OBJECTIVE: To facilitate clinical/genomic/biomedical research, constructing generalizable predictive models using cross-institutional methods while protecting privacy is imperative. However, state-of-the-art methods assume a "flattened" topology, while real-world research networks may consist of "network-of-networks" which can imply practical issues including training on small data for rare diseases/conditions, prioritizing locally trained models, and maintaining models for each level of the hierarchy. In this study, we focus on developing a hierarchical approach to inherit the benefits of the privacy-preserving methods, retain the advantages of adopting blockchain, and address practical concerns on a research network-of-networks. MATERIALS AND METHODS: We propose a framework to combine level-wise model learning, blockchain-based model dissemination, and a novel hierarchical consensus algorithm for model ensemble. We developed an example implementation HierarchicalChain (hierarchical privacy-preserving modeling on blockchain), evaluated it on 3 healthcare/genomic datasets, as well as compared its predictive correctness, learning iteration, and execution time with a state-of-the-art method designed for flattened network topology. RESULTS: HierarchicalChain improves the predictive correctness for small training datasets and provides comparable correctness results with the competing method with higher learning iteration and similar per-iteration execution time, inherits the benefits of the privacy-preserving learning and advantages of blockchain technology, and immutable records models for each level. DISCUSSION: HierarchicalChain is independent of the core privacy-preserving learning method, as well as of the underlying blockchain platform. Further studies are warranted for various types of network topology, complex data, and privacy concerns. CONCLUSION: We demonstrated the potential of utilizing the information from the hierarchical network-of-networks topology to improve prediction.
Blockchain technology and, in particular, blockchain-based cryptocurrencies offer us information that has never been seen before in the financial world. In contrast to fiat currencies, all transactions of crypto-currencies and crypto-tokens are permanently recorded on distributed ledgers and are publicly available. As a result, this allows us to construct a transaction graph and to assess not only its organization but to glean relationships between transaction graph properties and crypto price dynamics. The ultimate goal of this paper is to facilitate our understanding on horizons and limitations of what can be learned on crypto-tokens from local topology and geometry of the Ethereum transaction network whose even global network properties remain scarcely explored. By introducing novel tools based on topological data analysis and functional data depth into Blockchain Data Analytics, we show that Ethereum network (one of the most popular blockchains for creating new crypto-tokens) can provide critical insights on price strikes of crypto-tokens that are otherwise largely inaccessible with conventional data sources and traditional analytic methods.
Nazmiye Ceren Abay, Cüneyt Gürcan Akçora, Yulia R. Gel, Murat Kantarcıoğlu · 7 authors
With emergence of blockchain technologies and the associated cryptocurrencies, such as Bitcoin, understanding network dynamics behind Blockchain graphs has become a rapidly evolving research direction. Unlike other financial networks, such as stock and currency trading, blockchain based cryptocurrencies have the entire transaction graph accessible to the public (i.e., all transactions can be downloaded and analyzed). A natural question is then to ask whether the dynamics of the transaction graph impacts the price of the underlying cryptocurrency. We show that standard graph features such as degree distribution of the transaction graph may not be sufficient to capture network dynamics and its potential impact on fluctuations of Bitcoin price. In contrast, the new graph associated topological features computed using the tools of persistent homology, are found to exhibit a high utility for predicting Bitcoin price dynamics. %explain higher order interactions among the nodes in Blockchain graphs and can be used to build much more accurate price prediction models. Using the proposed persistent homology-based techniques, we offer a new elegant, easily extendable and computationally light approach for graph representation learning on Blockchain.
In this paper, we design and implement the first-ever decentralized replicated relational database with blockchain properties that we term blockchain relational database. We highlight several similarities between features provided by blockchain platforms and a replicated relational database, although they are conceptually different, primarily in their trust model. Motivated by this, we leverage the rich features, decades of research and optimization, and available tooling in relational databases to build a blockchain relational database. We consider a permissioned blockchain model of known, but mutually distrustful organizations each operating their own database instance that are replicas of one another. The replicas execute transactions independently and engage in decentralized consensus to determine the commit order for transactions. We design two approaches, the first where the commit order for transactions is agreed upon prior to executing them, and the second where transactions are executed without prior knowledge of the commit order while the ordering happens in parallel. We leverage serializable snapshot isolation (SSI) to guarantee that the replicas across nodes remain consistent and respect the ordering determined by consensus, and devise a new variant of SSI based on block height for the latter approach. We implement our system on PostgreSQL and present detailed performance experiments analyzing both approaches.
Abstract A blockchain can be considered a technological phenomenon that is made up of different interconnected and autonomous systems. Such systems are referred to here as cyber‐physical systems: complex interconnections of cyber and physical components. When cyber‐physical systems are interconnected, a new whole consisting of a system of systems is created by the autonomous systems and their intercommunication and interaction. In a blockchain, individual systems can independently make decisions on joint information transactions. The decision‐making procedures needed for this are executed based on fault‐tolerant communication and voting and consensus procedures, while the results of these decision‐making procedures are stored in distributed ledgers. Due to the intercommunication, interaction, and independent decision making by autonomous systems, the new whole of a blockchain is a complex entity. Complexity science rather than the usual reductionist scientific approach can help us better understand the behaviour of the new and continuously developing whole of a blockchain as a technological phenomenon.
Francois Belletti, Evan Sparks, Michael J. Franklin, Alexandre M. Bayen · 5 authors
Linear causal analysis is central to a wide range of important application spanning finance, the physical sciences, and engineering. Much of the existing literature in linear causal analysis operates in the time domain. Unfortunately, the direct application of time domain linear causal analysis to many real-world time series presents three critical challenges: irregular temporal sampling, long range dependencies, and scale. Moreover, real-world data is often collected at irregular time intervals across vast arrays of decentralized sensors and with long range dependencies which make naive time domain correlation estimators spurious. In this paper we present a frequency domain based estimation framework which naturally handles irregularly sampled data and long range dependencies while enabled memory and communication efficient distributed processing of time series data. By operating in the frequency domain we eliminate the need to interpolate and help mitigate the effects of long range dependencies. We implement and evaluate our new work-flow in the distributed setting using Apache Spark and demonstrate on both Monte Carlo simulations and high-frequency financial trading that we can accurately recover causal structure at scale.
In this decade, establishing structure-function relationships in human brain has become one of the most influential concepts in modern cognitive neuroscience since interactions among cerebral components are fundamental to explain cortical activities ([1]; [2]; [3]). \nIn literature such relationships have been defined in terms of structural, functional and effective connectivity. This distinction, mainly focused on the theoretic concept, is also related to the different measurement instruments and analytical tools used for acquiring and processing the data. The structural connectivity refers to a pattern of anatomical links among brain regions. Its analysis aims to characterize the architecture of complex networks underlying the cerebral functional organization. Magnetic Resonance Imaging and especially Diffusion Tensor Imaging can be used to convey information concerning the physical connection between neuronal populations. Functional/effective connectivity aims at identifying the presence and the strength of connections in terms of statistically significant dependency. The former is defined as the temporal correlation between neurophysiological events occurring in distributed neuronal groups and areas. The latter describes the causal influence that one neural system exerts over another either directly or indirectly in terms of temporal precedence and physical control ([4];[5]). Functional and effective connectivity can be estimated exploiting both Functional Magnetic Resonance Imaging (fMRI) and electrophysiological signals, such as Electroencephalography (EEG) and Magnetoencephalography (MEG), with different advantages and drawbacks, respectively. fMRI provides high spatial resolution (mm) but poor temporal precision (s) while EEG/MEG has more limited spatial resolution (cm) and higher temporal precision (ms). Because functional and effective connectivity are largely estimated over time, EEG and MEG are more suitable for calculating such connectivity. In literature several methods have been developed to characterize brain connectivity in terms of network topology, connections strength and causality, following two main approaches: the data-driven, where topology, causality and strength are all inferred from data, and the neural model-based, where the model topology is postulated from a priori knowledge and only the connections strength is estimated from the data. \n-\tData driven approach. The data driven approach includes linear, non-linear and information-based techniques. The linear ones provide a battery of indices derived by multivariate autoregressive models (MVAR) based on Granger causality principles ([6]) or MVAR frequency response ([7]). Such are Ordinary Coherence, Partial Coherence, Directed Transfer Function (DTF) and Partial Directed Coherence (PDC). These indexes measure the strength of the linear coupling between two signals; in addition DTF and PDC provide information about causal influence ([8]). \no\tAmong the non-linear techniques, phase synchronization has been shown to be very effcient in detecting interactions between oscillators. The phase locking values approach assumes that two dynamic systems may have their phases synchronized even if their amplitude are zero correlates ([9]). \no\tThe most representative information-based technique is the cross mutual information that measures the mutual dependence between two signals by quantifying the amount of information gained about one signal from measuring the other, as a function of delay between these two signals ([10]). \n-\tNeural model based approach. Representative methods are the Structural Equations Modelling (SEM) and the Dynamic Causal Modelling (DCM) ([11]; [12]). They are multivariate technique used to test hypothesis regarding the influences among interacting variables, but different concepts underlies these two methods. SEM approach assumes that neuronal dynamics are very fast in relation to signals uctuations and, hence, is based on a static neuronal model. This case, the neuronal activity has reached steady-state and changes in connectivity are led directly by changes in the covariance structure of the observed time series ([13]). On the other hand, in DCM the observed time series are modelled as a deterministic dynamical system in which external inputs causes changes in neural activity and therefore in connectivity values ([14]). \nMost approaches, like those based on Granger causality principles, have been examined in literature to quantify their ability in revealing cerebral connections ([15];[16]; [11]) but their simulation studies do not provide a comprehensive analysis because they use in silico data generated by self-referential linear methods which do not reproduce the complexity of brain. To overcome this issue, an innovative simulation approach has been developed in this work, based on a nonlinear neural mass model ([17]) totally independent of SEM and MVAR linear equation and able to address the complexity of neural networks. This no-self referential approach was exploited to generate in silico network data to be used as a benchmark, to quantitatively compare obtained results with true connections. The main objective of this work was to understand limits and advantages of MVAR indexes and SEM by exploiting the simulation study. Thus, it mainly serves as a proof-of-concept for connectivity measures under ideal conditions. Our purpose was to derive from simulation results some practical procedures in order to classify different brain states to support both cognitive research and clinical activity. First, research activity was focused to address connectivity on simulated data obtained on three regions networks characterized by different strength connections and based on different levels of non linearity. Second, a dataset, made available by Department of Medicine, University of Padova was used to explore application of these methods to real data by applying the simulation study suggestions. \nThis thesis consists of three main section. \nThe ffrst one includes Chapter 1-2-3 describing in detailed the considered connectivity measures, such are those based on Multivariate Autoregressive models and the Structural Equation Modelling, and the simulation study. The second part depicts in silico results and the application to EEG data. Finally, comments are reported in Discussion and Conclusions. \nChapter 1 explains how the connecting parameters of MVAR and SEM models are identified on EEG data and describes procedures commonly exploited to analyse connectivity. Chapter 2 reports an overview about the principal models used to generate in silico data, namely the neural mass models, and described the neural mass model exploited in this work. Finally, it characterizes network models adopted to simulate data and lists the procedure followed to generate in silico datasets. Chapter 3 summarizes the computations implemented to have more insights on our data by analysing the output of each methods. It describes the procedure used to evaluate the statistical signiffcance of each index results, such are the F-test for Granger causality index and the null distribution threshold using surrogate data for MVAR frequency indexes. Chapter 4 illustrates the results obtained with the simulation study. First, we reported the complete analysis for a representative subset of experiments, then for all datasets we showed topology and strength estimates. Chapter 6 delineates the procedure followed to study the connectivity in case of hepatic encephalopathy. Chapter 7 covers the Discussion and Conclusions. The Appendix is a parallel work aimed to understand the meaning of connectivity indexes computed via Structural Equation Modelling. By exploiting the neural mass model used to simulate cortical data, the objective is to quantify which measure its estimates represent. We demonstrated that Granger causality is a good estimator with high values both of sensitivity and specificity, while frequency indexes, DTF and PDC, are too much affected by the threshold choice and their interpretation in terms of absolute strength connection is not clear. As regard SEM, we proved the difficulty of its approach to describe just simple situations. Even if SEM is based on linear regression as well as MVAR models, it differently assumes there is no connection with past information, as if brain connectivity could describe time series relationships by the instant we observe it. Hence, it is not sufficiently robust to characterize neuronal dynamic activity. \n