yipala的博客

224444ssff

01 计算机视觉 02 计算机视觉应用03 机器学习训练图像的流程04 深度学习图像识别流程05 卷积神经网络06 物体检测07 传统目标检测方法基于深度神经网络的目标检测...

参考：http://ruder.io/optimizing-gradient-descent/index.html#adadeltahttps://blog.youkuaiyun.com/u010089444/article/details/76725843

Latent Space Autoregression for Novelty Detection AbstractNovelty detection is commonly referred to as the discrimination of observations that do not conform to a learned model of reg...

用于无监督异常检测的深度自编码器高斯混合模型DEEP AUTOENCODING GAUSSIAN MIXTURE MODEL FOR UNSUPERVISED ANOMALY DETECTION摘要1 INTRODUCTION2 RELATED WORK3 DEEP AUTOENCODING GAUSSIAN MIXTURE MODEL3.1 OVERVIEW3.2 ...

请认仔细阅读下文，就能理解onehot转化 what is form of one-hot ?即用包含0和1的tensor来表示数字标签，数字1所在的索引值（从0开始）即为我们的数字标签，例如我们有0-9的数字标签，则标签5所对应的one-hot形式为[0 , 0 , 0 , 0 , 0 ,1 , 0 , 0 , 0 , 0]，因为1所在位置的索引值为5。 And how to transf...