A set of simple studies and solutions¶
Predict whether a mammogram mass is benign or malignant¶
- Data: The dataset can be found from University of Irvine: Mammographic Mass.
- Goal: Build a Multi-Layer Perceptron and train it to classify masses as benign or malignant based on its features.
- Challenges: The data needs to be cleaned; many rows contain missing data, and there may be erroneous data identifiable as outliers as well.
-
Approach:
- Review data quality, and missing data. Drop if not a lot of records are wrong
- Transform the data to be usable by sklearn using numpy
See personal notebook in mammogram_mass folder
Computer vision with PyTorch: classify sushi, pizza and steak¶
- Data: The food 101 dataset from PyTorch vision
- Goal: Develop a NN to classify images
- Challenges: The number of layers
- Approach: Develop a basic NN and then compare it with existing CNN
Demonstration with pytorch scripts
- Create a virtual env and install requirements under the pytorch folder;
pip install -r requirements.txt - Under computer vision, load the data sets locally:
python prepare_image_dataset.py --classes sushi,steak,pizza - Do a simple classification using a Tiny VGG:
python classify_food.py - Use transfer learning (see explanations here):
python transfer_learning.py
Other use case¶
- Data:
- Goal:
- Challenges:
- Approach: