Temporal decay

TemporalHypergrid applies exponential decay before each update() call:

\[H_t \leftarrow \lambda \cdot H_{t-1} + \Delta_t\]

where \(\lambda \in (0, 1]\) is the decay factor and \(\Delta_t\) is the new data batch.

This is equivalent to an exponential moving average over batches, giving more weight to recent data. The effective half-life in batches is \(\log(0.5) / \log(\lambda)\) — e.g. with \(\lambda = 0.99\) and batches of 1000 points, the half-life is ~69 batches (69 000 points).

Snapshots are saved every snapshot_interval points and compared with evolution(method) to track drift over time.