These are some of my side projects I did when learning Machine Learning and to explore variant machine learning algorithms and models. It's more for dataset exploring and algorithms experimenting rather than problem solving.
Datasets are all public available.
The algorithms used in this project is rather old, these are the techniques before the debut of CNN, yet the data preprocessing, feature engineering, etc. can still provide some reference.
Techniques used:
- PCA
- Image Pre-Processing
- Image Visualization
- Edge Detection
- Daisy Feature Extraction
- Gabor Filter
Same image recognition task and dataset with the PCA project, yet building MLN(Multi-Layer Neural Network) and variuos CNN models from scratch using keras.
Models built and compared:
- Multi-Layer Neural Network
- Simple CNN with relu activation and Maxpooling
- CNN with dropout
- LeNet
- Basic ResNet
- RasNet50
Exploring PyTorch, built a simple multi-layer CNN for the FashionMNIST dataset.
- Using various Logistic Regression and optimization algorithms to prodict Cellphone Sales Prices.
- Compares the models with scikitlearn's model interms of accuracy and training time.
Used PCA to do feature reduction, train/validation/test split, Multi-Layer Neural Network from scratch.
Techniques Used:
- Stemming
- Stop Words
- Bag-of-word Representation
- Tf-Idf Representation
- Word Frequency
- Word Cloud Generation
- Masked Word Clouds