This book provides a structured treatment of the key principles and techniques for enabling efficient processing of deep neural networks (DNNs). DNNs are currently widely used for many artificial intelligence (AI) applications, including computer vision, speech recognition, and robotics. While DNNs deliver state-of-the-art accuracy on many AI tasks, it comes at the cost of high computational complexity. Therefore, techniques that enable efficient processing of deep neural networks to improve key
2025-04-15 10:21:45 20.36MB 机器学习 硬件优化
1
This dissertation proposes three circuit design techniques for successive-approximation register (SAR) analog-to-digital converters (ADCs). According to the measurement results of the proof-of-concept prototypes, the proposed techniques are able to improve the operating speed and achieve excellent energy efficiency. The proposed techniques and chip measurement results are sketched as follows: The first technique is a monotonic capacitor switching procedure. Compared to converters that use the conventional procedure, the average switching energy and total sampling capacitance are reduced by about 81.3% and 50%, respectively. A 10-bit, 50-MS/s SAR ADC with the proposed monotonic capacitor switching procedure is implemented in a 0.13-μm 1P8M CMOS technology. The prototype ADC consumes 0.92 mW from a 1.2-V supply, and the effective number of bit (ENOB) is 8.48 bits. The resulting figure of merit (FOM) is 52 fJ/conversion-step. However, the signal-dependent offset caused by the variation of the input common-mode voltage degrades the linearity of ADC. We proposed an improved comparator design to avoid the linearity degradation. Besides, to avoid a clock signal with frequency higher than sampling rate, we used an asynchronous control circuit to internally generate the necessary control signals. The revised prototype is also implemented in a 0.13-μm 1P8M CMOS technology. It consumes 0.826 mW from a 1.2-V supply and achieves an ENOB of 9.18 bits. The resultant FOM is 29 fJ/conversion-step.
2025-04-04 20:42:28 3.09MB ADC
1
Efficient filter design against interrupted sampling repeater jamming for wideband radar
2024-03-02 02:02:46 2.27MB 研究论文
1
通过卡尔曼滤波进行有效GP回归 基于两篇论文的存储库,其中包含相对于同类项目的简单实现代码: [1] A.Carron,M.Todescato,R.Carli,L.Schenato,G.Pillonetto,机器学习遇到了Kalman Filtering ,《 2016年第55届决策与控制会议论文集》,第4594-4599页。 [2] M.Todescato,A.Carron,R.Carli,G.Pillonetto,L.Schenato,通过卡尔曼滤波的有效时空高斯回归,ArXiv:1705.01485,已提交JMLR。 PS。 该代码尽管基于上述论文中使用的代码,但与之稍有不同。 它是它的后来的改进和简化版本。 而且,此处仍未提供[2]中介绍的用于实现自适应方法的代码。 文件内容是很容易解释的(有关每个文件的简要介绍,请参考相应的帮助): main.m:包含主程序 plotResul
1
l-曲线矩阵代码A-LSQR类型的方法提供了计算效率高的正则化p的自动最佳选择 Matlab代码* :(需要) #Matlab基于LSQR的算法的实现(建议):reconstruct_cw_lsqr.m(要求目标函数:opt_lambda_cw.m和Lanczos双双角化函数:lsqr_b_hybrid.m **) #Matlab基于L曲线的算法(传统方法)的实现:reconstruct_cw_l_curve.m(需要使用正则化工具**) #Matlab基于GCV的算法(传统方法)的实现:reconstruct_cw_OGCV.m(需要使用正则化工具**) #Matlab基于MRM的算法(传统方法)的实现:reconstruct_stnd_cw_OMRM.m 此Matlab代码用作以下工作的一部分: Jaya Prakash和Phaneendra K. Yalavarthy,“ LSQR型方法在漫射光学层析成像中提供了计算效率高的自动优化正则化参数的选择,” Medical Physics,40(3),033101(2013)。 创建于:2012年9月9日 更新日期:2012年9月11
2023-03-01 19:54:40 30KB 系统开源
1
文的其余部分结构如下:在第3节中,我们描述了我们的高效立体匹配方法。第4节报告了真实世界数据集的实验结果以及与Middlebury基准图像上的各种其他方法的比较
2023-01-16 17:51:19 5.51MB
1
高效的面边缘检测和定量的绩效考核 首先介绍递归过程为有效地计算三次面参数的边缘检测,这个过程可以通过计算在曲面参数方程,用固定数量的相互独立算子。 然后,我们引入一个独立的图像定量标准解析评测不同的边缘检测器(包括梯度和过零基础的方法)。
2023-01-06 08:55:37 225KB 高效边缘检测
1
Efficient gamut clipping for color image processing using LHS and YIQ 很经典的一篇图像增强处理的方法介绍,在LHS和YIQ空间都可以,调整饱和度,亮度等,还有超出RGB范围时,快速做gamut clipping,在显示和拍照里都用的到
2023-01-03 13:28:26 545KB gamutclipping LHS YUV
1
经典的leach-mac协议文献。作者:Wendi Rabiner Heinzelman, Anantha Chandrakasan, and Hari Balakrishnan。写于2002年
2022-12-11 21:20:06 309KB leach mac协议
1
Efficient MySQL Performance Best Practices and Techniques 2021 PDF格式,好书
2022-12-06 18:15:59 4.68MB MySQL
1