Feature engineering has been the key to the success of many prediction models. However, the process is nontrivial and o en requires manual feature engineering or exhaustive searching. DNNs are able to automatically learn feature interactions; however, they generate all the interactions implicitly, and are not necessarily e cient in learning all types of cross features. In this paper, we propose the Deep & Cross Network (DCN) which keeps the bene ts of a DNN model, and beyond that, it introduces a novel cross network that is more e cient in learning certain bounded-degree feature interactions. In particular, DCN explicitly applies feature crossing at each layer, requires no manual feature engineering, and adds negligible extra complexity to the DNN model. Our experimental results have demonstrated its superiority over the state-of-art algorithms on the CTR prediction dataset and dense classi cation dataset, in terms of both model accuracy and memory usage.

No abstract

Removing specular highlight in an image is a fundamental research problem in computer vision and computer graphics. While various methods have been proposed, they typically do not work well for real‐world images due to the presence of rich textures, complex materials, hard shadows, occlusions and color illumination, etc. In this paper, we present a novel specular highlight removal method for real‐world images. Our approach is based on two observations of the real‐world images: (i) the specular highlight is often small in size and sparse in distribution; (ii) the remaining diffuse image can be represented by linear combination of a small number of basis colors with the sparse encoding coefficients. Based on the two observations, we design an optimization framework for simultaneously estimating the diffuse and specular highlight images from a single image. Specifically, we recover the diffuse components of those regions with specular highlight by encouraging the encoding coefficients sparseness using L0 norm. Moreover, the encoding coefficients and specular highlight are also subject to the non‐negativity according to the additive color mixing theory and the illumination definition, respectively. Extensive experiments have been performed on a variety of images to validate the effectiveness of the proposed method and its superiority over the previous methods.

Background Few studies have focused on the dimensional accuracy of customized bone grafting by means of guided bone regeneration (GBR) with 3D‐Printed Individual Titanium Mesh (3D‐PITM). Purpose Digital technologies were applied to evaluate the dimensional accuracy of customized bone augmentation with 3D‐PITM with a two‐stage technique. Materials and methods Sixteen patients were included in this study. The CBCT data of post‐GBR (immediate post‐GBR) and post‐implantation (immediate post‐implant placement) were 3D reconstructed and compared with the pre‐surgical planned bone augmentation. The dimensional differences were evaluated by superimposition using the Materialize 3‐matic software. Results The superimposition analysis showed that the maximum deviations of contour between were 3.4 mm, and the average differences of the augmentation contour were 0.5 ± 0.4 and 0.6 ± 0.5 mm respectively. The planned volume of bone regeneration was approximately equal to the amount of regenerated bone present 6 to 9 months after the surgical procedure. On average, the vertical gain in bone height was about 0.5 mm less than planned. And, the horizontal bone gain on the straight buccal of the dental implants and 2 to 4 mm apical of the platform fell also about a 0.5 mm short on average. Statistically significant differences were observed between the augmented volume of virtual and post‐GBR, and the horizontal bone gain of post‐implantation on the level of 4 mm apical to the implant platform (P < .05). Conclusions The dimensional accuracy of customized bone augmentation with the 3D‐PITM approach needs further improvement and compared to other surgical approaches of bone augmentation.

No abstract