A5017357771- Complex Network Analysis Techniques
- Opinion Dynamics and Social Influence
- Evolutionary Game Theory and Cooperation
- Stochastic processes and statistical mechanics
- Complex Systems and Time Series Analysis
- COVID-19 epidemiological studies
- Natural Language Processing Techniques
- Recommender Systems and Techniques
- Mathematical and Theoretical Epidemiology and Ecology Models
- Evolution and Genetic Dynamics
- Game Theory and Applications
- Speech and dialogue systems
- Advanced Graph Neural Networks
- Theoretical and Computational Physics
- Topic Modeling
- Industrial Automation and Control Systems
- Advanced Control Systems Optimization
- Explainable Artificial Intelligence (XAI)
- Privacy-Preserving Technologies in Data
- Machine Learning and Data Classification
- Social Media and Politics
- Caching and Content Delivery
- Mobile Ad Hoc Networks
- Immunotherapy and Immune Responses
- Hepatitis B Virus Studies
ServiceNow (United States)
2023
University of California, Davis
2013-2018
Université Laval
2009-2012
BC Centre for Disease Control
2009
University of British Columbia
2008-2009
General Motors (United States)
2002
Adaptive networks have been recently introduced in the context of disease propagation on complex networks. They account for mutual interaction between network topology and states nodes. Until now, existing models analyzed using low-complexity analytic formalisms, revealing nevertheless some novel dynamical features. However, current methods failed to reproduce with accuracy simultaneous time evolution underlying topology. In framework adaptive SIS model Gross et al. [Phys. Rev. Lett. 96,...
Epidemics seldom occur as isolated phenomena. Typically, two or more viral agents spread within the same host population and may interact dynamically with each other. We present a general model where via an immunity mechanism they propagate simultaneously on networks connecting set of nodes. By exploiting correspondence between propagation dynamics dynamical process performing progressive network generation, we develop analytical approach that accurately captures interaction epidemics...
Considerable attention has been paid, in recent years, to the use of networks modeling complex real-world systems. Among many dynamical processes involving networks, propagation -- which final state can be obtained by studying underlying network percolation properties have raised formidable interest. In this paper, we present a bond model multitype with an arbitrary joint degree distribution that allows heterogeneity edge occupation probability. As previously demonstrated, approach...
Mathematical models of infectious diseases, which are in principle analytically tractable, use two general approaches. The first approach, generally known as compartmental modeling, addresses the time evolution disease propagation at expense simplifying pattern transmission. second approach uses network theory to incorporate detailed information pertaining underlying contact structure among individuals while disregarding progression during outbreaks. So far, only alternative that enables...
Analytical description of propagation phenomena on random networks has flourished in recent years, yet more complex systems have mainly been studied through numerical means. In this paper, a mean-field is used to coherently couple the dynamics network elements (nodes, vertices, individuals...) one hand and their recurrent topological patterns (subgraphs, groups...) other hand. SIS model epidemic spread social with community structure, approach yields set ODEs for time evolution system, as...
Controlling self-organizing systems is challenging because the system responds to controller. Here, we develop a model that captures essential mechanisms of Bak-Tang-Wiesenfeld (BTW) sandpiles on networks, self-organized critical (SOC) system. This enables studying simple control scheme determines frequency cascades and shapes systemic risk. We show optimal strategies exist for generic cost functions controlling subcritical may drive it criticality. approach could enable other systems.
Aligning visual features with language embeddings is a key challenge in vision-language models (VLMs). The performance of such hinges on having good connector that maps generated by vision encoder to shared embedding space the LLM while preserving semantic similarity. Existing connectors, as multilayer perceptrons (MLPs), often produce out-of-distribution or noisy inputs, leading misalignment between modalities. In this work, we propose novel vision-text alignment method, AlignVLM, weighted...
We introduce a mechanism which models the emergence of universal properties complex networks, such as scale independence, modularity and self-similarity, unifies them under scale-free organization beyond link. This brings new perspective on network where communities, instead links, are fundamental building blocks systems. show how our simple model can reproduce social information networks by predicting their community structure more importantly, nodes or communities interconnected, often in...
Abstract Multiplex networks (a system of multiple that have different types links but share a common set nodes) arise naturally in wide spectrum fields. Theoretical studies show such multiplex networks, correlated edge dynamics between the layers can profound effect on dynamical processes. However, how to extract correlations from real-world systems is an outstanding challenge. Here we introduce Markov chain quantify found longitudinal data networks. By comparing results obtained perspective...
The mechanisms underlying cascading failures are often modeled via the paradigm of self-organized criticality. Here we introduce a simple network model where nodes self-organize to be either weakly or strongly protected against failure in manner that captures trade-off between degradation and reinforcement inherent many systems. If strong cannot fail, any is contained single, isolated cluster weak produces power-law distributions sizes. We classify large, rare events involve only single as...
We introduce a formalism for computing bond percolation properties of class correlated and clustered random graphs. This graphs is generalization the Configuration Model where nodes different types are connected via hyperedges, edges that can link more than 2 nodes. argue multitype approach coupled with use hyperedges reproduce wide spectrum complex patterns, thus enhances our capability to model real networks. As an illustration this claim, we highlight unusual behaviors size composition...
There is strong evidence that human papillomavirus (HPV) necessary for the development of cervical cancer. A prophylactic HPV vaccine with high reported efficacy was approved in North America 2006.A mathematical model transmission dynamics used to simulate different scenarios natural disease outcomes and intervention strategies. sensitivity analysis performed compensate uncertainties surrounding key epidemiological parameters.The expected impact vaccines have on cancer incidence prevalence...
By generating the specifics of a network structure only when needed (on-the-fly), we derive simple stochastic process that exactly models time evolution susceptible-infectious dynamics on finite-size networks. The small number dynamical variables this birth-death Markov greatly simplifies analytical calculations. We show how dual description, treating large scale epidemics with Gaussian approximation and outbreaks branching process, provides an accurate distribution even for rather approach...
Many complex systems have been shown to share universal properties of organization, such as scale independence, modularity and self-similarity. We borrow tools from statistical physics in order study structural preferential attachment (SPA), a recently proposed growth principle for the emergence aforementioned properties. corresponding stochastic process terms its time evolution, asymptotic behavior scaling steady state. Moreover, approximations are introduced facilitate modelling real...
The onset of large-scale connectivity in a network (i.e., percolation) often has major impact on the function system. Traditionally, graph percolation is analyzed by adding edges to fixed set initially isolated nodes. Several years ago, it was shown that nodes as well can yield an infinite order transition, which much smoother than traditional second-order transition. More recently, via competitive process lead delayed, extremely abrupt transition with significant jump large but finite...
The Bak-Tang-Wiesenfeld (BTW) sandpile process is an archetypal, stylized model of complex systems with a critical point as attractor their dynamics. This phenomenon, called self-organized criticality, appears to occur ubiquitously in both nature and technology. Initially introduced on the two-dimensional lattice, BTW has been studied network structures great analytical successes estimation macroscopic quantities, such exponents asymptotically power-law distributions. In this article, we...
We introduce a set of iterative equations that exactly solves the size distribution components on small arbitrary graphs after random removal edges.We also demonstrate how these can be used to predict node partitions (i.e., constrained each component) in undirected graphs.Besides opening way theoretical prediction percolation large but finite size, we show our results find application graph theory, epidemiology, and fragmentation theory.
Recent empirical work highlights the heterogeneity of social competitions such as political campaigns: proponents some ideologies seek debate and conversation, others create echo chambers. While symmetric static network structure is typically used a substrate to study competitor dynamics, can instead be interpreted signature strategies, yielding competition dynamics on adaptive networks. Here we demonstrate that tradeoffs between aggressiveness defensiveness (i.e., targeting adversaries vs....
Online social network (OSN) based applications often rely on user interactions to propagate information or recruit more users, producing a sequence of actions called adoption process cascades. This paper presents the first attempt quantitatively study cascade such OSN-based by analyzing detailed activity data from popular Facebook gifting application. In particular, due challenge monitoring over all possible channels OSN platforms, we focus characterizing that relies only user-based...
In-context learning (ICL) approaches typically leverage prompting to condition decoder-only language model generation on reference information. Just-in-time processing of a context is inefficient due the quadratic cost self-attention operations, and caching desirable. However, transformer states can easily require almost as much space parameters. When right isn't known in advance, ICL be challenging. This work addresses these limitations by introducing models that, inspired encoder-decoder...
Multimodal AI has the potential to significantly enhance document-understanding tasks, such as processing receipts, understanding workflows, extracting data from documents, and summarizing reports. Code generation tasks that require long-structured outputs can also be enhanced by multimodality. Despite this, their use in commercial applications is often limited due access training restrictive licensing, which hinders open access. To address these limitations, we introduce BigDocs-7.5M, a...
We argue that, when establishing and benchmarking Machine Learning (ML) models, the research community should favour evaluation metrics that better capture value delivered by their model in practical applications. For a specific class of use cases -- selective classification we show not only can it be simple enough to do, but has import consequences provides insights what look for ``good'' ML model.
Recent empirical work highlights the heterogeneity of social competitions such as political campaigns: proponents some ideologies seek debate and conversation, others create echo chambers. While symmetric static network structure is typically used a substrate to study competitor dynamics, can instead be interpreted signature strategies, yielding competition dynamics on adaptive networks. Here we demonstrate that tradeoffs between aggressiveness defensiveness (i.e., targeting adversaries vs....