Link Prediction in Social Networks Using Markov Random Field

Link prediction is an important task for social networks analysis, which also has applications in other domains such as information retrieval, recommender systems and e-commerce. The task is related to predicting the probable connection between two nodes in the netwok. These links are subjected to loss because of the improper creation or the lack of reflection of links in the networks; so it`s possible to develop or complete these networks and recycle the lost items and information through link prediction. In order to discover and predict these links we need the information of the nodes in the network. The information are usually extracted from the network`s graph and utilized as factors for recognition. There exist a variety of techniques for link prediction, amongst them, the most practical and current one is supervised learning based approach. In this approach, the link prediction is considered as binary classifier that each pair of nodes can be 0 or 1. The value of 0 indicates no connection between nodes and 1 means that there is a connection between them. In this research, while studying probabilistic graphical models, we use Markov random field (MRF) for link prediction problem in social networks. Experimentl results on Flicker dataset showed the proposed method was better than previous methods in precision and recall.

Download Full-text

Mapping Mouse Brain Slice Sequence to a Reference Brain Without 3D Reconstruction

10.1101/357475 ◽

2018 ◽

Cited By ~ 2

Author(s):

Jing Xiong ◽

Jing Ren ◽

Liqun Luo ◽

Mark Horowitz

Keyword(s):

Random Field ◽

Markov Random Field ◽

Brain Slice ◽

Neural Circuits ◽

Brain Slices ◽

Nonrigid Registration ◽

Histogram Of Oriented Gradients ◽

Markov Random ◽

L2 Norm ◽

Better Than

AbstractHistological brain slices are widely used in neuroscience to study anatomical organization of neural circuits. Since data from many brains are collected, mapping the slices to a reference atlas is often the first step in interpreting results. Most existing methods rely on an initial reconstruction of the volume before registering it to a reference atlas. Because these slices are prone to distortion during sectioning process and often sectioned with nonstandard angles, reconstruction is challenging and often inaccurate. We propose a framework that maps each slice to its corresponding plane in the atlas to build a plane-wise mapping and then perform 2D nonrigid registration to build pixel-wise mapping. We use the L2 norm of the Histogram of Oriented Gradients (HOG) of two patches as the similarity metric for both steps, and a Markov Random Field formulation that incorporates tissue coherency to compute the nonrigid registration. To fix significantly distorted regions that are misshaped or much smaller than the control grids, we trained a context-aggregation network to segment and warp them to their corresponding regions with thin plate spline. We have shown that our method generates results comparable to an expert neuroscientist and is significantly better than reconstruction-first approaches.

Download Full-text