Spatial-temporal combination and multi-head flow-attention network for traffic flow prediction.

Traffic flow prediction based on spatial-temporal data plays a vital role in traffic management. However, it still faces serious challenges due to the complex spatial-temporal correlation in nonlinear spatial-temporal data. Some previous methods have limited ability to capture spatial-temporal correlation, and ignore the quadratic complexity problem in the traditional attention mechanism. To this end, we propose a novel spatial-temporal combination and multi-head flow-attention network (STCMFA) to model the spatial-temporal correlation in road networks. Firstly, we design a temporal sequence multi-head flow attention (TS-MFA), in which the unique source competition mechanism and sink allocation mechanism make the model avoid attention degradation without being affected by inductive biases. Secondly, we use GRU instead of the linear layer in traditional attention to map the input sequence, which further enhances the temporal modeling ability of the model. Finally, we combine the GCN with the TS-MFA module to capture the spatial-temporal correlation, and introduce residual mechanism and feature aggregation strategy to further improve the performance of STCMFA. Extensive experiments on four real-world traffic datasets show that our model has excellent performance and is always significantly better than other baselines.

Isolation of cancer stem like cells from human adenosquamous carcinoma of the lung supports a monoclonal origin from a multipotential tissue stem cell.

There is increasing evidence that many solid tumors are hierarchically organized with the bulk tumor cells having limited replication potential, but are sustained by a stem-like cell that perpetuates the tumor. These cancer stem cells have been hypothesized to originate from transformation of adult tissue stem cells, or through re-acquisition of stem-like properties by progenitor cells. Adenosquamous carcinoma (ASC) is an aggressive type of lung cancer that contains a mixture of cells with squamous (cytokeratin 5+) and adenocarcinoma (cytokeratin 7+) phenotypes. The origin of these mixtures is unclear as squamous carcinomas are thought to arise from basal cells in the upper respiratory tract while adenocarcinomas are believed to form from stem cells in the bronchial alveolar junction. We have isolated and characterized cancer stem-like populations from ASC through application of selective defined culture medium initially used to grow human lung stem cells. Homogeneous cells selected from ASC tumor specimens were stably expanded in vitro. Primary xenografts and metastatic lesions derived from these cells in NSG mice fully recapitulate both the adenocarcinoma and squamous features of the patient tumor. Interestingly, while the CSLC all co-expressed cytokeratins 5 and 7, most xenograft cells expressed either one, or neither, with <10% remaining double positive. We also demonstrated the potential of the CSLC to differentiate to multi-lineage structures with branching lung morphology expressing bronchial, alveolar and neuroendocrine markers in vitro. Taken together the properties of these ASC-derived CSLC suggests that ASC may arise from a primitive lung stem cell distinct from the bronchial-alveolar or basal stem cells.