code

讲基于q学习的应用的相关文献，不错的学习例子(Q learning about the application based on the literature, examples of good learning)

工程计算MATLAB code to calculate the reorthogonalized sine tapers input: N = the length of the time series data to be tapered p = the number of tapers requested I = the gap structure a vector of length N I(t) = 1 if there is data at time t, t=1, ..., N I(t) = 0 if there is a gap at time t output: X = N-by-p vector of the reorthogonalized sine taper(engineering MATLAB code to calculate the reorthog onalized sine tapers input : N = the length of the time series data to be tapere d p = the number of tapers I requested the gap str = ucture a vector of length N of I (t) = 1 if there is dat a at time t, t = 1, ..., I N (t) = 0 if there is a gap at time t output : X = N-by-p vector of the reorthogonalized sine t aper)

mobile user location matlab coding

说明：  计算时变重力场变化的程序，通过此程序可以计算地球质量变化等等。(Changes in time-varying gravity field calculation program, this program can be calculated by changes in Earth' s mass and so on.)

说明：  图像匹配程序，应用matlab将某给定图像在指定的图像集中找出与其最匹配的图像(Image matching procedures, the application of matlab image of a given concentration of the specified image to find its best matching image)

说明：  与连续信号的解调相比，短时突发信号持续时间短，不适用性能良好的反馈环路算法，需要使用前向幵环算法。与一般突发信号相比，短时突发信号没有训练序列、同步头等辅助信息，所有参数需要利用有限的数据估计。与合作通信相比，盲接收信号参数的初始偏差较大，估计范围小的算法不适用。每个突发包的数据有限，解调过程不允许丢失任何符号。(Compared with the demodulation of continuous signal, short burst signal has a shorter duration and is not suitable for the feedback loop algorithm with good performance. Forward loop algorithm is needed.)

完整的mat格式的USPS手写字府库，里面包括7291个训练样本和2007个测试样本(USPS handwritten database of mat format,include 7291 cases for training and 2007 cases for testing.)

本程序利用奇异值分解对3通道彩色图像进行压缩分解，具体步骤如下： 压缩过程： 1. 选取子图像大小K值，把图像分解成M×M个子图像，IMG(s),s=1,2,…, M2，其中M=N/K，原始图像IMG大小为N×N。 2. 计算这M2个子图像的平均值average，对每幅子图像减去均值图像得到新图像。 3. 计算相关矩阵R，其元素定义为 。 4. 计算R的特征值与特征向量，计算每幅子图像与最大特征向量的内积，便得到编码，即压缩后的图像。 (This procedure using singular value decomposition of 3-channel color image compression decomposition, concrete steps are as follows: compression process: 1. Select the sub-image size K value, the image is decomposed into M × M sub-image, IMG (s), s = 1, 2, ..., M2, in which M = N/K, the original image IMG size N × N. 2. M2 calculate the average of sub-image average, for each sub-image minus the new images mean images. 3. Calculation of correlation matrix R, defined as its elements. 4. Calculation of the characteristics of R value and eigenvector calculated each sub-image with the largest eigenvector of the inner product, they will have to code, that is, the compressed images.)

标准压缩感知（CS）理论决定了可靠的信号恢复是可能给M= O（KLOG（N / K））的测量。我们证明了它可以通过利用超越简单的稀疏性和可压缩性由包括价值观和信号系数的位置之间的依赖关系更加逼真信号模型大大降低Mwithout牺牲的鲁棒性。(The standard compressive sensing (CS) theory dictates that robust signal recovery is  possible from  M=O(Klog(N/K))  measurements. We demonstrate that it is possible to substantially  decrease  Mwithout sacrificing robustness by leveraging more  realistic signal models that go beyond simple sparsity and  compressibility by including dependencies between values and  locations of the signal coefficients.   )

说明：  对图像进行频域的低通滤波（包括理想低通、布特沃斯低通、高斯斯通）滤波）、高频强调滤波，并给出滤波后的图像和频谱(The image is filtered by low-pass filter in frequency domain (including ideal low-pass, butworth low-pass, Gaussian stone filter) and high-frequency emphasis filter, and the filtered image and spectrum are given)