Here are some suggestions for projects to do. Either pick something from this list or use it as inspiration.
If you choose a project use strikthrought to remove it -> crossed
Explain an optimization method not covered in the course though an interactive notebook. Gold standard is Distill
- Conjugate gradient descent
- Cross-entropy method
- Levenberg–Marquardt algorithm (for nonlinear regression and inverse kinematics)
- training adverserial neural networks
- Q-learning
- link between fractals and the Newton method
- MAP-elites
- information-geometric opitmization algorithms
- the big deal with multi-objective optimization
- Thompson sampling: optimal gambling 🎲
- Christofides' algorithm for solving the TSP
- ...
Solve a problem using a method from the course or a new one.
- time table scheduling
- my own RNA folder!
- color transfer using optimal transportation done right
- analytic solution of a ODE using genetic programming
- graph matching using simulated annealing
- optimal task for every person using the Hungarian algorithm
- finding negative cycles using the Bellman–Ford algorithm
- multiple sequence alignment using Genetic Algorithms
- using MaxEnt to find the best probability distribution
- finding design patterns in phage lytic proteins using evolutionary algorithms
- homebrew SVM
- recovering images from a few pixels using compressed sensing (L1-norm minimization)
- visalizing complex networks using quadratic optimization
- ...
Experiment and learn. Reproduce the results of a paper (in simplified form). Show me something exciting about optimization! Use visuals to tell a story.
- fitting a neural network by optimizing one weigth at a time
- search paths of Hill Climbing: an exhaustive search
- convex restrictions in physical design
- non-covergence as a regularization mechanism in machine learning
- improving gradient descent or learning using gradient startvation
- improving the convergence of evolutionary algorithms by varying the objective function
- the optimization landscape of ANNs
- comparision of optimizers for deep learning
- Newton methods for lazy bums: is the diagonal sufficient?
- utilitarian ethics: optimization for the Good Place?
- Maze solver using bacteria (simulation of an experiment)
- discovering natural laws
- ...