- 2019년 경북대학교 여름계절학기 지능시스템개론 강의를 듣고 정리하였습니다.
- 기계학습의 기초적인 수학 이론을 공부하고, 이를
python
으로 구현해봅니다. tensorflow
,pytorch
등의 딥러닝 라이브러리를 사용하지 않고,
numpy
,matplotlib.pyplot
모듈만을 사용해 모든 것을 구현하는 것을 목표로 합니다.
- Batch Gradient Descent
- Stochastic Gradient Descent
- Closed-form solution (Ordinary Least Square)
- Compute and compare solutions for
- unregularized linear
- unregularized parabolic
- unregularized 5th-order polynomial
- regularized 5th-order polynomial (RIDGE)
- Implementing FFNN for classification problem
- Back Propagation with Gradient Descent
- Back Propagation
- Resilient Propagation
- Gradient Clipping
notebook links
- K-means A : Clustering some synthetic data
- K-means B : Clustering some real data
- PCA A : Reducing the dimension of som synthetic data
- PCA B : Reduing demension of some real data
- K-means
- PCA
- Spam Mail Detector with Naive Bayes Classifier
- Discriminative model
- learns the conditional probability distribution
p(y|x)
- learns
p(y|x)
directly from the data and then try to classify data - generally give better performance in classification tasks
- learns the conditional probability distribution
- Generative model
- learns the joint probability distribution
p(x, y)
- learns
p(x, y)
which can be transformed intop(y|x)
later to classify the data - we can use
p(x, y)
to generate new data similar to existing data
- learns the joint probability distribution