LatentTrack Experiments

This repository contains the training and evaluation pipeline for LatentTrack and baseline models used in sequential forecasting and uncertainty experiments.

The workflow is intentionally split into three stages:

1. Training & evaluation (main.py)
2. Multi-seed aggregation & representative seed selection (gen_multi_seed_figs.py)
3. Single-seed visualization (gen_single_seed_figs.py)

This separation keeps training reproducible, comparisons fair across seeds, and figures consistent with the paper.

---

Repository Workflow Overview