My Progress

1. Neurons and the brain Origins: Algorithms that try to mimic the brain Development RoadMap: speech -> images -> text(NLP) -> ... Since the development of Large Neural Network and GPU, we could perform many different task using the "Big Data". 2. Example - Demand Prediction We use sigmoid function to solve the classification problem. We called the function we use f(x). In the neural network, we..

1. Cost function1.1 IntuitionWe use logistic regression / sigmoid function to estimate the data's label or category. How do we choose w and b?For linear regression, we used squared error cost.Linear regression is a convex form, so we could use standard gradient descent equation to figure out local minimum. Since logistic regression is a non-convex form, it has multiple local minimum. Thus, we ca..

1. IntuitionLinear Regression is not a good method for Classification Problem.Why?One outlier on the right changes the linear regression function. It moves decision boundary dramatically. 2. Logistic Regression2.1 FormulaLogistic function allows the function to be curved unlike the linear regression. This is a sigmoid function aka logistic function. x-axis represents the number z. It outputs a v..

1. What is Feature Engineering? Using intuition to design new features, by transforming or combinging original features. Example) House price. If we are given the dimension of the house such as length and width, we can create a new variable to include for price prediction of a house. For example, we create a area variable with length and width. 2. Feature engineering in Polynomial Regression We ..

1. Checking Gradient descent for convergence How to check if Gradient descent is working well? 1.1 Graph This graph is called Learning curve. As the iteration increases, the minimum cost should decrease. If the graph remains constant after enough iterations, we call that it has converged 1.2 Epsilon / Automatic convergence test Let epsilon be 0.001 If the cost function decreaes by epsilon in one..

1. Intuition House example: Lets say we are predicting the price of a house based on its size and number of bed rooms. Number for size of a house is relatively larger than the number of bed rooms. This large difference between numbers will make it hard to accurately predict the price. How is this related to gradient descent? Since the number for size is large, it takes smaller w value to make a ..

1. Linear Regression Purpose: To predict the output based on the given examples or dataset 1. 1 Terminology Univariate linear regression: Linear regression with one(single feature x) variable fw,b(x) = wx + b 1.2 Code Necessary Libraries #Numpy, a popular library for scientific computing import numpy as np #Matplotlib, a popular library for plotting data import matplotlib.pyplot as plt x_train =..

2023 Summer AI/언어모델 시장은 크게 3가지로 나눌수있다.1. 기업들간 생성 AI 경쟁 과열2. Open source와 Closed Source 언어모델들간의 경쟁3. AI 발전에 대한 우려 1.기업들간의 생성 AI 경쟁 과열 모두가 알다시피 GPT가 큰 인기를 얻게 되기 시작하면서 시장의 트렌드가 언어모델 또는 AI에 집중되었다. 수많은 기업들이 LLM(Large Language Model)/ 언어모델을 개발을 하면 경쟁을 하기 시작한다. 기업들간의 경쟁1. ChatGPT by OpenAI2. Bard by Google3. Bing AI by Microsoft4. LLaMA by Meta5. 기타 등등이렇게 수많은 언어모델들이 앞다투어 나오기 시작한다. 이 트렌드에 힘 입어 수많은 기업들이 ..