In the previous post, we can predict the value using simple linear regression model. Let’s think about different situations.

• SPAM(Y/N)
• Fraudulent(Y/N)
• Tumor(Y/N)

In the case the output is discrete like example, Linear regression cannot predict correct answer. Then how can we deal with the discrete case?

How can the computer think by self? Normally, they can’t think like human, they will use their way to learn and think.

This post is based on the Stanford video lecture,”Machine learning”, by Andrew Ng.

개발자의 필수도구이자 협업도구로 널리 사용하고 있는 git에 대하여 소개하고, 효과적인 사용방법과 흐름에 대하여 설명하고자 한다.

Realworld is not deterministic environment. We can’t sure our decision has same result in same environment. It is called stochastic environment.

Non-deterministic

[\hat Q(s,a) \leftarrow r + \alpha[\gamma \max_{a’} \hat Q (s’,a’)]
\alpha:\text{learning ratio}
\gamma:\text{discount ratio}
\hat Q : \text{approximation value for }Q\]

[Q(s,a) \leftarrow (1- \alpha)Q(s,a) + \alpha[r+\gamma \max_{a’} \hat Q (s’,a’)]]

자연어 처리에서 사용되는 다양한 모델들 중 확률 기반하여 토픽이나 키워드 추출에 사용되는 여러 모델들이 있다.

자연어 처리에 기본적으로 사용되는 확률 기반 모델들(TF-IDF, BM25, N-gram, LDA, LSA, pLSA)를 소개한다.

3D localization (3D $\rightarrow$ 2D)

To recognize image on the computer, the feature is one of the important part for it. And there are some algorithms to detect features like edge and corner.

And depending on feature type, we can consider which algorithm is useful for our image.

The reason why we analyze images is to get the information and features.

So, We should know how to get the information from images. Many algorithms exist for filtering, edges and corner. It is used for many way to extract and exaggerate features.

Image processing also important to remove noise in images and transform images before extracting features.

In this post, i will introduce some algorithm for image processing.