Deep learning has achieved outstanding results in various tasks in machine learning under the background of rapid increase in equipment’s computing capacity. However, while achieving higher performance and effects, model size is larger, training and inference time longer, the memory and storage occupancy increasing, the computing efficiency shrinking, and the energy consumption augmenting. Consequently, it’s difficult to let these models run on edge devices such as micro and mobile devices. Model compression technology is gradually emerging and researched, for instance, model quantization. Quantization aware training can take more accuracy loss resulting from data mapping in model training into account, which clamps and approximates the data when updating parameters, and introduces quantization errors into the model loss function. In quantization, we found that some stages of the two super-resolution model networks, SRGAN and ESRGAN, showed sensitivity to quantization, which greatly reduced the performance. Therefore, we use higher-bits integer quantization for the sensitive stage, and train the model together in quantization aware training. Although model size was sacrificed a little, the accuracy approaching the original model was achieved. The ESRGAN model was still reduced by nearly 67.14% and SRGAN model was reduced by nearly 68.48%, and the inference time was reduced by nearly 30.48% and 39.85% respectively. What’s more, the PI values of SRGAN and ESRGAN are 2.1049 and 2.2075 respectively.

Graph neural networks (GNNs) have emerged as powerful approaches to learn knowledge about graphs and vertices. The rapid employment of GNNs poses requirements for processing efficiency. Due to incompatibility of general platforms, dedicated hardware devices and platforms are developed to efficiently accelerate training and inference of GNNs. We conduct a survey on hardware acceleration for GNNs. We first include and introduce recent advances of the domain, and then provide a methodology of categorization to classify existing works into three categories. Next, we discuss optimization techniques adopted at different levels. And finally we propose suggestions on future directions to facilitate further works.

This article finds that the reusability of vertices in the same graph in graph processing differs, and the high-reuse and low-reuse vertices are stored together. These phenomena lead to the inability of existing GPU architectures to capture the reusability of graph processing. The most advanced cache optimization strategies cannot implement different management strategies for data with different reusability, which is an essential reason for graph processing’s poor performance. Therefore, we propose a TSN cache scheme for the GPU platform. This scheme employs distinct management strategies for data with varying reusability in the cache, effectively leveraging the locality of these different data types. In addition, the TSN cache scheme can also reduce the probability of cache thrashing caused by low-reuse data. This article evaluates multiple graph algorithms and datasets and shows that the TSN cache scheme achieves an average speedup of 1.38 compared with the baseline scheme.

As a popular research field of computer vision, super resolution(SR) has received more and more attention in recent years. Although the deep learning methods have achieved good results in SR, there are still some problems. For example, the previous models are often based on single depth mechanism. This means that the SR reconstruction problem of all images is regarded as of equal complexity. And we found that some images have more details and are suitable for recovering in complex models, while other images have less texture information and are suitable for recovering in simple models. At the same time, the size of the training set is too large, which creats a lot of resource overhead. To solve these problems, this paper proposes a new SR framework can be customized according to image features. We choose 3 representative models for testing and in test our framework can reduce the size of the training set by 41.9%. For MSRResNet, we can reduce the training time from 2517 minutes to 2449 minutes. The reconstruction quality of 61 % test images has been improved and the average perceptual index has dropped from 5.1912 to 5.155833, at the same time the reconstruction time has been optimized from 85 seconds to 59 seconds. For SRGAN, the training time can be reduced from 1920 minutes to 1812 minutes. The reconstruction quality of 58 % test images have been improved and the average perceptual index has dropped from 2.0869 to 2.0509, while the reconstruction time has been optimized from 82 seconds to 46 seconds. For ESRGAN, the training time can be reduced from 4368 minutes to 4249 minutes. The reconstruction quality of 78 % test images has been improved and the average perceptual index has dropped from 2.2041 to 2.0535. The reconstruction time has been optimized from 138 seconds to 95 seconds. Our framework can improve the effect of super-resolution models while reducing the resource overhead.