Home

Normalization in Deep Learning

Normalization has the advantages to reduce the training time and avoid overfitting as a way of regularization. Challenges during training: Choices regarding data preprocessing often make an enormous difference in the final results The variables (e.g. affine transformation outputs in MLP) in intermediate layers may take values with widely ...

Read more

BLEU Score

BLEU, bilingual evaluation understudy, is a standard algorithm to evaluate the quality of translated text in machine translation. For example, we want to translate a French sentence into English, as we know, there is no single right answer for translation, and different people may have different ways to translate the same sentences. So we will c...

Read more

Decision Tree

Decision Tree A decision tree is a flowchart-like tree structure in which each internal node represents a split based on a feature condition while each leaf node represents a class label. And it is non-parametric supervised learning method which can be used for both classification and regression tasks. The difference between classification and ...

Read more

Notes for Gradient Boosting

I don’t know why I wrote it in Chinese….. Lately I will change it to English….. 最近准备面试,又复习了这一块的知识,想当初老师讲的也极其简单粗暴(如图1。。。),找了些中文和英文的资料,发现中西结合更有效得搞出了这一篇。我也有一模一样的全英文版本,想要的小伙伴可以私信我。Gradient boosting uses an ensemble of decision trees to predict a target label. 简单的说,Gradient boosing 是利用决策树的集成来预测目标值的一种方法。提示一下,在学习gradient boosting之前,需要先了解decision tree(...

Read more

Notes for A5 in Stanford CS224n!

I am learning the model in A5 of Stanford CS224n and I am also a beginner for PyTorch. So I hope I could fully understand the code. All the code below are typed manually with added comments and it may have some typo. Please check Stanford CS224n website for the original code. Even typing is time consuming, it makes me think about the question wh...

Read more

就很喜欢 今生是第一次

这是一部又搞笑又丧又治愈的韩剧,非常贴近当下的年轻人,韩国的编剧真的是太厉害了,大纲抽出来的话也不是什么从未有过的剧情,但是就是会有很多上一秒让你笑的拍桌子,下一秒又难过到你的心坎里。是啊,谁的人生不是第一次,既然是第一次,磕磕碰碰什么的就enjoy啦。 世熙: 不要,我为什么周末要跟马代表一起过? 马代表: 反正你也自己一个人吃着没味道的盒饭,为什么这么对我啊 世熙: 你为什么这样评价我的时间,马代表,像你这样渴望被别人认可,在社会关系中找存在感的人对你来说可能是没有味道的盒饭,对我来说可是像沙漠绿洲的盒饭。在见面之后,要一直听你唠叨这样吃饭对我来说是多么强大的压力。 当我下定决定为了梦想奋斗的时候,我以为我的人生,就要肚子走过这黑暗的隧道,但是没想到会是这么黑暗,也没想...

Read more