Humans prolifically engage in mental time travel. We dwell on past actions and experience satisfaction or regret. More than storytelling, these recollections change how we act in the future and endow us with a computationally important ability to link actions and consequences across spans of time, which helps address the problem of long-term credit assignment: the question of how to evaluate the utility of actions within a long-duration behavioral sequence. Existing approaches to credit assignment in AI cannot solve tasks with long delays between actions and consequences. Here, we introduce a paradigm where agents use recall of specific memories to credit past actions, allowing them to solve problems that are intractable for existing algorithms. This paradigm broadens the scope of problems that can be investigated in AI and offers a mechanistic account of behaviors that may inspire models in neuroscience, psychology, and behavioral economics.

The automatic parking is being massively developed by car manufacturers and providers. Until now, there are two problems with the automatic parking. First, there is no openly-available segmentation labels of parking slot on panoramic surround view (PSV) dataset. Second, how to detect parking slot and road structure robustly. Therefore, in this paper, we build up a public PSV dataset. At the same time, we proposed a highly fused convolutional network (HFCN) based segmentation method for parking slot and lane markings based on the PSV dataset. A surround-view image is made of four calibrated images captured from four fisheye cameras. We collect and label more than 4,200 surround view images for this task, which contain various illuminated scenes of different types of parking slots. A VH-HFCN network is proposed, which adopts an HFCN as the base, with an extra efficient VH-stage for better segmenting various markings. The VH-stage consists of two independent linear convolution paths with vertical and horizontal convolution kernels respectively. This modification enables the network to robustly and precisely extract linear features. We evaluated our model on the PSV dataset and the results showed outstanding performance in ground markings segmentation. Based on the segmented markings, parking slots and lanes are acquired by skeletonization, hough line transform and line arrangement.

No abstract

Citrus is one of the most widely cultivated fruit in the world. However, citrus diseases are becoming more and more serious, which has caused substantial economic losses to citrus growers. With the rapid developments of mobile device, mobile services computing play an increasingly important role in our daily lives. How to develop an intelligent diagnosis system for citrus diseases based on mobile services computing and bridge the gap between citrus growers and plant diagnostic experts is worth studying. In this paper, we build an image dataset of six kinds of citrus diseases with the help of experts and realize an intelligent diagnosis system for citrus diseases by constructing the simplified densely connected convolutional networks (DenseNet). The system is realized using the WeChat applet in the mobile device, with which users can upload images and receive diagnostic results and comments. The experimental results show that the recognition accuracy of citrus diseases exceeds 88% and the predict time consumption has also been reduced by simplifying the structure of the DenseNet.

Radar returned signals are processed by stretch processing, resulting in mixer outputs. Based on the time-frequency decomposition of the cross S-method (CSM) of two adjacent mixer outputs, a range-spread target detector is proposed in this paper. As a preparatory work, we propose a signal synthesis method (SSM) based on the singular value decomposition. The SSM synthesizes two signals in their normalized forms from their cross Wigner distribution (CWD) and concentrates their energy on two singular values. This detector consists of three steps. First, we derive the CSM from the S-method (SM). The CSM is close to the sum of the CWDs of the components in one mixer output and their counterparts in the other. Second, we can decompose the CSM by the SSM, thereby obtaining singular values. Third, the time-frequency decomposition feature, i.e., the ratio of the sum of several biggest singular values to the median or mean of the rest, is defined to demonstrate the concentration of the singular values and used to detect the range-spread target. The proposed detector is evaluated by the raw radar data without range migration correction. Results show that it outperforms the conventional detectors. In addition, we prove that the proposed detector has the constant false-alarm rate (CFAR) property.Index Terms-Cross S-method (CSM), constant false-alarm rate (CFAR), range-spread target detector, signal decomposition, singular values.