No abstract

This paper proposes a probabilistic method that infers the transport modes and the physical path of trips from smartphone data that were recorded during travels. This method synthesizes multiple kinds of data from smartphone sensors which provide relevant location or transport mode information: GPS, Bluetooth, and Accelerometer. The method is based on a smartphone measurement model that calculates the likelihood of observing the smartphone data in the multimodal transport network. The output of this probabilistic method is a set of candidate true paths, and the probability of each path being the true one. The transport mode used on each arc is also inferred. Numerical experiments include map visualizations of some example trips, and an analysis on the performance of the transport mode inference.

Three-dimensional cell culture techniques can better reflect the in vivo characteristics of tumor cells compared with traditional monolayer cultures. Compared with their 2D counterparts, 3D-cultured tumor cells showed enhanced resistance to the cytotoxic T cell-mediated immune response. However, it remains unclear whether 3D-cultured tumor cells have an enhanced resistance to NK cell cytotoxicity. In this study, a total of 363 differentially expressed proteins were identified between the 2D- and 3D-cultured U251 cells by comparative proteomics, and an immune-associated protein-protein interaction (PPI) network based on these differential proteins was constructed by bioinformatics. Within the network, HLA-E, as a molecule for inhibiting NK cell activation, was significantly up-regulated in the 3D-cultured tumor cells. Then, we found that the 3D-cultured U251 cells exhibited potent resistance to NK cell cytotoxicity in vitro and were prone to tumor formation in vivo. The resistance of the 3D-cultured tumor cells to NK cell lysis was mediated by the HLA-E/NKG2A interaction because the administration of antibodies that block either HLA-E or NKG2A completely eliminated this resistance and significantly decreased tumor formation. Taken together, our findings indicate that HLA-E up-regulation in 3D-cultured cells may result in enhanced tumor resistance to NK cell-mediated immune response.

Zero-shot learning has been a tough problem since no labeled data is available for unseen classes during training, especially for classes with low similarity. In this situation, transferring from seen classes to unseen classes is extremely hard. To tackle this problem, in this paper we propose a self-training based method to efficiently leverage unlabeled data. Traditional self-training methods use fixed heuristics to select instances from unlabeled data, whose performance varies among different datasets. We propose a reinforcement learning framework to learn data selection strategy automatically and provide more reliable selection. Experimental results on both benchmarks and a real-world e-commerce dataset show that our approach significantly outperforms previous methods in zero-shot text classification.

No abstract