Technique "Vector Quantization"
Component0731215963Jump to navigation Jump to search
See Publication "How to create a mind", page 135.
- Reduce data complexity
- Reduce multiple data dimensions to one (a single number as the quantized representation of a cluster)
- Improve ability to find invariants
- Reduce data to equally likely probabilities
- Make it possible to use one-dimensional pattern recognizers
- How many clusters? E.g. 256.
- Register the first 256 one-point clusters.
- Take the 257th point and find its distance X to its closest neighbor (out of the already registered 256 points).
- If X is greater than the smallest distance of any pair of the already registered 256, then it is a new one-point cluster.
- Collapse the two one-point clusters that are closest together into a single cluster.
- Process remaining points while always maintaining 256 clusters.
- Find each cluster's geometric center point (vector).