** ACTA 기계학습과 딥러닝 (20시간 과정)**
(2주, 매주 10시간: 주중 야간 줌수업  3 시간씩   2 회 ,   토   4 시간 대면수업 )
 
대상:  ACTA 기계학습과 딥러닝은 중급 파이썬 수준의 엔지니어들이 기계학습과 딮러닝의 역량을 단기간에 강화하고자 하는 경우에 적합한 과정

Machine Learning (10시간)
  1. Basics – Curse of dimensionality, bias-variance tradeoff, K-nearest neighbors
  2. Linear Regression – Parameter estimation by least squares,
  3. Classification – Logistic regression & Maximum likelihood
  4. Resampling methods – Training, validation, and test datasets
  5. Model selection and regularization – Ridge and Lasso, forward & backward selections
  6. Non-linear methods – Tree-based, support vector machine, K-means clustering

Deep Learning (10시간)
  1. MLP – gradient descent, hidden layers, architecture design
  2. Optimization & Regularization – initialization, stochastic gradient descent, ADAM, early stopping
  3. CNN – parameter sharing, equivariance to translation, computational benefit, pooling
  4. RNN – computational graphs, teacher forcing, bidirectional RNN, vanishing & exploding gradients, LSTM
  5. Transformer – query/key/value, attention, masked learning & predictive learning
  6. Advanced topics (high level concept explanations) – generalization, representation, unsupervised learning, LLM, etc.