Following are some resources to help you explore related projects.
Projects
- 2018 Data Science Bowl (link)
- Freesound General-Purpose Audio Tagging Challenge (link)
- Toxic Comment Classification Challenge (link)
- Cdiscount’s Image Classification Challenge (link)
- Mercedes-Benz Greener Manufacturing (link)
- Web Traffic Time Series Forecasting (link)
- Personalized Medicine: Redefining Cancer Treatment (link)
Helpful Resources
- cs229 2014 Projects
- cs229 2016 Projects
- Kaggle for datasets and competitions link
- UC Irvine Machine Learning Repository for datasets [link]
- A list of Machine Learning Project Ideas link
Getting Started with your Project
I have seen some groups stuck in the initial phase of their project where they have no idea what to do. If so, watch following video from Andrew Ng course for getting started advice. Build your first system quickly then iterate
And if you got some extra time, skim through following, very short course that will definitely help you devise a proper strategy for your project.
Structuring Machine Learning Projects
Hyper parameter Tuning
Since all of you will soon start working on your projects, we will be sharing some useful materials from time to time. One key problem in machine learning is hyper-parameter optimization. A rudimentary method is to use Grid Search, which is extremely exhaustive. Another way is to do random search for optimal hyper-parameters as explained in this paper.
Random Search for Hyper-Parameter Optimization
But searching in hyper-parameter space blindly is not a good idea. You may want to use the knowledge of previous hyper-parameter performance evaluations to guide your search for the optimal ones ergo Bayesian optimization. You can easily find many research papers and materials on this topic.
Bayesian Optimization with scikit-learn
Tutorial on Bayesian Optimization for Machine Learning
Here is another very interesting paper from nips.
An Empirical Bayes Approach to Optimizing Machine Learning Algorithms
This is an active area of research and you can find very good study material as well.