Physics 7097 "Machine Learning"
Fall 2020
Project Assignments
Below is the list of students (in alphabetical order by their initials) and the selected topics.
- 01. JA: "OPTICS clustering"
- 02. SB: "Voting Classifier and Voting Regressor"
- 03. DC: "Novelty and Outlier Detection"
- 04. AD: "Predicting cryogenic thermalization with neural networks"
- 05. PE: "Expanding LISA Optical Pathlength Noise Simulations" (regression and dimensionality reduction)
- 06. SG: "Affinity propagation clustering"
- 07. NG: "Using topological data science to detect cosmic voids"
- 08. SKa: "A basic time series forecasting of the stock market using LSTM"
- 09. SKu: "Predicting match results in the English Premier League"
- 10. LL: "The Winning Formula"
- 11. IM: "Recovering Binary Black Hole Mergers with Convolutional Neural Networks"
- 12. TM: "Enhancing detection of Gravitational waves with Machine Learning" (uses XGBoost classifier)
- 13. VR: "Graph Neural Networks"
- 14. NR: "Binary classification of Higgs from 4 lepton background processes using Neural network""
- 15. AR: "Transformer modelling in music or something easier"
- 16. MS: "Manifold learning: t-SNE"
- 17. BS: "Dispersed Multiphase Flow Generation using 3D Steerable CNN"
- 18. SS: "Hierarchical clustering"
- 19. CS: "Decomposing signals in components"
- 20. LT: "Glitch classification using a convolutional autoencoder"
Below are some topics which we did not get a chance to mention or discuss in detail in class. They are all implemented in scikit-learn and make for suitabe final projects. You will need to briefly explain the basic idea, show some simple example as an illustration of a typical implementation, and present a performance comparison against a competitor method.
- Novelty and outlier detection.
- Ensemble methods (two topics):
- Decomposing signals in components (any one method)
- Manifold learning (any one of the 8 methods)
- Clustering
- Autoencoders
- Restricted Boltzmann machines
Return to Physics 7097 home page