Extracting automata from neural networks using active learning.

Deep learning is one of the most advanced forms of machine learning. Most modern deep learning models are based on an artificial neural network, and benchmarking studies reveal that neural networks have produced results comparable to and in some cases superior to human experts. However, the generated neural networks are typically regarded as incomprehensible black-box models, which not only limits their applications, but also hinders testing and verifying. In this paper, we present an active learning framework to extract automata from neural network classifiers, which can help users to understand the classifiers. In more detail, we use Angluin's L* algorithm as a learner and the neural network under learning as an oracle, employing abstraction interpretation of the neural network for answering membership and equivalence queries. Our abstraction consists of value, symbol and word abstractions. The factors that may affect the abstraction are also discussed in the paper. We have implemented our approach in a prototype. To evaluate it, we have performed the prototype on a MNIST classifier and have identified that the abstraction with interval number 2 and block size 1 × 28 offers the best performance in terms of F1 score. We also have compared our extracted DFA against the DFAs learned via the passive learning algorithms provided in LearnLib and the experimental results show that our DFA gives a better performance on the MNIST dataset.

Prevalence and spatial distribution patterns of human echinococcosis at the township level in Sichuan Province, China.

BACKGROUND: Echinococcosis is a global zoonotic parasitic disease caused by Echinococcus larvae. This disease is highly endemic in Sichuan Province, China. This study investigates the prevalence and spatial distribution characteristics of human echinococcosis at the township level in Sichuan Province, geared towards providing a future reference for the development of precise prevention and control strategies. METHODS: Human prevalence of echinococcosis was evaluated using the B-ultrasonography diagnostic method in Sichuan Province between 2016 and 2019. All data were collected, collated, and analyzed. A spatial distribution map was drawn to intuitively analyze the spatial distribution features. Eventually, the spatial autocorrelation was specified and local indicators of spatial association (LISA) clustering map was drawn to investigate the spatial aggregation of echinococcosis at the township level in Sichuan Province. RESULTS: The prevalence of echinococcosis in humans of Sichuan Province was 0.462%, among which the occurrence of cystic echinococcosis (CE) was 0.221%, while that of alveolar echinococcosis (AE) was 0.244%. Based on the results of the spatial distribution map, a predominance of echinococcosis in humans decreased gradually from west to east and from north to south. The Global Moran's I index was 0.77 (Z = 32.07, P < 0.05), indicating that the prevalence of echinococcosis in humans was spatially clustered, exhibiting a significant spatial positive correlation. Further, the findings of local spatial autocorrelation analysis revealed that the "high-high" concentration areas were primarily located in some townships in the northwest of Sichuan Province. However, the "low-low" concentration areas were predominantly located in some townships in the southeast of Sichuan Province. CONCLUSIONS: Our findings demonstrated that the prevalence of echinococcosis in humans of Sichuan Province follows a downward trend, suggesting that the current prevention and control work has achieved substantial outcomes. Nevertheless, the prevalence in humans at the township level is widely distributed and differs significantly, with a clear clustering in space. Therefore, precise prevention and control strategies should be formulated for clusters, specifically strengthening the "high-high" clusters at the township level.

Geographic distribution of echinococcosis in Tibetan region of Sichuan Province, China.

BACKGROUND: Echinococcosis is a parasitic zoonosis caused by Echinococcus larvae parasitism causing high mortality. The Tibetan Region of Sichuan Province is a high prevalence area for echinococcosis in China. Understanding the geographic distribution pattern is necessary for precise control and prevention. In this study, a spatial analysis was conducted to explore the town-level epidemiology of echinococcosis in the Sichuan Tibetan Region and to provide guidance for formulating regional prevention and control strategies. METHODS: The study was based on reported echinococcosis cases by the end of 2017, and each case was geo-coded at the town level. Spatial empirical Bayes smoothing and global spatial autocorrelation were used to detect the spatial distribution pattern. Spatial scan statistics were applied to examine local clusters. RESULTS: The spatial distribution of echinococcosis in the Sichuan Tibetan Region was mapped at the town level in terms of the crude prevalence rate, excess hazard and spatial smoothed prevalence rate. The spatial distribution of echinococcosis was non-random and clustered with the significant global spatial autocorrelation (I = 0.7301, P = 0.001). Additionally, five significant spatial clusters were detected through the spatial scan statistic. CONCLUSIONS: There was evidence for the existence of significant echinococcosis clusters in the Tibetan Region of Sichuan Province, China. The results of this study may assist local health departments with developing better prevention strategies and prompt more efficient public health interventions.