Custom implementations of machine learning algorithms.
| Algorithm / Architecture | Language | Framework |
|---|---|---|
| CART | Julia | Base |
| Transformer | Python | PyTorch |
| GRU | Python | PyTorch |
| LSTM | Python | PyTorch |
| DenseNet | Python | PyTorch |
| Mixup data augmentation | Python | NumPy |
| ResNet | Python | PyTorch |
| Conv2d | Julia | Base |
| One-Step Actor Critic REINFORCE | Python | PyTorch |
| Vanilla Policy Gradient w/ GAE | Python | PyTorch |
| Neural Network | Python | NumPy |
| Support Vector Machine | Python | NumPy |
| K-Nearest Neighbor | Python | Base |
| Radial Basis Function Network | Python | Base |
| Perceptron | Python | NumPy |
Plots of decision regions for classifying digits 1 versus not 1 are shown below. Average intensity and symmetry were two features that were computed from the raw image data. In each case, regularization and cross validation was used.
- Neural Network
- Support Vector Machine
- K-Nearest Neighbor
- Radial Basis Function Network
- Perceptron
