Learning Class-imbalanced Logical Rules with Bayesian Networks – The computational time of supervised learning in sparse or high dimensional context is prohibitive, and we propose a novel sparse or high dimensional algorithm, Sparse Logical Rules (SSRC). The algorithm, SSRC, plays a central role in the classification of sparse logic programs by making use of Bayesian networks or a Bayesian network with a hidden layer. The SSRC algorithm learns the logic model by applying Bayesian networks to sparse input data. We evaluate a limited set of sparse logic programs from multiple scenarios in a real-world data lab and show that SSRC can easily handle the large number of data. When compared with the state-of-the-art Sparse Logic Programs (SLLP) algorithms, SSRC has the best performance and leads to the lowest computation time, while being a simple and efficient machine learning algorithm.

We propose a framework for the visualization of data in a deep space. Instead of computing the parameters of the object, we assume that it is an image. We propose a novel method for solving the well-known Image Stereo Problem. This problem can be used to solve complex visualizations of data by utilizing features extracted from the object. As previously formulated, the image has a pixel-depth. However, the depth can be represented by multiple images. We propose a novel concept of a multi-model convolutional CNN architecture, which is capable of visualizing multiple datasets in a convolutional CNN. We first evaluate the performance of the method with the input from a single dataset and compare it with several state-of-the-arts. The method achieves a good average test error of +7.3% and an average test per 100 views on the MNIST dataset. We also demonstrate the method with both synthetic and real data.

A Linear Tempering Paradigm for Hidden Markov Models

On the Existence of a Constraint-Based Algorithm for Learning Regular Expressions

# Learning Class-imbalanced Logical Rules with Bayesian Networks

A Framework for Learning Discrete Event-based Features from Data

A Novel Approach to Video Analysis Using Vector Mean Field and Kernel Density FunctionsWe propose a framework for the visualization of data in a deep space. Instead of computing the parameters of the object, we assume that it is an image. We propose a novel method for solving the well-known Image Stereo Problem. This problem can be used to solve complex visualizations of data by utilizing features extracted from the object. As previously formulated, the image has a pixel-depth. However, the depth can be represented by multiple images. We propose a novel concept of a multi-model convolutional CNN architecture, which is capable of visualizing multiple datasets in a convolutional CNN. We first evaluate the performance of the method with the input from a single dataset and compare it with several state-of-the-arts. The method achieves a good average test error of +7.3% and an average test per 100 views on the MNIST dataset. We also demonstrate the method with both synthetic and real data.