How charan9665/advml works
This repository combines an offline MovieLens reinforcement-learning study with a personalized movie web application served from its trained Q-table. The research workflow loads and prepares logged ratings, replays user histories to train policies, then compares and evaluates them. The web workflow authenticates users, serves recommendations and movie features, and persists account-specific state; the React client calls the API and renders those features. The backend’s catalog and poster resolver are useful internal layers. Several backend routes and frontend page details are unsampled, so those links rely only on explicit imports, API calls, or documented behavior.
Main entry points: main.py Study runner · app.py FastAPI service · experiments.py Parameter experiments · train.py Policy training
Components 23
Offline RL study
- Study runner
main.py - Experiment settings
config.py - MovieLens loader
src/data_loader.py - Data exploration
src/eda.py - Chronological splits
src/preprocessing.py - Replay environment
src/environment.py - Q-learning agents
src/agents.py - Policy training
src/train.py - Baseline policies
src/baselines.py - Policy evaluation
src/evaluate.py - Parameter experiments
src/experiments.py - Results plots
src/plots.py
Recommendation backend
- FastAPI service
webapp/backend/app.py - Movie catalog
webapp/backend/app.py - SQLite user and cache data
webapp/backend/app.py
Web client
- React application
webapp/frontend/src/App.jsx - Application pages
webapp/frontend/src/pages.jsx - API client
webapp/frontend/src/api.js - Movie cards
webapp/frontend/src/components/MovieCard.jsx
Movie artwork
- Poster display
webapp/frontend/src/components/Poster.jsx - TMDB poster resolver
webapp/backend/tmdb_posters.py
Other components
- Movie viewer
- TMDB
Connections 28
- Study runner → MovieLens loader: loads data
main.py - Study runner → Data exploration: runs exploration
main.py - Study runner → Chronological splits: prepares splits
main.py - Study runner → Policy training: trains policy
main.py - Study runner → Parameter experiments: tunes parameters
main.py - Study runner → Policy evaluation: evaluates policies
main.py - Study runner → Results plots: saves plots
main.py - Study runner → Baseline policies: creates baselines
main.py - Policy training → Replay environment: replays histories
src/train.py - Policy training → Q-learning agents: updates policy
src/train.py - Policy evaluation → Replay environment: replays test histories
src/evaluate.py - Parameter experiments → Policy training: runs training
src/experiments.py - Parameter experiments → Policy evaluation: compares policies
src/experiments.py - Parameter experiments → Results plots: plots comparisons
src/experiments.py - Movie viewer → React application: uses
webapp/frontend/src/App.jsx - React application → Application pages: renders pages
webapp/frontend/src/App.jsx - React application → API client: requests data
webapp/frontend/src/App.jsx - Application pages → API client: requests page data
webapp/frontend/src/pages.jsx - API client → FastAPI service: sends HTTP requests
webapp/frontend/src/api.js - FastAPI service → Movie catalog: loads MovieLens and Q table
webapp/backend/app.py - FastAPI service → Q-learning agents: uses Q policy
webapp/backend/app.py - FastAPI service → TMDB poster resolver: creates resolver
webapp/backend/app.py - FastAPI service → SQLite user and cache data: reads and writes
webapp/backend/app.py - TMDB poster resolver → SQLite user and cache data: caches artwork
webapp/backend/tmdb_posters.py - TMDB poster resolver → TMDB: searches titles
webapp/backend/tmdb_posters.py - Movie cards → Poster display: renders artwork
webapp/frontend/src/components/MovieCard.jsx - Poster display → API client: resolves artwork
webapp/frontend/src/components/Poster.jsx - FastAPI service → Movie viewer: returns recommendations and features