approach). Exact results were obtained for the second hierarchical degree in scale free networks.

cond-mat.stat-mech↗

arXiv2004

The complex channel networks of bone structure

Bone structure in mammals involves a complex network of channels (Havers and Volkmann channels) required to nourish the bone marrow cells. This work describes how three-dimensional reconstructions of such systems can be obtained and represented in terms of complex networks. Three important findings are reported: (i) the fact that the channel branching density resembles a power law implies the existence of distribution hubs; (ii) the conditional node degree density indicates a clear tendency of connection between nodes with degrees 2 and 4; and (iii) the application of the recently introduced concept of hierarchical clustering coefficient allows the identification of typical scales of channel redistribution. A series of important biological insights is drawn and discussed

q-bio.TO↗