Search

As neural networks are ubiquitous in the real-world environment, we encounter a variety of situations in which neural network-based systems fail. To address these failures, we need to identify their root causes. However, due to the black-box nature of neural networks, it is considered a difficult task to explain/understand internal functions, and cause analysis is usually conducted based on personal experience. To alleviate these problems, we propose a method to compute an element-wise contribution of inputs on current decisions. To this end, we focus on the physical meaning of neuron activations, and formulate an optimization problem that finds partial activations that support current decisions. Then, by accumulating these partial activations in the backward direction, we evaluate the contributions of each element in an input vector to the current decision. Experimental results have shown that the proposed method outperforms conventional methods in terms of cause localization for a range of failure scenarios.

Accurate prediction of scoliotic curve progression is crucial for guiding treatment decisions in adolescent idiopathic scoliosis (AIS). Traditional methods of assessing the likelihood of AIS progression are limited by variability and rely on static measurements. This study developed and validated machine learning models for classifying progressive and non-progressive scoliotic curves based on gait analysis using wearable inertial sensors. Gait data from 38 AIS patients were collected using seven inertial measurement unit (IMU) sensors, and hip–knee (HK) cyclograms representing inter-joint coordination were generated. Various machine learning algorithms, including support vector machine (SVM), random forest (RF), and novel deep convolutional neural network (DCNN) models utilizing multi-plane HK cyclograms, were developed and evaluated using 10-fold cross-validation. The DCNN model incorporating multi-plane HK cyclograms and clinical factors achieved an accuracy of 92% in predicting curve progression, outperforming SVM (55% accuracy) and RF (52% accuracy) models using handcrafted gait features. Gradient-based class activation mapping revealed that the DCNN model focused on the swing phase of the gait cycle to make predictions. This study demonstrates the potential of deep learning techniques, and DCNNs in particular, in accurately classifying scoliotic curve progression using gait data from wearable IMU sensors.

Insulin-like peptides (ILPs) mediate various physiological processes in insects. Specifically, ILP expression is required for immature development in different insects. The western flower thrips, Frankliniella occidentalis, is polyphagous, but its occurrence and population density vary among different hosts. This study assesses the developmental variations in the thrips through quantitative analysis of their ILP expressions. Two types of ILPs (Fo-ILP1 and Fo-ILP2) were identified from the genome of F. occidentalis, and both ILPs were predicted to have the characteristics of signal peptides and B-C-A chains linked by cysteines. A phylogenetic analysis indicates that these two ILPs in the thrips are clustered with the ILP1 of Drosophila melanogaster, suggesting their physiological roles in growth. In addition, the two ILP genes were relatively highly expressed at all feeding stages, but their expression was reduced during the nonfeeding prepupal and pupal stages. Furthermore, RNA interference of each ILP expression led to significant developmental retardation. In validating the ILP expression in the thrips’ development, five different varieties of host hot peppers were assessed in a choice test, along with the immature development of F. occidentalis. The expression levels of the two ILP genes were highly correlated with variations in the immature developmental rates of different hot pepper varieties. These suggest that Fo-ILP1 and Fo-ILP2 mediate the immature development of F. occidentalis by sensing different nutritional values of hot peppers. This study is the first report on ILPs in thysanopteran insects.

Xenorhabdus nematophila is the symbiotic bacterium of an entomopathogenic nematode, Steinernema carpocapsae, When the nematode enters a target insect, the symbiotic bacteria are released into the hemocoel. After inducing host immunosuppression, the bacteria multiply in the hemocoel and cause fatal septicemia. For optimal field application to control insect pests, culturing mass numbers of the nematodes would be costly. In this study, Bacillus thuringiensis (Bt) was chosen as an alternative natural vector, which would be relatively economical for field application. Bt infection of gut epithelium would form a bacterial passage between the gut lumen and hemocoel, which facilitates the orally fed X. nematophila to infect the hemocoel. Diamondback moth, Plutella xylostella (L.) (Lepidoptera: Plutellidae), used in this study was tolerant to Bt because only 10% mortality was noted in response to 2 times higher concentration than recommended for commercial B. t. kurstaki, although this species was susceptible only during early instars. The orally fed X. nematophila caused significant mortality to early instars of P. xylostella, but not late instars. When both X. nematophila and Bt were fed to late instars of P. xylostella, they showed significantly enhanced mortality, in which X. nematophila cells were recovered from the hemocoel of the treated P. xylostella. However, when only X. nematophila was fed, no cells were recovered from the hemolymph. This study suggests that X. nematophila can be applied to control P. xylostella in a mixture with Bt in the field without its nematode host. [ABSTRACT FROM AUTHOR]