Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

What is Twitter

Social networking websites allow users to create and share content. Big information cascades of post resharing can form as users of these sites reshare others' posts with their friends and followers. One of the central challenges in understanding such cascading behaviors is in forecasting information outbreaks, where a single post becomes widely popular by being reshared by many users. In this paper, we focus on predicting the final number of reshares of a given post. We build on the theory of self-exciting point processes to develop a statistical model that allows us to make accurate predictions. Our model requires no training or expensive feature engineering. It results in a simple and efficiently computable formula that allows us to answer questions, in real-time, such as: Given a post's resharing history so far, what is our current estimate of its final number of reshares? Is the post resharing cascade past the initial stage of explosive growth? And, which posts will be the most reshared in the future?We validate our model using one month of complete Twitter data and demonstrate a strong improvement in predictive accuracy over existing approaches. Our model gives only 15% relative error in predicting final size of an average information cascade after observing it for just one hour.

/pdf/seismic-a-self-exciting-point-process-model-for-predicting-1wchxwsfms.pdf

SEISMIC: A Self-Exciting Point Process Model for Predicting Tweet Popularity

Dynamics of information diffusion and its applications on complex networks

Text summarization is the process of automatically creating a shorter version of one or more text documents. It is an important way of finding relevant information in large text libraries or in the Internet. Essentially, text summarization techniques are classified as Extractive and Abstractive. Extractive techniques perform text summarization by selecting sentences of documents according to some criteria. Abstractive summaries attempt to improve the coherence among sentences by eliminating redundancies and clarifying the contest of sentences. In terms of extractive summarization, sentence scoring is the technique most used for extractive text summarization. This paper describes and performs a quantitative and qualitative assessment of 15 algorithms for sentence scoring available in the literature. Three different datasets (News, Blogs and Article contexts) were evaluated. In addition, directions to improve the sentence extraction results obtained are suggested.

Assessing sentence scoring techniques for extractive text summarization

Ranking nodes by their spreading ability in complex networks is a fundamental problem which relates to wide applications. Local metric like degree centrality is simple but less effective. Global metrics such as betweenness and closeness centrality perform well in ranking nodes, but are of high computational complexity. Recently, to rank nodes effectively and efficiently, a semi-local centrality measure has been proposed as a tradeoff between local and global metrics. However, in semi-local centrality, only the number of the nearest and the next nearest neighbors of a node is taken into account, while the topological connections among the neighbors are neglected. In this paper, we propose a local structural centrality measure which considers both the number and the topological connections of the neighbors of a node. To evaluate the performance of our method, we use the Susceptible–Infected–Recovered (SIR) model to simulate the epidemic spreading process on both artificial and real networks. By measuring the rank correlation between the ranked list generated by simulation results and the ones generated by centrality measures, we show that our method can rank the spreading ability of nodes more accurately than centrality measures such as degree, k -shell, betweenness, closeness and local centrality. Further, we show that our method can better distinguish the spreading ability of nodes.

Ranking the spreading ability of nodes in complex networks based on local structure

Popularity prediction on microblogging platforms aims to predict the future popularity of a message based on its retweeting dynamics in the early stages. Existing works mainly focus on exploring effective features for prediction, while ignoring the underlying arrival process of retweets. Also, the effect of user activity variation on the retweeting dynamics in the early stages has been neglected. In this paper, we propose an extended reinforced Poisson process model with time mapping process to model the retweeting dynamics and predict the future popularity. The proposed model explicitly characterizes the process through which a message gain its retweets, by capturing a power-law temporal relaxation function corresponding to the aging in the ability of the message to attract new retweets and an exponential reinforcement mechanism characterizing the "richer-get-richer" phenomenon. Further, we introduce the notation of weibo time and integrate a time mapping process into the proposed model to eliminate the effect of user activity variation. Extensive experiments on two Weibo datasets, with 10K and 18K messages respectively, well demonstrate the effectiveness of our proposed model in popularity prediction.

Modeling and Predicting Retweeting Dynamics on Microblogging Platforms

Predicting popularity of online contents is of remarkable practical value in various business and administrative applications. Existing studies mainly focus on finding the most effective features for prediction. However, some effective features, such as structural features which are extracted from the underlying user network, are hard to access. In this paper, we aim to identify features that are both effective and effortless (easy to obtain or compute). Experiments on Sina Weibo show the effectiveness and effortlessness of the temporal features and satisfying prediction performance can be obtained based on only the temporal features of first 10 retweets.

Effective and effortless features for popularity prediction in microblogging network

The study on influence modeling is to understand the information diffusion and word-of-mouth marketing. Independent Cascade Model (ICM) is the most widely studied theoretical diffusion model. However, until now it is unknown whether ICM matches the real information diffusion in online social networks. In this paper, we demonstrate that ICM cannot model accurately for the structure of information diffusion over real networks through our experiments. Meanwhile, we propose a more suitable diffusion model named Three Steps Cascade Model (TSCM) to simulate information diffusion process in online social networks. We focus on the influence maximization problem under TSCM. First, we show that this optimization problem is NP hard. Then we prove that the greedy algorithm can guarantee an influence spread within 63 % of the optimal value. Finally we devise an efficient algorithm which is scalable for large social networks. The experiment results on large-scale real networks show the robustness and utility of our approach.

Efficient influence maximization under TSCM: a suitable diffusion model in online social networks

We describe a framework for human action retrieval in still web images by verb queries, for instance "phoning". Firstly, we build a group of visual discriminative instances for each action class, called "Exemplarlets". Thereafter we employ Multiple Kernel Learning (MKL) to learn an optimal combination of histogram intersection kernels, each of which captures a state-of-the-art feature channel. Our features include the distribution of edges, dense visual words and feature descriptors at different levels of spatial pyramid. For a new image we can detect the hot-region using a sliding-window detector learnt via MKL. The hotregion can imply latent actions in the image. After the hot-region has been detected, we build a inverted index in the visual search path, which we called Visual Inverted Index (VII). Finally, fusing the visual search path and the text search path, we can get the accurate results either relevant to text or to visual information. We show both the detection and retrieval results on our newly collected dataset of six actions as well as demonstrate improved performance over existing methods.

Shuai Gao

Papers

Ranking the spreading ability of nodes in complex networks based on local structure

Modeling and Predicting Retweeting Dynamics on Microblogging Platforms

Effective and effortless features for popularity prediction in microblogging network

Efficient influence maximization under TSCM: a suitable diffusion model in online social networks

Actions in stillweb images: visualization, detection and retrieval