COE491 SMART MOBILITY DESIGN FOR DESIGNER, ENGINEER, AND DATA SCIENTIST

FALL 2023

CLASS MATERIALS

Lecture Slides: KLMS

REFERENCE VIDEOS

Youtube (In Korean): Link

CODE

Prepare
1) This class uses Google Colab for practice. If you prefer, you can install Jupyter Notebook and TensorFlow on your personal computer. See below for how to prepare.
TensorFlow + Colab + Python
2) This class uses bracket data, which is structural design data. Please download below.
data_2000, data_1000, data_500

1. Convolutional Neural Network (CNN)

1) Colab
  --Bracket(Pytorch): CNN_Sequential
  --Bracket(TensorFlow): CNN_Sequential
  --MNIST(TensorFlow): CNN_Sequential, CNN_Functional

2) Jupyter Notebook
  --Bracket(TensorFlow): CNN_Sequential
  --MNIST(TensorFlow): CNN_Sequential, CNN_Functional

# Compare learning methods (MNIST)
Sequential+ModelFit / Sequential+GradientTape /
Functional+ModelFit / Functional+GradientTape /
Subclassing+ModelFit / Subclassing+GradientTape
Ensemble

2. Autoencoder (AE) & Anomaly Detection

1) Colab
  --Bracket: Autoencoder, Anomaly Detection

2) Jupyter Notebook
  --Bracket: Autoencoder, Anomaly Detection

# Compare learning methods (Bracket)
ModelFit / GradientTape

3. Generative Adversarial Network (GAN)

1) Colab
  --Bracket: GAN

2) Jupyter Notebook
  --Bracket: GAN

# Compare learning methods (Bracket)
ModelFit / GradientTape

Option 1: Variational AutoEncoder (VAE)

1) Colab
  --Bracket: VAE

2) Jupyter Notebook
  --Bracket: VAE

# Compare learning methods (Bracket)
ModelFit / GradientTape

Option 2: Advanced Models

1) Convolutional Autoencoder (CAE)
  --Bracket: ModelFit / GradientTape

2) Deep Convolutional GAN (DCGAN)
  --MNIST: Keras

3) Conditional GAN (cGAN)
  --MNIST: Keras

4) Boundary Equilibrium GAN (BEGAN)
  --MNIST: Keras

Competition

CNN Bracket(PyTorch): Test
CNN Bracket(TensorFlow): Test