What happens when two contexts share the same input unit?
There is a loss of compression ratio, because the contexts may predict different characters, and there is no way for the neural network to distinguish between them. However in text, contexts are not uniformly distributed, so if there is a collision, one context is likely to dominate over the other, minimizing the loss of compression.
Related Questions
- What happens if, under the new formula, the calculated local share is LESS than the district is currently raising through school property taxes?
- What happens if, under the new formula, the calculated local share is MORE than the district is currently raising through school property taxes?
- What happens if one of the input structures cannot be processed?
