Sharing some of my favorite projects that I have completed throughout Thinkful's curriculum.
Out of my early projects I really enjoyed how these two came out. This area of the curriculum is where I got most of my practice utilizing SK-Learn, Pandas, Matplotlib and Seaborn. With these tools I practiced general data cleaning, data visualization and predictive modeling.
The unit 4 capstone centered around natural language processing. Covering tools such as NLTK and spaCy. Using these tools I was then able to train a classifier that could identify the source material of a paragraph from one of ten Shakespeare plays.
During some of my downtime I took on a few extra projects. The first being Kaggle's MNIST challenge where I learned how to implement Convolutional Neural Networks. The next project used a similar approach when detecting breast cancer in X-rays.
Before finishing the program I applied to some jobs. As a pre-screening question I was asked to implement K Means Clustering from scratch. Even though I knew how K Means clustering was done mathematically it proved to be a difficult coding challenge.
The final capstone focused on image segmentation using a U-Net architecture. I trained two variants that showed promising results. The first used a VGG11 encoder to utilize transfer learning. The second was trained completely from scratch and used convolutional transpose layers to handle the upscaling portion of the U-Net architecture. Overall both models performed well, but the transfer learning model was a clear winner.