Year 12 · Student investigation
Does a better reconstruction reveal every concept?
Keep an individual record even when you investigate in pairs. Record the exact settings, source and data split.
Will increasing dictionary size reduce both average error and error on examples containing the rare feature?
What did you change? What stayed the same? What did you notice?
Use a very rare feature. If the held-out set has no rare cases, report “not measured”, not zero error. Compare seeds 11, 23 and 47 with the same settings. Report every result and the range. Lock your final settings before opening seed 89; changing settings afterwards turns that check into development data.
A three-setting comparison of held-out reconstruction, rare-feature count and rare-feature error.
This activity can show _ . It cannot tell us _ . A new case I would test is ____ .
Use fictional data. Download or print your work before leaving.