In the previous post, I introduced the small neural unit, Logistic regression being considered neural network without hidden layers.

The main difference of neural network from logistic regression is stacking layers called hidden layers.

When stacking each hidden layer, we also stack activation function.

This post is about the activation function and why we need it.

Machine learning Neural network Deep learning Python Tensorflow

The Perceptron is a simplest version of Neural network. It is similar to way Logistic regression works.

I already posted about the Logistic regression on my post for Machine learning.

To remind, I will explain more detail of Logistic regression.

Machine learning Neural network Deep learning Python Tensorflow

Deep learning start from the idea to make thinking machine.

Then how to make the thinking machine?

People use the brain to think something. If computer can imitate the human’s brain, Can computer think like human?

Machine learning Neural network Deep learning Python Tensorflow

2018 Pycon Korea에서 뱅크샐러드가 주최한 “텍사스 홀덤” 알고리즘 대회가 있었습니다.

마감전까지 시간내에 동작하는 로직 구현을 하지못해서 참여는 못하였지만, 고민하는 과정이 의미있었습니다.

해당 알고리즘에서 계산 한것은 아래와 같습니다.

  • 이길 확률
  • 내 패의 강함
  • 상대 패의 강함
  • 배팅 비율

P.S. 구현은 해두고 시간 제한이 걸려서 이벤트를 참석하지 않았습니다. 서버에서 1초 내에 구현 되는 2번째 로직으로 제출을 했는데, 노트북에서는 시간이 더걸리더군요… 이후 3번째 로직은 노트북에서 동작하도록 마감 이후에 변경하였습니다.

Development Mathematics Python

One of the most popular library for Deep learning and Machine learning is Tensorflow by google.

This libray support GPU computing and powerful visualization for your training process.

In this post, I will introduce the below topics about tensorflow.

  • Tensorflow concept
  • Computation
  • Optmize model
  • Visualize model

The sample codes and definitions are from Tensorflow cookbook1 and Wikipedia.

  1. https://github.com/nfmcclure/tensorflow_cookbook ↩

Machine learning Tensorflow Linear algebra

One of the most interesting part in Machine learning is knowing what you want based on the your logs.

Like netflix, User rates movies using zero to five stars, Then the system will recommend movies you may like by analyzing your personal taste.

Machine learning Tensorflow Anomaly detection

How do we detect whether value is anomalous? Like alerting system based on log data, the system must know the some input is different from normal cases.

  • Fraud detection
  • Manufacturing
  • Monitoring log

In this post, I will introduce how to make an Anomaly detection algorithm using Gaussian.

Machine learning Tensorflow Anomaly detection

When your data samples are in high-dimensional feature space, you should consider the curse of dimensionality1 decreasing accuracy of your model.

High dimensional is hard to understand and not always helpful for your data.

In this post, I will introduce about the way to reduce to map high-dimensional features into low-dimensional features by PCA.

  1. https://en.wikipedia.org/wiki/Curse_of_dimensionality ↩

Machine learning Tensorflow Dimensional reduction