贝叶斯分类器

本文是2016 台湾资料科学年会前导课程“一天搞懂深度学习”的全部讲义PPT（共268页），由台湾大学电机工程学助理教授李宏毅主讲。作者在文中分四个部分对神经网络的原理、目前存在形态以及未来的发展进行了介绍。深度学习的每一个核心概念在文中都有相关案例进行呈现，通俗易懂。一天的时间搞懂深度学习？其实并不是没有可能。(This is the entire handout PPT (268 pages in total) of "a day to understand deep learning", the leading course of the 2016 Taiwan annual conference of data science. It was given by li hongyi, assistant professor of electrical engineering at Taiwan university.In this paper, the author introduces the principle, existing form and future development of neural network in four parts.Each core concept of deep learning is presented in relevant cases, which are easy to understand.)

SRCNN Superresolution imteplated by tensorflow SRCNN tensorflow 实现(SRCNN Superresolution imteplated by tensorflow)

用于运行LSTM的预测代码，例子是国际航班客流量。使用语言为python.(The forecast code used to run LSTM is an example of international flight traffic. The language used is python.)

Django测试，Django测试Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，Django测试，

说明：  pca-svm的人脸识别，利用pca进行人脸特征提取，再利用svm进行聚类分析(Face recognition based on PCA SVM)

python 卡曼滤波的python代码和说明文档，程序都在文档里(python kalmanfilter)

对于视频中动态背景下运动前景部分提取，一种基于单高斯模型和混合高斯模型的动态背景模型提取前景信息。(For moving foreground extraction in dynamic background, a dynamic background model based on single Gaussian model and mixed Gaussian model is proposed to extract foreground information.)

selenium+PYHTON环境的自动化，模仿百度联想词编写的代码，方便大家使用

说明：  在许多情况下，利用深度学习算法搭建的神经网络模型都需妥进行某 种形式的优化。 这非常重要，只有经过优化的网络，才能在训练之后达到 不错的解决问题的效果。 优化的最直接目的就是使参数更加准确地更新。 一般神经网络的训练过程大致可以分为两个阶段：第一个阶段先通过 前向传播算法计算得到预测值，并将预测值和真实值做对比，得出两者之 间的差距；在第二个阶段，通过反向传播算法计算损失函数对每一个参数 的梯度，再根据梯度和学习率使用梯度下降算法更新每一个参数。(In many cases, the neural network model built by deep learning algorithm needs to be optimized in some form. This is very important, only after the optimization of the network, in order to achieve good results in solving problems after training. The most direct purpose of optimization is to update parameters more accurately. The training process of general neural network can be roughly divided into two stages: in the first stage, the predicted value is calculated by the forward propagation algorithm, and the difference between the predicted value and the real value is obtained; in the second stage, the loss function is calculated by the back-propagation algorithm for each parameter According to the gradient and learning rate, the gradient descent algorithm is used to update each parameter.)

说明：  675991+Keras图像深度学习实战.侯宜军%28带书签文字版%29(Keras Image Deep Learning Practical Combat. Hou Yijun%28 with Bookmark Text Version)