The document discusses local variation analysis of hashtag spike trains on Twitter. It defines local variation (LV) as the ratio of the difference between the inter-event interval of the forward event and the inter-event interval of the backward event, at discrete time points of the spike train. The document suggests that LV can help characterize whether a time series is Poissonian or not, and can provide insights into the dynamics of hashtag propagation on Twitter.
From Data to Knowledge in Mons, Belgium: "Taken from the cover of Atlas of Science, Katy Borner (MIT Press, 2010), where one reads: Noise becomes data when it has a cognitive pattern. Data becomes information when assembled into a coherent whole, which can be related to other information. Information becomes knowledge when integrated with other information in a form useful for making decisions and determining actions. Knowledge becomes understanding when related to other knowledge in a manner useful, in anticipating, judging and acting. Understanding becomes wisdom when informed by purpose, ethics, principles, memory and projection. George Santayana"
NET 303 Policy Primer - Instagram's Terms of Uselauraclairecapel
Instagram's terms of use allow the platform to use any posted content without compensation and share user information with Facebook. While users retain copyright over their content, Instagram is granted broad rights to use, modify and share that content. The terms also give Instagram wide discretion to change the service or terms at any time. Criticism arose when new terms allowed paid display of user content without consent or a way to opt out. Users ultimately have little control over how their content is used once posted to Instagram.
The endocrine system consists of glands that produce hormones and regulate bodily functions. The hypothalamus and hypophysis make up the central regulatory formations of the endocrine system. Peripheral endocrine glands include the thyroid gland, parathyroid glands, adrenal glands, pancreas, gonads, and others. These glands produce hormones that are released into the bloodstream to target and regulate distant tissues and organs. The endocrine system works through feedback loops to maintain homeostasis.
This short document contains a link and encourages the reader to click on it to access or obtain something. No other context or information is provided about what would be accessed or obtained by clicking the link.
Este documento contiene propuestas para varios experimentos científicos sencillos que se pueden realizar en casa, incluyendo cómo generar electricidad con limones, hacer fuego con una papa, simular los efectos de la lluvia ácida en las plantas, y estudiar las reacciones de combustión apagando una vela bajo un vaso.
Eating disorders are serious mental illnesses that cause severe disturbances to eating behaviors and how one feels about their body weight and shape. The most common types are anorexia nervosa, bulimia nervosa, binge eating disorder, and eating disorder not otherwise specified. People with anorexia see themselves as overweight even when dangerously underweight and restrict food intake. Those with bulimia have recurrent episodes of eating large amounts of food and feeling a lack of control followed by purging. Binge eating disorder involves losing control over eating without purging. Treatments include adequate nutrition, reducing excessive exercise, stopping purging, psychotherapy, medical care, nutritional counseling, and medications.
From Data to Knowledge in Mons, Belgium: "Taken from the cover of Atlas of Science, Katy Borner (MIT Press, 2010), where one reads: Noise becomes data when it has a cognitive pattern. Data becomes information when assembled into a coherent whole, which can be related to other information. Information becomes knowledge when integrated with other information in a form useful for making decisions and determining actions. Knowledge becomes understanding when related to other knowledge in a manner useful, in anticipating, judging and acting. Understanding becomes wisdom when informed by purpose, ethics, principles, memory and projection. George Santayana"
NET 303 Policy Primer - Instagram's Terms of Uselauraclairecapel
Instagram's terms of use allow the platform to use any posted content without compensation and share user information with Facebook. While users retain copyright over their content, Instagram is granted broad rights to use, modify and share that content. The terms also give Instagram wide discretion to change the service or terms at any time. Criticism arose when new terms allowed paid display of user content without consent or a way to opt out. Users ultimately have little control over how their content is used once posted to Instagram.
The endocrine system consists of glands that produce hormones and regulate bodily functions. The hypothalamus and hypophysis make up the central regulatory formations of the endocrine system. Peripheral endocrine glands include the thyroid gland, parathyroid glands, adrenal glands, pancreas, gonads, and others. These glands produce hormones that are released into the bloodstream to target and regulate distant tissues and organs. The endocrine system works through feedback loops to maintain homeostasis.
This short document contains a link and encourages the reader to click on it to access or obtain something. No other context or information is provided about what would be accessed or obtained by clicking the link.
Este documento contiene propuestas para varios experimentos científicos sencillos que se pueden realizar en casa, incluyendo cómo generar electricidad con limones, hacer fuego con una papa, simular los efectos de la lluvia ácida en las plantas, y estudiar las reacciones de combustión apagando una vela bajo un vaso.
Eating disorders are serious mental illnesses that cause severe disturbances to eating behaviors and how one feels about their body weight and shape. The most common types are anorexia nervosa, bulimia nervosa, binge eating disorder, and eating disorder not otherwise specified. People with anorexia see themselves as overweight even when dangerously underweight and restrict food intake. Those with bulimia have recurrent episodes of eating large amounts of food and feeling a lack of control followed by purging. Binge eating disorder involves losing control over eating without purging. Treatments include adequate nutrition, reducing excessive exercise, stopping purging, psychotherapy, medical care, nutritional counseling, and medications.
Apomediation refers to guiding consumers to health information without being required to access that information. This is done through "apomediaries" who have little control over the information. There are ethical concerns that incorrect or dangerous information could spread without oversight. Nurses must act as apomediaries to help clients find appropriate information while addressing ethical considerations like verifying information and advocating for the client's well-being.
The document discusses demand functions and supply functions. It provides examples of demand functions with different slopes:
1) Negative slope - as price increases, demand decreases.
2) Constant slope - a change in demand does not affect price.
3) Undefined slope - demand is not affected by price.
4) Increasing slope - demand increases as price increases.
It also discusses cases of supply functions with positive, constant, or negative slopes and how quantity supplied relates to price in each case. Examples are provided of constructing a demand equation from data points and using the equation to solve for price, demand, or other variables.
The Flexi Biogas system is a flexible above-ground biogas system manufactured in Kenya that is simpler and less costly to build and operate than traditional fixed dome biogas systems. It consists of a plastic digester bag housed in a greenhouse tunnel. Organic waste placed in the bag produces biogas through anaerobic digestion, which is then piped to appliances. The Flexi Biogas system is portable, inexpensive at $410, and takes only one day to install compared to $1,000 and 21 days for fixed dome systems. It also has a shorter retention time of 15 days versus 45 days.
The document provides an introduction to outcomes and outcome frameworks for nonprofits. It discusses how the nonprofit sector has traditionally focused on activities and outputs but is now shifting toward an outcomes approach. An outcomes approach emphasizes measuring the effects and changes that result from a program rather than just the activities. It also discusses some key tools used in outcomes-based planning, tracking, reporting, and assessment, including logic models, results-based accountability (RBA), targeting outcomes of programs (TOP), and balanced scorecards. The document uses examples and visual snapshots to explain these different outcome frameworks and tools from 30,000 feet.
Dokumen tersebut membahas tentang pemeriksaan jumlah trombosit dalam diagnosis laboratorium, termasuk bahan pemeriksaan, metode pemeriksaan secara langsung dan tidak langsung, serta estimasi jumlah trombosit pada sediaan apus darah tepi.
Vitamin C Serum ampoule, containing various functional ingredients like antioxidant, moisturiser, and even stem cell culture, to enhance and keep your skin more active, elastic, and younger. Big hit seller in S. Korea, sold USD 15 million in last six months.
El coaching tiene sus orígenes en Sócrates y su método mayéutico de guiar a otros hacia la verdad. El documento describe cómo el coaching puede ayudar a una empresa mediante el desarrollo del potencial humano, mejorando el trabajo en equipo y compromiso. Propone implementar sesiones de coaching de hasta 5 minutos diarias entre líderes y empleados para establecer comunicación bidireccional, identificar necesidades de entrenamiento, y desarrollar posibles sucesores.
The document discusses analyzing the temporal dynamics of hashtag diffusion on Twitter. It presents two methods: (1) Analyzing heterogeneity in hashtag popularity, which finds that a small number of hashtags are very popular while most are unpopular; (2) Performing a local analysis of hashtag spike trains to quantify non-stationarity using the Lv metric, which compares variations in the intervals between hashtag mentions. The analysis aims to characterize information diffusion and find differences between popular and unpopular hashtags.
[IRTalks@The University of Glasgow] A Topology-aware Analysis of Graph Collab...Daniele Malitesta
Slides for the invited talk "A Topology-aware Analysis of Graph Collaborative Filtering" at the Glasgow Information Retrieval Group (University of Glasgow).
Event link: https://samoa.dcs.gla.ac.uk/events/viewtalk.jsp?id=19320
Paper: https://arxiv.org/pdf/2308.10778.pdf
The document discusses similarities between the temporal patterns of communication in online social networks like Twitter and the dynamics of beads floating on liquid surfaces.
It hypothesizes that both systems self-organize under restricted resources, causing Twitter users to collectively spread messages and beads to form moving groups, creating dynamic heterogeneity. Local variation analysis is applied to characterize burstiness in user time series on the Higgs boson rumor, finding low popularity users communicate in bursts while high popularity users are more regular. Comparisons with bead experiments suggest interpreting dynamic heterogeneity in critical systems could help characterize viral hashtags in Twitter.
Anomaly Detection in Sequences of Short Text Using Iterative Language ModelsCynthia Freeman
The document discusses various methods for anomaly detection in time series data. It begins by defining time series and anomalies, noting that anomaly detection is challenging due to issues like lack of labeled data and data imbalance. It then covers characteristics of time series like seasonality, trends, and concept drift, and how to detect them. Various anomaly detection methods are outlined, including STL, SARIMA, Prophet, Gaussian processes, and RNNs. Evaluation methods and factors to consider in choosing a detection method are also discussed. The document provides an overview of approaches to determining the optimal anomaly detection model for a given time series and application.
RSC: Mining and Modeling Temporal Activity in Social MediaAlceu Ferraz Costa
Presentation of the KDD 2015 paper describing the RSC model:
RSC: Mining and Modeling Temporal Activity in Social Media
Alceu Ferraz Costa, Yuto Yamaguchi, Agma Juci Machado Traina, Caetano Traina Jr., and Christos Faloutsos
The 21st SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 2015
FAST School of ComputingProject Differential Equations (MTChereCheek752
FAST School of Computing
Project Differential Equations (MT-224)
Due Date: 14th, June 2021. Max Marks: 70
A Brief Literature Review:
We have studied the population growth model i.e., if P represents population. Since the
population varies over time, it is understood to be a function of time. Therefore we use the
notation P (t) for the population as a function of time. If P (t) is a differentiable function,
then the first derivative
dP
dt
represents the instantaneous rate of change of the population
as a function of time, which is proportional to present population in case of the exponential
growth and decay of populations and radioactive substances. Mathematically
dP
dt
∝ P.
We can verify that the function P (t) = P0e
rt satisfies the initial-value problem
dP
dt
= rP, P (0) = P0.
This differential equation has an interesting interpretation. The left-hand side represents
the rate at which the population increases (or decreases). The right-hand side is equal to a
positive constant multiplied by the current population. Therefore the differential equation
states that the rate at which the population increases is proportional to the population at
that point in time. Furthermore, it states that the constant of proportionality never changes.
One problem with this function is its prediction that as time goes on, the population grows
without bound. This is unrealistic in a real-world setting. Various factors limit the rate of
growth of a particular population, including birth rate, death rate, food supply, predators,
diseases and so on. The growth constant r usually takes into consideration the birth and
death rates but none of the other factors, and it can be interpreted as a net (birth minus
death) percent growth rate per unit time. A natural question to ask is whether the population
growth rate stays constant, or whether it changes over time. Biologists have found that in
many biological systems, the population grows until a certain steady-state population is
reached. This possibility is not taken into account with exponential growth. However, the
concept of carrying capacity allows for the possibility that in a given area, only a certain
number of a given organism or animal can thrive without running into resource issues.
• The carrying capacity of an organism in a given environment is defined to be the maxi-
mum population of that organism that the environment can sustain indefinitely.
• We use the variable K to denote the carrying capacity. The growth rate is represented by
the variable r. Using these variables, we can define the logistic differential equation.
dP
dt
= rP
(
1 −
P
K
)
.
1
• An improvement to the logistic model includes a threshold population. The threshold
population is defined to be the minimum population that is necessary for the species
to survive. We use the variable T to represent the threshold population. A differential
equation that incorporates both the threshold population T and carrying capacit ...
The document describes an activity analysis and visualization project with the following objectives:
1. Build a system to support groups in learning how to work more effectively through visualizing collaboration data logs.
2. Develop different types of visualizations like activity radars and interaction networks to provide insights into participation, interactions, and timelines of events.
3. Apply data mining techniques to find frequent patterns and sequences of events that characterize aspects of teamwork.
Building graphs to discover information by David Martínez at Big Data Spain 2015Big Data Spain
Graphs can be built from raw data to discover information by representing relationships between data points as graph connections. Techniques like locality sensitive hashing can be used to efficiently construct graphs from high-dimensional data by mapping similar points to the same "buckets". Once a graph is built, algorithms can find structure like connected components, detect anomalies using local outlier factor, perform clustering, and make inferences about unlabeled nodes. Building graphs is a powerful approach for transforming raw data into useful information through network analysis and machine learning on graphs.
Practice of Streaming Processing of Dynamic Graphs: Concepts, Models, and Sys...Subhajit Sahu
Highlighted notes on Practice of Streaming Processing of Dynamic Graphs: Concepts, Models, and Systems.
While doing research work under Prof. Dip Banerjee, Prof, Kishore Kothapalli.
This is a huge review paper discussing a lot about several graph streaming frameworks, and graph databases. How can i summarize this! GPU frameworks given are cuSTINGER, EvoGraph, Hornet, faimGraph, GPMA. Gap between databases and frameworks seems to be closing.
This document discusses extracting event-related information from social media data that contains geo-temporal tags. It presents several scenarios where users post photos on social media from locations and events. It then describes using techniques like Ripley's K-function and cross-K-function to analyze spatial point patterns of geo-tagged terms and identify clusters indicating points of interest. This spatial analysis can help extract and rank socially tagged terms for locations. The document also discusses scaling techniques and using the spatial analysis to support image search related to geo-temporal events.
The Semantic Evolution of Online CommunitiesMatthew Rowe
This paper studies the semantic evolution of online communities over time. It constructs semantic graphs based on concepts and entities discussed in community forums between given time intervals. It analyzes the macro evolution of these graphs across communities by measuring graph properties like node count, diameter, entropy, and specializations over time. It finds that for concept graphs, these measures tend to converge over time, while for entity graphs, node count increases linearly but other measures converge. This provides insights into how semantics in online communities develop and differ between communities.
Incremental View Maintenance for openCypher QueriesGábor Szárnyas
Presented at the Fourth openCypher Implementers Meeting
Numerous graph use cases require continuous evaluation of queries over a constantly changing data set, e.g. fraud detection in financial systems, recommendations, and checking integrity constraints. For relational systems, incremental view maintenance has been researched for three decades, resulting in a wide body of literature. The property graph data model and the openCypher language, however, are recent developments, and therefore lack established techniques to perform efficient view maintenance. In this talk, we give an overview of the view maintenance problem for property graphs, discuss why it is particularly difficult and present an approach that tackles a meaningful subset of the language.
Apomediation refers to guiding consumers to health information without being required to access that information. This is done through "apomediaries" who have little control over the information. There are ethical concerns that incorrect or dangerous information could spread without oversight. Nurses must act as apomediaries to help clients find appropriate information while addressing ethical considerations like verifying information and advocating for the client's well-being.
The document discusses demand functions and supply functions. It provides examples of demand functions with different slopes:
1) Negative slope - as price increases, demand decreases.
2) Constant slope - a change in demand does not affect price.
3) Undefined slope - demand is not affected by price.
4) Increasing slope - demand increases as price increases.
It also discusses cases of supply functions with positive, constant, or negative slopes and how quantity supplied relates to price in each case. Examples are provided of constructing a demand equation from data points and using the equation to solve for price, demand, or other variables.
The Flexi Biogas system is a flexible above-ground biogas system manufactured in Kenya that is simpler and less costly to build and operate than traditional fixed dome biogas systems. It consists of a plastic digester bag housed in a greenhouse tunnel. Organic waste placed in the bag produces biogas through anaerobic digestion, which is then piped to appliances. The Flexi Biogas system is portable, inexpensive at $410, and takes only one day to install compared to $1,000 and 21 days for fixed dome systems. It also has a shorter retention time of 15 days versus 45 days.
The document provides an introduction to outcomes and outcome frameworks for nonprofits. It discusses how the nonprofit sector has traditionally focused on activities and outputs but is now shifting toward an outcomes approach. An outcomes approach emphasizes measuring the effects and changes that result from a program rather than just the activities. It also discusses some key tools used in outcomes-based planning, tracking, reporting, and assessment, including logic models, results-based accountability (RBA), targeting outcomes of programs (TOP), and balanced scorecards. The document uses examples and visual snapshots to explain these different outcome frameworks and tools from 30,000 feet.
Dokumen tersebut membahas tentang pemeriksaan jumlah trombosit dalam diagnosis laboratorium, termasuk bahan pemeriksaan, metode pemeriksaan secara langsung dan tidak langsung, serta estimasi jumlah trombosit pada sediaan apus darah tepi.
Vitamin C Serum ampoule, containing various functional ingredients like antioxidant, moisturiser, and even stem cell culture, to enhance and keep your skin more active, elastic, and younger. Big hit seller in S. Korea, sold USD 15 million in last six months.
El coaching tiene sus orígenes en Sócrates y su método mayéutico de guiar a otros hacia la verdad. El documento describe cómo el coaching puede ayudar a una empresa mediante el desarrollo del potencial humano, mejorando el trabajo en equipo y compromiso. Propone implementar sesiones de coaching de hasta 5 minutos diarias entre líderes y empleados para establecer comunicación bidireccional, identificar necesidades de entrenamiento, y desarrollar posibles sucesores.
The document discusses analyzing the temporal dynamics of hashtag diffusion on Twitter. It presents two methods: (1) Analyzing heterogeneity in hashtag popularity, which finds that a small number of hashtags are very popular while most are unpopular; (2) Performing a local analysis of hashtag spike trains to quantify non-stationarity using the Lv metric, which compares variations in the intervals between hashtag mentions. The analysis aims to characterize information diffusion and find differences between popular and unpopular hashtags.
[IRTalks@The University of Glasgow] A Topology-aware Analysis of Graph Collab...Daniele Malitesta
Slides for the invited talk "A Topology-aware Analysis of Graph Collaborative Filtering" at the Glasgow Information Retrieval Group (University of Glasgow).
Event link: https://samoa.dcs.gla.ac.uk/events/viewtalk.jsp?id=19320
Paper: https://arxiv.org/pdf/2308.10778.pdf
The document discusses similarities between the temporal patterns of communication in online social networks like Twitter and the dynamics of beads floating on liquid surfaces.
It hypothesizes that both systems self-organize under restricted resources, causing Twitter users to collectively spread messages and beads to form moving groups, creating dynamic heterogeneity. Local variation analysis is applied to characterize burstiness in user time series on the Higgs boson rumor, finding low popularity users communicate in bursts while high popularity users are more regular. Comparisons with bead experiments suggest interpreting dynamic heterogeneity in critical systems could help characterize viral hashtags in Twitter.
Anomaly Detection in Sequences of Short Text Using Iterative Language ModelsCynthia Freeman
The document discusses various methods for anomaly detection in time series data. It begins by defining time series and anomalies, noting that anomaly detection is challenging due to issues like lack of labeled data and data imbalance. It then covers characteristics of time series like seasonality, trends, and concept drift, and how to detect them. Various anomaly detection methods are outlined, including STL, SARIMA, Prophet, Gaussian processes, and RNNs. Evaluation methods and factors to consider in choosing a detection method are also discussed. The document provides an overview of approaches to determining the optimal anomaly detection model for a given time series and application.
RSC: Mining and Modeling Temporal Activity in Social MediaAlceu Ferraz Costa
Presentation of the KDD 2015 paper describing the RSC model:
RSC: Mining and Modeling Temporal Activity in Social Media
Alceu Ferraz Costa, Yuto Yamaguchi, Agma Juci Machado Traina, Caetano Traina Jr., and Christos Faloutsos
The 21st SIGKDD Conference on Knowledge Discovery and Data Mining (KDD), 2015
FAST School of ComputingProject Differential Equations (MTChereCheek752
FAST School of Computing
Project Differential Equations (MT-224)
Due Date: 14th, June 2021. Max Marks: 70
A Brief Literature Review:
We have studied the population growth model i.e., if P represents population. Since the
population varies over time, it is understood to be a function of time. Therefore we use the
notation P (t) for the population as a function of time. If P (t) is a differentiable function,
then the first derivative
dP
dt
represents the instantaneous rate of change of the population
as a function of time, which is proportional to present population in case of the exponential
growth and decay of populations and radioactive substances. Mathematically
dP
dt
∝ P.
We can verify that the function P (t) = P0e
rt satisfies the initial-value problem
dP
dt
= rP, P (0) = P0.
This differential equation has an interesting interpretation. The left-hand side represents
the rate at which the population increases (or decreases). The right-hand side is equal to a
positive constant multiplied by the current population. Therefore the differential equation
states that the rate at which the population increases is proportional to the population at
that point in time. Furthermore, it states that the constant of proportionality never changes.
One problem with this function is its prediction that as time goes on, the population grows
without bound. This is unrealistic in a real-world setting. Various factors limit the rate of
growth of a particular population, including birth rate, death rate, food supply, predators,
diseases and so on. The growth constant r usually takes into consideration the birth and
death rates but none of the other factors, and it can be interpreted as a net (birth minus
death) percent growth rate per unit time. A natural question to ask is whether the population
growth rate stays constant, or whether it changes over time. Biologists have found that in
many biological systems, the population grows until a certain steady-state population is
reached. This possibility is not taken into account with exponential growth. However, the
concept of carrying capacity allows for the possibility that in a given area, only a certain
number of a given organism or animal can thrive without running into resource issues.
• The carrying capacity of an organism in a given environment is defined to be the maxi-
mum population of that organism that the environment can sustain indefinitely.
• We use the variable K to denote the carrying capacity. The growth rate is represented by
the variable r. Using these variables, we can define the logistic differential equation.
dP
dt
= rP
(
1 −
P
K
)
.
1
• An improvement to the logistic model includes a threshold population. The threshold
population is defined to be the minimum population that is necessary for the species
to survive. We use the variable T to represent the threshold population. A differential
equation that incorporates both the threshold population T and carrying capacit ...
The document describes an activity analysis and visualization project with the following objectives:
1. Build a system to support groups in learning how to work more effectively through visualizing collaboration data logs.
2. Develop different types of visualizations like activity radars and interaction networks to provide insights into participation, interactions, and timelines of events.
3. Apply data mining techniques to find frequent patterns and sequences of events that characterize aspects of teamwork.
Building graphs to discover information by David Martínez at Big Data Spain 2015Big Data Spain
Graphs can be built from raw data to discover information by representing relationships between data points as graph connections. Techniques like locality sensitive hashing can be used to efficiently construct graphs from high-dimensional data by mapping similar points to the same "buckets". Once a graph is built, algorithms can find structure like connected components, detect anomalies using local outlier factor, perform clustering, and make inferences about unlabeled nodes. Building graphs is a powerful approach for transforming raw data into useful information through network analysis and machine learning on graphs.
Practice of Streaming Processing of Dynamic Graphs: Concepts, Models, and Sys...Subhajit Sahu
Highlighted notes on Practice of Streaming Processing of Dynamic Graphs: Concepts, Models, and Systems.
While doing research work under Prof. Dip Banerjee, Prof, Kishore Kothapalli.
This is a huge review paper discussing a lot about several graph streaming frameworks, and graph databases. How can i summarize this! GPU frameworks given are cuSTINGER, EvoGraph, Hornet, faimGraph, GPMA. Gap between databases and frameworks seems to be closing.
This document discusses extracting event-related information from social media data that contains geo-temporal tags. It presents several scenarios where users post photos on social media from locations and events. It then describes using techniques like Ripley's K-function and cross-K-function to analyze spatial point patterns of geo-tagged terms and identify clusters indicating points of interest. This spatial analysis can help extract and rank socially tagged terms for locations. The document also discusses scaling techniques and using the spatial analysis to support image search related to geo-temporal events.
The Semantic Evolution of Online CommunitiesMatthew Rowe
This paper studies the semantic evolution of online communities over time. It constructs semantic graphs based on concepts and entities discussed in community forums between given time intervals. It analyzes the macro evolution of these graphs across communities by measuring graph properties like node count, diameter, entropy, and specializations over time. It finds that for concept graphs, these measures tend to converge over time, while for entity graphs, node count increases linearly but other measures converge. This provides insights into how semantics in online communities develop and differ between communities.
Incremental View Maintenance for openCypher QueriesGábor Szárnyas
Presented at the Fourth openCypher Implementers Meeting
Numerous graph use cases require continuous evaluation of queries over a constantly changing data set, e.g. fraud detection in financial systems, recommendations, and checking integrity constraints. For relational systems, incremental view maintenance has been researched for three decades, resulting in a wide body of literature. The property graph data model and the openCypher language, however, are recent developments, and therefore lack established techniques to perform efficient view maintenance. In this talk, we give an overview of the view maintenance problem for property graphs, discuss why it is particularly difficult and present an approach that tackles a meaningful subset of the language.
Moving on Twitter: Using Episodic Hotspot and Drift Analysis to Detect and Ch...Hansi Senaratne
Today, a tremendous source of spatio-temporal data is user generated, so-called volunteered geographic information (VGI). Among the many VGI sources, microblogged services, such as Twitter, are extensively used to disseminate information on a near real-time basis. Interest in analysis of microblogged data has been motivated to date by many applications ranging from trend detection, early disaster warning, to urban management and marketing. One important analysis perspective in understanding microblogged data is based on the notion of drift, considering a gradual change of real world phenomena observed across space, time, content, or a combination thereof. The scientific contribution provided by this paper is the presentation of a systematic framework that utilises on the one hand a Kernel Density Estimation (KDE) to detect hotspot clusters of Tweeter activities, which are episodically sequential in nature. These clusters help to derive spatial trajectories. On the other hand we introduce the concept of drift that characterises these trajectories by looking into changes of sentiment and topics to derive meaningful information. We apply our approach to a Twitter dataset comprising 26,000 tweets. We demonstrate how phenomena of interest can be detected by our approach. As an example, we use our approach to detect the locations of Lady Gaga’s concert tour in 2013. A set of visualisations allows to analyse the identified trajectories in space, enhanced by optional overlays for sentiment or other parameters of interest.
Search Engines for Machine Learning: Presented by Joe Blue, MapRLucidworks
This document discusses building and deploying machine learning models for recommender systems. It describes using user interaction data to train recommendation models to suggest additional items. It outlines challenges in deploying models and improving results. It provides examples of using Apache Solr and MapR technologies to build a workflow for analyzing user history data, updating item metadata, and delivering recommendations.
Demystifying Parallel and Distributed Deep Learning: An In-Depth Concurrency ...inside-BigData.com
In this deck from 2018 Swiss HPC Conference, Torsten Hoefler from (ETH) Zürich presents: Demystifying Parallel and Distributed Deep Learning: An In-Depth Concurrency Analysis.
“Deep Neural Networks (DNNs) are becoming an important tool in modern computing applications. Accelerating their training is a major challenge and techniques range from distributed algorithms to low-level circuit design. In this talk, we describe the problem from a theoretical perspective, followed by approaches for its parallelization.
Specifically, we present trends in DNN architectures and the resulting implications on parallelization strategies. We discuss the different types of concurrency in DNNs; synchronous and asynchronous stochastic gradient descent; distributed system architectures; communication schemes; and performance modeling. Based on these approaches, we extrapolate potential directions for parallelism in deep learning.”
Learn more: https://www.arxiv.org/abs/1802.09941
and
http://hpcadvisorycouncil.com
Sign up for our insideHPC Newsletter: http://insidehpc.com/newsletter
State-of-the-art time-series prediction with continuous-time recurrent neural networks.
Neural networks with continuous-time hidden state representations have become unprecedentedly popular within the machine learning community. This is due to their strong approximation capability in modeling time-series, their adaptive computation modality, their memory and parameter efficiency. In this talk Ramin will discuss how this family of neural networks work and why they realize attractive degrees of generalizability across different application domains.
OUR SPEAKER
Ramin Hasani, PhD, Machine Learning Scientist at TU Wien, expert in robotics, including previously being a scholar MIT CSAL, presents technical aspects of continuous-time neural networks.
1. The document presents a method for detecting social events using mobile phone data by analyzing the mobility and social behavior of mobile phone users.
2. It aims to detect unusual large gatherings of people (social events) and identify frequent locations like home or work.
3. The method uses Bayesian inference on antenna connections to estimate user locations and identifies events as weeks with unusually high numbers of probable attendees based on comparing ordinary and event presence probabilities.
Big Data Day LA 2016/ Big Data Track - Twitter Heron @ Scale - Karthik Ramasa...Data Con LA
Twitter generates billions and billions of events per day. Analyzing these events in real time presents a massive challenge. Twitter designed and deployed a new streaming system called Heron. Heron has been in production nearly 2 years and is widely used by several teams for diverse use cases. This talk looks at Twitter's operating experiences and challenges of running Heron at scale and the approaches taken to solve those challenges.
Similar to Local Variation of Collective Attention in Hashtag Spike Trains (20)
"Financial Odyssey: Navigating Past Performance Through Diverse Analytical Lens"sameer shah
Embark on a captivating financial journey with 'Financial Odyssey,' our hackathon project. Delve deep into the past performance of two companies as we employ an array of financial statement analysis techniques. From ratio analysis to trend analysis, uncover insights crucial for informed decision-making in the dynamic world of finance."
Codeless Generative AI Pipelines
(GenAI with Milvus)
https://ml.dssconf.pl/user.html#!/lecture/DSSML24-041a/rate
Discover the potential of real-time streaming in the context of GenAI as we delve into the intricacies of Apache NiFi and its capabilities. Learn how this tool can significantly simplify the data engineering workflow for GenAI applications, allowing you to focus on the creative aspects rather than the technical complexities. I will guide you through practical examples and use cases, showing the impact of automation on prompt building. From data ingestion to transformation and delivery, witness how Apache NiFi streamlines the entire pipeline, ensuring a smooth and hassle-free experience.
Timothy Spann
https://www.youtube.com/@FLaNK-Stack
https://medium.com/@tspann
https://www.datainmotion.dev/
milvus, unstructured data, vector database, zilliz, cloud, vectors, python, deep learning, generative ai, genai, nifi, kafka, flink, streaming, iot, edge
The Ipsos - AI - Monitor 2024 Report.pdfSocial Samosa
According to Ipsos AI Monitor's 2024 report, 65% Indians said that products and services using AI have profoundly changed their daily life in the past 3-5 years.
STATATHON: Unleashing the Power of Statistics in a 48-Hour Knowledge Extravag...sameer shah
"Join us for STATATHON, a dynamic 2-day event dedicated to exploring statistical knowledge and its real-world applications. From theory to practice, participants engage in intensive learning sessions, workshops, and challenges, fostering a deeper understanding of statistical methodologies and their significance in various fields."
Local Variation of Collective Attention in Hashtag Spike Trains
1. • A typical snapsho
The white spots
of the beads floa
waves.
cedaysan@gmail.com
http://fcxn.wordpress.com
http://xn.unamur.be
r driving want to be mobile. As a
witter users collectively advertise and
orm groups to move together. Both
f-organize and create dynamic
ty.
e, the interpretation of the dynamic
eity of the beads in a critical limit
elp to characterize viral memes
(#hashtags) in twitter.
Refs:
1 C. Sanlı et al. (a
2 L. Berthier (201
• A typical snapshot o
The white spots indi
of the beads floating
waves.
cedaysan@gmail.com
http://fcxn.wordpress.com
http://xn.unamur.be
driving want to be mobile. As a
ter users collectively advertise and
m groups to move together. Both
organize and create dynamic
the interpretation of the dynamic
ty of the beads in a critical limit
p to characterize viral memes
hashtags) in twitter.
Refs:
1 C. Sanlı et al. (arXi
2 L. Berthier (2011).
Local Variation of Collective Attention
in Hashtag Spike Trains
!
(collaboration with Renaud Lambiotte)5mm
• Inset: The displacement
field demonstrates local
heterogeneities in the flow.
• A typical snapshot of an experiment:
The white spots indicate the positions
of the beads floating on surface
waves.
cedaysan@gmail.com
http://fcxn.wordpress.com
http://xn.unamur.be
es social
sages and
As a
ertise and
er. Both
mic
s:
dynamic
al limit
mes
Refs:
1 C. Sanlı et al. (arXiv - 2013).
2 L. Berthier (2011).
International AAAI Conference on Weblogs and Social Media (ICWSM-15) Workshop 3: Modeling and Mining
Temporal Interactions, 26th May 2015, Oxford, the UK.
@CeydaSanli
2. Online
SEPT. 29, 2014
Photo
Credit
Tomi Um
Brendan Nyhan
Information Diffusion in Twitter
Local Variation Spike Trains 1C. Sanli, CompleXity Networks, UNamur
• tweets
• retweets
• mentions
Tomi Um
• following -
followers
3. y Rumors Outrace the Truth
ne
014
Brendan Nyhan
Hashtag Diffusion in Twitter
Local Variation Spike Trains 2C. Sanli, CompleXity Networks, UNamur
Tomi Um
• hashtag
• hashtag spike train
time
count
4. What do we address in this talk?
Local Variation Spike Trains 3C. Sanli, CompleXity Networks, UNamur
• How can we measure local temporal
behaviour of the hashtag diffusion?
• Is there a difference in the dynamics
between popular and less used
hashtags?
• Can we measure (and predict) collective
attention by the hashtag dynamics?
5. Key Results: Local Variation
Local Variation Spike Trains 4C. Sanli, CompleXity Networks, UNamur
0
5
10
15
20
25
30
15
20
25
30
<p>=91127
<p>=18553
<p>= 1678
<p>= 318
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
<p>=91127
<p>=18553
<p>= 1678
<p>= 318
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
P()LVP()LV
Real activity(a)
Random activity(b)
0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
2
2.5
3
LV(t1)
LV(t2)
0.4
0.6
0.8
1
(LV(t1),LV(t2))
(a)
(b)
<p
<p
<p
<p
<p
<p
<p
<p
<p
<p
bursty regular
0
5
10
15
20
0 1 2 3 4 5
0
5
10
15
20
25
30
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
<p>=91127
<p>=18553
<p>= 1678
<p>= 318
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
LV
P()LVP()LV
Random activity(b)
FIG. 7. Probability density function (PDF) of the local vari-
ation LV of real hashtag propagation (a) and random hash-
tag time sequence (b). Two distinct shapes are visible: (a)
From high p to low p, the peak position of P(LV ) shifts from
low values of LV to higher values of LV . (b) P(LV ) always
0
5
10
15
20
0 1 2 3 4 5
0
5
10
15
20
25
30
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
<p>=91127
<p>=18553
<p>= 1678
<p>= 318
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
LV
P()LVP()LV
Random activity(b)
FIG. 7. Probability density function (PDF) of the local vari-
ation LV of real hashtag propagation (a) and random hash-
tag time sequence (b). Two distinct shapes are visible: (a)
From high p to low p, the peak position of P(LV ) shifts from
low values of LV to higher values of LV . (b) P(LV ) always
0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
LV(t1)
LV(t2
101
102
103
104
105
0
0.2
0.4
0.6
0.8
1
<p>
r(LV(t1),LV(t2))
(b)
<
<
<
<
<
bursty regular
FIG. 8. Linear correlation of LV through real hashta
trains. (a) The linear relation of the first and the
halves of the empirical spike trains, LV (t1) and LV
spectively, are investigated. The legend ranks hpi in d
colors and symbols. (b) The Pearson correlation co
r(LV (t1), LV (t2)) between these quantities show tha
the temporal correlation through moderately popular
is maximum, r reaches the minimum values for both
• hashtag dynamics • artificial dynamics
• collective attention
• C. Sanlı and R. Lambiotte, PLoS ONE 10(7): e0131704 (2015).
6. Hashtag Spike Train
Local Variation Spike Trains 5C. Sanli, CompleXity Networks, UNamur
time
count
t t0 f
f a time delay between successive events, inter-event interva
dent events, the distrubution of inter-event interval is Poissonia
are observed and therefore forward propogation of a signal is a
ry. Thus, quantifying ⌧ is crucial.
is an alternative way to characterize whether a time series is P
For a stationarly process, Lv is a ratio of the di↵erence be
of forward event and the inter-event interval of backward eve
ent intervals. Suppose that a signal propogates in distinct tim
. . . ⌧N . Then, at ⌧i, the inter-event interval of forward event i
r-event interval of backward event is ⌧i = ⌧i ⌧i 1. Conseq
0 50 100 150 200 250 300 350
0
5
1
τh
(hour)
h
< r
h
>= 41
< r
h
>= 35
es versus life time (⌧h) of the corresponsi
domly selected #hashtag activity from r
τ
h
(hour)
FIG. 2. Rank of #hashtag rh versus life ti
2
. . .
τ
h
(hour)
FIG. 2. Rank of #hashtag rh versus life ti
2
hpi =
p
7. Control Parameters
Local Variation Spike Trains 6C. Sanli, CompleXity Networks, UNamur
t
t
0
f
150 200 250 300 350
τh
(hour)
h
< r
h
>= 35
time (⌧h) of the corresponsing
d #hashtag activity from real
τ
h
(hour)
ashtag rh versus life time of #hashtag ⌧h.
2
• .
• .
• .
• .
• .
: number of spikes
: initial time
: final time
: life time
= popularity
τ
h
(hour)
ashtag rh versus life time of #hashtag ⌧h.
2
hpi =
p
clude interaction among agents. Considering online social
ess self-organized optimizing of popularity of information.
of #hashtag propogation, user activity, and user #hashtag
y between successive events, inter-event interval ⌧, is a
he distrubution of inter-event interval is Poissonian. If not,
and therefore forward propogation of a signal is a function
ntifying ⌧ is crucial.
ive way to characterize whether a time series is Poissonian
arly process, Lv is a ratio of the di↵erence between the
: inter-hashtag spike interval
8. Circadian Pattern and Local Signal
Local Variation Spike Trains 7C. Sanli, CompleXity Networks, UNamur
9. Driving Factors in our Twitter Network
Local Variation Spike Trains 8C. Sanli, CompleXity Networks, UNamur
1. circadian human behaviour (internal)
2. political election (external)
+ complex decision-making (both internal
and external)
10. Data Set
Local Variation Spike Trains 9C. Sanli, CompleXity Networks, UNamur
• 9 days of the French election 2012 (May 5th),
• total activity ~ 10 million, hashtag activity~ 3
million, unique hashtags ~ 300.000,
•
!
!
!
• number of total users ~ 475.000,
• number of users tweet or retweet any
hashtags at least ones ~ 230.000.
#ledebat 180946
#hollande 143636
#sarkozy 116906
#votehollande 99908
#france2012 20635
#fh2012 67759
11. Top Most Used Hashtags
Local Variation Spike Trains 10C. Sanli, CompleXity Networks, UNamur
DAILY CYCLE OF #HASHTAGS
00:0012:0000:0012:0000:0012:0000:0012:0000:0012:0000:0012:0000:00
0
10
20
30
40
50
60
70
80
hour
count/min.
#ledebat
#hollande
#sarkozy
#votehollande
#fh2012
#france2012
debate election
12. Local Variation Spike Trains 11C. Sanli, CompleXity Networks, UNamur
Statistics of Hashtags
13. Heterogeneity in Popularity
Local Variation Spike Trains 12C. Sanli, CompleXity Networks, UNamur
3
popu-
quiva-
d the
5,697
f-plot
htags
= 1.
n the
f Fig.
old of
5% of
nally,
ctan-
than
10
0
10
1
10
2
10
3
10
4
10
5
10
6
10
0
10
1
10
2
10
3
10
4
10
5
rank hashtag
popularity:p 10
6
(a)
P(p)
83%
0.15%
0.0001% 60%
• C. Sanlı and R. Lambiotte, PLoS ONE 10(7): e0131704 (2015).
14. Heterogeneity in Time
Local Variation Spike Trains 13C. Sanli, CompleXity Networks, UNamur
10−1 100 101 102 10310−3
10−2
10−1
100
101
102
103
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
<p>= 11
<p>= 2
0.8
0.85
0.9
∆τ (hour)
P()∆τCDF
(b)
12 hours
1 day
2 days
3 days
FIG. 3. The cumulative (a), CDF( ⌧), and probability (b),
P( ⌧), distributions of the inter-hashtag spike intervals. We
#hash2#hash3
merged
#hash
ficial
ash
10−3
10−2
10−1
100
101
102
103
<p>=91127
<p>=18553
<p>= 1678
<p>= 318
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
<p>= 11
<p>= 2
0.8
0.85
0.9
0.95
1
P()∆τCDF()∆τ
(a)
(b)
12 hours
1 day
2 days
3 days
#hash1#hash2#hash3
merged
#hash
2
103
<p>=91127
<p>=18553
<p>= 1678
<p>= 318
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
<p>= 11
<p>= 2
0.8
0.85
0.9
0.95
1
CDF()∆τ
popular hashtags
• C. Sanlı and R. Lambiotte, PLoS ONE 10(7): e0131704 (2015).
15. Local Variation Spike Trains 14C. Sanli, CompleXity Networks, UNamur
Local Analysis on Hashtag
Spike Trains
16. Local Variation
Local Variation Spike Trains 15C. Sanli, CompleXity Networks, UNamur
time
count
For a stationarly process, Lv is a
terval of forward event and the int
these inter-event intervals. Suppose
⌧1 . . . , ⌧i 1, ⌧i, ⌧i+1, . . . ⌧N . Then, at
⌧i+1 ⌧i and the inter-event interval o
is
For a stationarly process, Lv is a
terval of forward event and the inter
these inter-event intervals. Suppose
⌧1 . . . , ⌧i 1, ⌧i, ⌧i+1, . . . ⌧N . Then, at ⌧i
⌧i+1 ⌧i and the inter-event interval of
is
✓
For a stationarly process, Lv is a r
terval of forward event and the inter
these inter-event intervals. Suppose t
⌧1 . . . , ⌧i 1, ⌧i, ⌧i+1, . . . ⌧N . Then, at ⌧i,
⌧i+1 ⌧i and the inter-event interval of b
is
N 1X ✓
For a stationarly process, Lv is a
terval of forward event and the int
these inter-event intervals. Suppose
⌧1 . . . , ⌧i 1, ⌧i, ⌧i+1, . . . ⌧N . Then, at ⌧
⌧i+1 ⌧i and the inter-event interval o
is
N 1X ✓
a stationarly process, Lv is a ratio of the di↵erenc
of forward event and the inter-event interval of ba
nter-event intervals. Suppose that a signal propog
⌧i 1, ⌧i, ⌧i+1, . . . ⌧N . Then, at ⌧i, the inter-event inter
⌧i and the inter-event interval of backward event is ⌧
N 1X ✓ ◆2
ocess, Lv is a ratio of the di↵erence between the inter-event in-
and the inter-event interval of backward event to the sum of
als. Suppose that a signal propogates in distinct time such as
⌧N . Then, at ⌧i, the inter-event interval of forward event is ⌧i+1 =
vent interval of backward event is ⌧i = ⌧i ⌧i 1. Consequently, Lv
2
N 1X
i=2
✓
(⌧i+1 ⌧i) (⌧i ⌧i 1)
(⌧i+1 ⌧i) + (⌧i ⌧i 1)
◆2
=
✓
⌧i+1 ⌧i
⌧i+1 + ⌧i
◆2
. (1)
earance of a time series in distinct times. Multiple activity in same
• K. Miura et al. Neural Computation 18, 2359-2386 (2006).
• S. Shinomoto et al. Neural Computation 15, 2823-2842 (2003).
from the nonstationarity of the hashtag propagation. Simila
sed on this distribution, such as its variance or Fano factor,
similar way. For this reason, we consider here the so-called l
y defined to determine intrinsic temporal dynamics of neuro
].
antities such as P( ⌧), LV compares temporal variations w
specifically defined for nonstationary processes [27]
LV =
3
N 2
N 1X
i=2
✓
(⌧i+1 ⌧i) (⌧i ⌧i 1)
(⌧i+1 ⌧i) + (⌧i ⌧i 1)
◆2
he total number of spikes and . . ., ⌧i 1, ⌧i, ⌧i+1, . . . represent
e of a single hashtag spike train. Eq. 1 also takes the form [
LV =
3
N 2
N 1X
i=2
✓
⌧i+1 ⌧i
⌧i+1 + ⌧i
◆2
17. Limits of Local Variation
Local Variation Spike Trains 16C. Sanli, CompleXity Networks, UNamur
v =
3
N 2
N 1X
i=2
✓
(⌧i+1 ⌧i) (⌧i ⌧i
(⌧i+1 ⌧i) + (⌧i ⌧i
total appearance of a time series in d
any #hashtags at least ones = 2
tion 2012
one and tweets of any language
c-
p
1
c-
n
y
s,
,
s
y,
-
LV =
3
N 2
N 1X
i=2
✓
(⌧i+1 ⌧i) (⌧i ⌧i 1)
(⌧i+1 ⌧i) + (⌧i ⌧i 1)
◆2
(1)
Here, N is the total number of spikes and . . ., ⌧i 1, ⌧i,
⌧i+1, . . . represents successive time sequence of a single
hashtag spike train. Eq. 1 also takes the form [27]
LV =
3
N 2
N 1X
i=2
✓
⌧i+1 ⌧i
⌧i+1 + ⌧i
◆2
(2)
where ⌧i+1 = ⌧i+1 ⌧i and ⌧i = ⌧i ⌧i 1. ⌧i+1 quan-
tifies forward delay and ⌧i represents backward waiting
time for an event at ⌧i. Importantly, the denominator
normalizes the quantity such as to account for local vari-
ations of the rate at which events take place. By defini-
tion, LV takes values in the interval [0:3].
The local variation L presents properties making it
time
count
time
count
bursty >1
regular <1
random=1
18. Classification of Spike Trains
Local Variation Spike Trains 17C. Sanli, CompleXity Networks, UNamur
• S. Shinomoto et al. PLoS Comput. Biol. 15, 2823-2842 (2003).
e collected from awake,
our of the 15 areas were
sets were generated in
used to record neuronal
nter-trial intervals. All
perimentation were in
onal Institutes of Health
ent committee at the
nts were performed.
d spike train for each
task trial periods and
ring rate differs greatly.
000 ISIs, or those with
ignored; 1,307 neurons
computed for the entire
for each neuron. They
for analyzing fractional
n individual neuron was
mputed for 20 fractional
the spike data.
In comparison with Cv, local metrics, such as Lv, SI, Cv2, and IR,
detect firing irregularities fairly invariantly with firing rate
fluctuations. However, these metrics are still somewhat dependent
on firing rate fluctuations. Assuming that rate dependence is
caused by the refractory period that follows a spike, we can
Figure 1. Spike sequences that have identical sets of inter-
spike intervals. Intervals are aligned (A) in a regular order, (B)
randomly, and (C) alternating between short and long.
doi:10.1371/journal.pcbi.1000433.g001
ploscompbiol.org 2 July 2009 | Volume 5 | Issue 7 | e1000433
ts were performed.
d spike train for each
task trial periods and
ing rate differs greatly.
000 ISIs, or those with
ignored; 1,307 neurons
computed for the entire
for each neuron. They
or analyzing fractional
individual neuron was
puted for 20 fractional
he spike data.
Figure 1. Spike sequences that have identical sets of inter-
spike intervals. Intervals are aligned (A) in a regular order, (B)
randomly, and (C) alternating between short and long.
doi:10.1371/journal.pcbi.1000433.g001
oscompbiol.org 2 July 2009 | Volume 5 | Issue 7 | e1000433
dual neurons can be
nalyzed differences in
rtical areas and found a
t closely corresponded
area; neuronal firing is
er-order motor areas,
prefrontal area. Thus,
al areas that may be
utations.
collected from awake,
ur of the 15 areas were
sets were generated in
sed to record neuronal
nter-trial intervals. All
erimentation were in
nal Institutes of Health
nt committee at the
ts were performed.
d spike train for each
task trial periods and
ng rate differs greatly.
000 ISIs, or those with
ignored; 1,307 neurons
computed for the entire
for each neuron. They
instantaneous ISI variability, SI, the geometric average of the
rescaled cross-correlation of ISIs [37,38], Cv2, the coefficient of
variation for a sequence of two ISIs [39], and IR, the difference of
the log ISIs [34] were also used.
Figure 1 displays three types of spike sequences comprising
identical sets of exponentially distributed ISIs. In terms of the ISI
distributions, all of these are regarded as Poisson processes,
accordingly Cv values are all identical at 1. However, these
sequences clearly differ in how their ISIs are arranged; Lv may be
able to detect these differences.
In comparison with Cv, local metrics, such as Lv, SI, Cv2, and IR,
detect firing irregularities fairly invariantly with firing rate
fluctuations. However, these metrics are still somewhat dependent
on firing rate fluctuations. Assuming that rate dependence is
caused by the refractory period that follows a spike, we can
LV =
N 2 i=2
(⌧i+1 ⌧i) + (⌧i ⌧i 1)
e, N is the total number of spikes and . . ., ⌧i 1, ⌧i, ⌧i+1, . . . represen
e sequence of a single hashtag spike train. Eq. 1 also takes the form [
LV =
3
N 2
N 1X
i=2
✓
⌧i+1 ⌧i
⌧i+1 + ⌧i
◆2
ere ⌧i+1 = ⌧i+1 ⌧i and ⌧i = ⌧i ⌧i 1. ⌧i+1 quantifies forward
resents backward waiting time for an event at ⌧i. Importantly, the de
malizes the quantity such as to account for local variations of the rat
nts take place. By definition, LV takes values in the interval [0:3].
The local variation LV presents properties making it an interesting can
lysis of hashtag spike trains [23–27]. In particular, LV is on average eq
random process is either a stationary or a non-stationary Poisson pro
only condition that the time scale over which the firing rate ⇠(t) fluct
n the typical time between spikes. Deviations from 1 originate from l
relations in the underlying signal, either under the form of pairwise c
=1.4
Unlike quantities such as P( ⌧), LV compares temporal variations w
es and is specifically defined for nonstationary processes [27]
LV =
3
N 2
N 1X
i=2
✓
(⌧i+1 ⌧i) (⌧i ⌧i 1)
(⌧i+1 ⌧i) + (⌧i ⌧i 1)
◆2
e, N is the total number of spikes and . . ., ⌧i 1, ⌧i, ⌧i+1, . . . represen
e sequence of a single hashtag spike train. Eq. 1 also takes the form [
LV =
3
N 2
N 1X
i=2
✓
⌧i+1 ⌧i
⌧i+1 + ⌧i
◆2
ere ⌧i+1 = ⌧i+1 ⌧i and ⌧i = ⌧i ⌧i 1. ⌧i+1 quantifies forward
resents backward waiting time for an event at ⌧i. Importantly, the de
malizes the quantity such as to account for local variations of the rat
nts take place. By definition, LV takes values in the interval [0:3].
The local variation LV presents properties making it an interesting can
lysis of hashtag spike trains [23–27]. In particular, LV is on average eq
random process is either a stationary or a non-stationary Poisson pro
=1.0
cted in a similar way. For this reason, we consider here the so-called
, originally defined to determine intrinsic temporal dynamics of neuro
ns [23–27].
Unlike quantities such as P( ⌧), LV compares temporal variations w
es and is specifically defined for nonstationary processes [27]
LV =
3
N 2
N 1X
i=2
✓
(⌧i+1 ⌧i) (⌧i ⌧i 1)
(⌧i+1 ⌧i) + (⌧i ⌧i 1)
◆2
e, N is the total number of spikes and . . ., ⌧i 1, ⌧i, ⌧i+1, . . . represen
e sequence of a single hashtag spike train. Eq. 1 also takes the form [
LV =
3
N 2
N 1X
i=2
✓
⌧i+1 ⌧i
⌧i+1 + ⌧i
◆2
ere ⌧i+1 = ⌧i+1 ⌧i and ⌧i = ⌧i ⌧i 1. ⌧i+1 quantifies forward
resents backward waiting time for an event at ⌧i. Importantly, the de
malizes the quantity such as to account for local variations of the rat
=0.1
bursty
irregular!
(random)
regular
19. Analysis of Hashtag Spike Trains
Local Variation Spike Trains 18C. Sanli, CompleXity Networks, UNamur
time
#hash1
time
time
time
#hash2#hash3
merged
#hash
20. Generating Random Spike Trains
Local Variation Spike Trains 19C. Sanli, CompleXity Networks, UNamur
time
merged
#hash
htags. Each activity time gives us a spike in a merged
we set the appearance to 1 even though we observe mu
in a second.
mization is satisfied by a random permutation. We us
f MatLab such as randperm(T, p). Here, T represents th
pike train which we keep all information of real #hash
rank, the total exact appearance, of individual #hashtag
uniformly distributed unique numbers out of T. For ou
5
time
artificially
generated
#hash
B. Local variable of #hashtag s
For a stationarly process, Lv is a
terval of forward event and the int
these inter-event intervals. Suppose
⌧1 . . . , ⌧i 1, ⌧i, ⌧i+1, . . . ⌧N . Then, at
⌧i+1 ⌧i and the inter-event interval o
is
B. Local variable of #hashtag sp
For a stationarly process, Lv is a
terval of forward event and the inter
these inter-event intervals. Suppose
⌧1 . . . , ⌧i 1, ⌧i, ⌧i+1, . . . ⌧N . Then, at ⌧i
⌧i+1 ⌧i and the inter-event interval of
is
B. Local variable of #hashtag sp
For a stationarly process, Lv is a r
terval of forward event and the inter
these inter-event intervals. Suppose t
⌧1 . . . , ⌧i 1, ⌧i, ⌧i+1, . . . ⌧N . Then, at ⌧i,
⌧i+1 ⌧i and the inter-event interval of b
is
B. Local variable of #hashtag s
For a stationarly process, Lv is a
terval of forward event and the int
these inter-event intervals. Suppose
⌧1 . . . , ⌧i 1, ⌧i, ⌧i+1, . . . ⌧N . Then, at
⌧i+1 ⌧i and the inter-event interval o
is
Local variable of #hashtag spike trains
a stationarly process, Lv is a ratio of the di↵erenc
of forward event and the inter-event interval of b
nter-event intervals. Suppose that a signal propog
⌧i 1, ⌧i, ⌧i+1, . . . ⌧N . Then, at ⌧i, the inter-event inter
⌧i and the inter-event interval of backward event is ⌧
[ ]
=
21. Distribution of Local Variation
Local Variation Spike Trains 20C. Sanli, CompleXity Networks, UNamur
0
5
10
15
20
25
30
15
20
25
30
<p>=91127
<p>=18553
<p>= 1678
<p>= 318
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
<p>=91127
<p>=18553
<p>= 1678
<p>= 318
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
P()LVP()LV
Real activity(a)
Random activity(b)
0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
2
2.5
3
LV(t1)
LV(t2)
0.4
0.6
0.8
1
(LV(t1),LV(t2))
(a)
(b)
<p
<p
<p
<p
<p
<p
<p
<p
<p
<p
bursty regular
0
5
10
15
20
0 1 2 3 4 5
0
5
10
15
20
25
30
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
<p>=91127
<p>=18553
<p>= 1678
<p>= 318
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
LV
P()LVP()LV
Random activity(b)
FIG. 7. Probability density function (PDF) of the local vari-
ation LV of real hashtag propagation (a) and random hash-
tag time sequence (b). Two distinct shapes are visible: (a)
From high p to low p, the peak position of P(LV ) shifts from
low values of LV to higher values of LV . (b) P(LV ) always
0
5
10
15
20
0 1 2 3 4 5
0
5
10
15
20
25
30
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
<p>=91127
<p>=18553
<p>= 1678
<p>= 318
<p>= 174
<p>= 117
<p>= 86
<p>= 68
<p>= 56
<p>= 47
<p>= 41
<p>= 35
LV
P()LVP()LV
Random activity(b)
FIG. 7. Probability density function (PDF) of the local vari-
ation LV of real hashtag propagation (a) and random hash-
tag time sequence (b). Two distinct shapes are visible: (a)
From high p to low p, the peak position of P(LV ) shifts from
low values of LV to higher values of LV . (b) P(LV ) always
0 0.5 1 1.5 2 2.5 3
0
0.5
1
1.5
LV(t1)
LV(t2
101
102
103
104
105
0
0.2
0.4
0.6
0.8
1
<p>
r(LV(t1),LV(t2))
(b)
<
<
<
<
<
bursty regular
FIG. 8. Linear correlation of LV through real hashta
trains. (a) The linear relation of the first and the
halves of the empirical spike trains, LV (t1) and LV
spectively, are investigated. The legend ranks hpi in d
colors and symbols. (b) The Pearson correlation co
r(LV (t1), LV (t2)) between these quantities show tha
the temporal correlation through moderately popular
is maximum, r reaches the minimum values for both
• hashtag dynamics • artificial dynamics
• C. Sanlı and R. Lambiotte, PLoS ONE 10(7): e0131704 (2015).
22. Statistics of Local Variation
Local Variation Spike Trains 21C. Sanli, CompleXity Networks, UNamur
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
real
random
µ()LV
real hashtags:
decay in µ
with increasing p
30
40
50
es
µ (LV) = 1
real
random
0
(a)
1 2 3 4 5 6
0
0.2
0.4
0.6
0.8
log (<p>)10
real hashtags:
decay in µ
with increasing p
−20
−10
0
10
20
30
40
50z−values
µ (LV) = 1
real
random
0
µ0µ =
(a)
(b)
bursty
regular
• C. Sanlı and R. Lambiotte, PLoS ONE 10(7): e0131704 (2015).
23. Local Variation Spike Trains 22C. Sanli, CompleXity Networks, UNamur
Local Variation of Collective Attention
24. Trains in Collective Attention
Local Variation Spike Trains 23C. Sanli, CompleXity Networks, UNamur
8 am 12 pm 4 pm 8 pm 0 am4 am0 am
hour
debate day
regular day
election day
hashtagspiketrains:#ledebat
• #ledebat: debate at
7-11 pm
25. Local Variation of Collective Attention
Local Variation Spike Trains 24C. Sanli, CompleXity Networks, UNamur
0
0.5
1
1.5
2
2.5
3
#ledebat
#hollande
#sarkozy
#votehollande
#avecsarkozy
#ledebat
(a) debate day
0 am 3 am 6 am 9 am 12 pm3 pm 6 pm 9 pm 0 am
0
500
1000
1500
2000
2500
3000
3500
4000
(b)
0 am 3 am 6 am 9 am 12 pm3 pm 6 pm 9 pm 0 am
#hollande
#sarkozy
0 am 3 am 6 am 9 am 12 pm3 pm 6 pm 9 pm 0 am
regular day election day
#ledebat
#hollande
#sarkozy
#votehollande
#avecsarkozy
#ledebat
hour hour hour
L(t)Vtweetcount/hour
announcement
of the result
at 7 pm
debate
at 7-11 pm
26. What we understand
Local Variation Spike Trains 25C. Sanli, CompleXity Networks, UNamur
• Local variation is very simple, but powerful to
quantify local temporal characteristics of complex
hashtag dynamics.
• There is a direct relation between the popularity and
the inter-hashtag spike interval: While popular
hashtags present regular activation, less used
hashtags indicate bursty spiking.
• Collective attention can be determined by the local
variation: From random irregular spiking to regular
signal. Further detail analysis will help to predict
collective attention.
27. • Online analysis tool for the optimisation of social
media campaigns (EU project),
Acknowledgement
Local Variation Spike Trains 26C. Sanli, CompleXity Networks, UNamur
ceday
http://fcx!
Month 6 General Meeting
htt
• Restricted amount of sources forces social
and physical systems to present
emergence of order.
• Twitter users want to spread their messages and
beads under driving want to be mobile. As a
result, the twitter users collectively advertise and
the beads form groups to move together. Both
systems self-organize and create dynamic
heterogeneity.
Therefore, the interpretation of the dynamic
heterogeneity of the beads in a critical limit
would help to characterize viral memes
(#hashtags) in twitter.
Refs:
1 H. Simon (1971).
2 L. Weng et al. (2012).
3 J. P. Gleeson et al. (2014).
0 12 24 36 48 60 72 84
0
10
20
30
40
time (hours)
numberoftweets/unitt
!
Month 6 General Meeting
beads under driving want to be
result, the twitter users collecti
the beads form groups to mov
systems self-organize and cre
heterogeneity.
Therefore, the interpretatio
heterogeneity of the beads in
would help to characterize
(#hashtags) in twit
Refs:
1 H. Simon (1971).
2 L. Weng et al. (2012).
3 J. P. Gleeson et al. (2014).
0 12 24 36 48 60 72 84
0
10
time (hours)
number
• FNRS (le Fonds de la Recherche Scientifique),
Wallonie, Belgium.
• Takaaki Aoki (Kagawa University, Japan) and
Taro Takaguchi (Nat. Inst. of Informatics, Japan),