Fusing features from different layers is essential to improve the ship target detection ability in the synthetic aperture radar (SAR) images. Mainstream methods usually perform simple addition or concatenation operations on adjacent feature layers without properly merging their semantic and spatial information, whereas the traditional skip connections are unable to explore sufficient information by the same scale. To address these issues, a ship detection network based on adjacent context guide fusion module and dense weighted skip connection (AFDN) in SAR images is proposed: Adjacent context guide fusion module is specially designed to capture the long-range dependencies of high-level features as weights to multiply with low-level features to fuse adjacent features more efficiently. Furthermore, the dual-path enhanced pyramid is constructed to refine and fuse multi-scale features. Finally, a dense weighted skip connection is proposed by weighted fusion of features of all sizes before the decoder to enrich the feature space. Our anchor-free AFDN outputs the spatial density map and clusters to obtain the rotatable bounding box. The experimental results indicate that the method proposed in this paper surpasses previous ship detection methods and achieves high accuracy on SSDD and AIRSARShip-1.0 datasets. INDEX TERMSAdjacent context guide fusion module, dense weighted skip connection, dual-path enhanced pyramid, spatial density map, ship detection, synthetic aperture radar (SAR).

The mainstream Go AI algorithms represented by AlphaZero and KataGo suffer from lowquality samples in the early training period and low exploration efficiency when performing traditional Monte Carlo Tree Search (MCTS). For the shortcomings mentioned above: The variable scale training is proposed, i.e., introducing a variable scale board with boundary conditions of randomly placed stones at the boundary periphery, to pre-train a small-scale network for recommending local move strategy and ownership. This network is used to improve the backbone network's moving policy and state value, enhancing the quality of game samples in the early stages of training. To improve the efficiency and convergence speed of the search, we propose the Parallel Monte Carlo Tree Search with Potential-Upper-Bound (PUB-PMCTS), i.e., executing multiple unevaluated searches sequentially and then evaluating multiple leaf nodes in parallel; also, the variance of the node's action values are used to forecast the potential upper limit of the node. In addition, we add a self-attention mechanism in the network to extract global context information of features and add maximum entropy loss to grow the exploration ability of the model. With the improvements described above, the bot TransGo is designed. Experimental results show that in a 13×13 Go environment, TransGo has more stable performance and higher game level in the early training period compared with other algorithms. After four days of training with TransGo, KataGo, and AlphaZero: TransGo improved by 102 Elo compared to KataGo and over 1000 Elo compared to AlphaZero.