Accurate causal discovery in telecommunication alarm event sequences is crucial for reliable root cause analysis, but presents significant challenges due to complex topological dependencies and inherent alarm redundancy. To address these challenges, we propose the Topological-Temporal Convolution Transformer Hawkes process (TTCTH), a novel framework for learning hidden causal relationships in alarm event sequences. To start with, a topological-temporal graph convolution is designed to initialize a causal graph for mainly exploring local dependencies among event types in the topological and temporal domains. We next design three separable transformer variants of topological transformer, causal transformer and temporal transformer, which are sequentially stacked to globally capture causal dependencies among event types. In particular, the topological transformer fuses prior topological information into alarm sequence representation. The causal transformer models causal dependencies with self-attention mechanism to capture spatially causal relations among event types. The temporal transformer is developed to model long-range causal dependencies across multiple events. We then leverage the high-dimensional hidden state representation, encoded by the stacked transformers, to construct feature evolution with the constraint-based Hawkes process. In order to observe the performance of TTCTH, we conduct extensive experiments on real-world alarm datasets. It demonstrates TTCTH’s excellence in comparison with ten baselines.

Activity prediction aims to predict what activities will occur in the future. In smart homes, to facilitate the daily living of the residents, automated or assistive services are provided. To provide these services, the ability of activity prediction is necessary. When we make a prediction, most of the existing works focus on predicting information about the next activity. However, in a smart home environment, compared with just predicting information about the next activity, another type of activity prediction problem has more practical value: predicting what activities will occur in the coming time period of a certain length. The necessity of this type of prediction is due to the purpose of the smart homes and the character of the activities. Many activities in smart homes need preparation time before being performed. Through this type of prediction, activities could be predicted sufficient time before being performed, and there will be adequate time for the smart home system to prepare corresponding automated or assistive services. As more than one activity could occur within the time period in which the prediction is made, this problem is a multilabel classification problem. In this article, we first give a formulation of the problem. Then, we propose a deep learning model to address it. The proposed model consists of the convolutional part, the long short-term memory layer, and the multilabel output layer. Experiments on real-world datasets show the effectiveness of our model.

The nucleus shape is of greatly importance for disease diagnosis and cell function. In this paper, a new model for nucleus shape is proposed to more accurately represent it, which mainly contains two components: boundary axis representation and weighted least-squares B-spline fitting. Boundary axis representation is achieved by segmenting the nuclear object extracted from original images into two parts with the major axis and recognizing the object boundary of each part. Then weighted least-squares B-spline fitting is employed to obtain two curves to represent the nucleus shape. Experiments and comparisons demonstrate the effectiveness of the proposed model. Moreover, outliers can be discarded by using weighted idea to remarkably improve the fitting accuracy.

For detection of protein expression in biomedicai image, shape measurement of protein expression mostly depends on semi-automatic analysis of image analysis software which makes the results vulnerable to subjective factors, since the automatic analysis is too complicated to operate. Therefore, a novel algorithm based on generalized data field (GDF) is proposed to determine the region of protein expression. Instead of being directly divided into the measured object and background, all the data objects, namely pixels of an image, are naturally clustered into multiple classes based on potential distribution in generalized data field. Each class represents protein expression in different degree, which precisely describes the details of protein expression. Compared with image-pro plus software analysis, KM and EM, experiment results demonstrate that the protein expression can be extracted easily and objectively from an image by GDF. Furthermore, noises of background are eliminated by the smoothing procedure of GDF.