Overview
What this challenge is about.
Pick a public trajectory dataset (e.g., Argoverse 2, Waymo Open, or ETH-UCY). Implement three models with comparable parameter counts (around 5M each): an LSTM baseline, a vanilla causal transformer, and a state-space-model (SSM) baseline like S4 or Mamba. Train on the same compute budget (2 GPU-days each). Evaluate average displacement error (ADE), final displacement error (FDE), and inference latency. Run 5 seeds per condition. Write a 5-page workshop-style report with one striking plot per architecture comparison.
The Brief
What you'll do, and what you'll demonstrate.
Compare LSTM, transformer, and SSM for trajectory prediction with controlled compute and write the workshop-style report.
Earning criteria — what you'll demonstrate
- Implement RNN, transformer, and SSM at comparable parameter counts
- Run a fair architecture comparison under shared compute
- Evaluate sequence models with standard trajectory-prediction metrics
- Write a workshop-style architecture comparison report
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Research Scientist
Comparing architecture families with proper statistical rigor is the daily work of research scientists at AV research centers and AI labs.
This challenge sharpens
- transformers
- state-space-models
- experiment-design
ML Researcher
Designing fair architecture comparisons and writing the workshop report is the applied ML-research work that bridges research and product.
This challenge sharpens
- transformers
- rnn
- trajectory-prediction
Applied AI Scientist
Architecture-choice analyses inform every applied-AI project; this challenge gives the student the methodology to lead one.
This challenge sharpens
- pytorch
- trajectory-prediction
- experiment-design