While practicing DSA on LeetCode, I’ve been thinking about how different learning formats push understanding in different ways.
One format I recently explored was a Reinforcement Learning–based problem setup, where instead of writing a direct solution, you train an agent and refine behavior through reward shaping and iterations. It felt similar to how we debug greedy or DP solutions — except the feedback loop is probabilistic and visual.
It made me appreciate ideas like:
- defining the right “state”
- shaping rewards similar to constraints
- tuning parameters the way we tune edge cases in code
Sharing this in case anyone here is interested in exploring RL alongside traditional DSA practice
Curious if others here mix algorithmic problem-solving with ML/RL experiments, or prefer keeping them separate.