Vector Connectivity
A vector rule is a list of integer offsets [dx, dy, dz]. For every source
neuron at (x, y, z), FEAGI tries to create a synapse to the destination
neuron at (x+dx, y+dy, z+dz) when that voxel exists.
Click the guide icon in the vector editor to return here.
Editor format
Each row is one offset:
| dx | dy | dz | Result |
|---|---|---|---|
| 0 | 0 | 0 | Identity: dest voxel equals source voxel |
| 1 | 0 | 0 | One step in +X |
| 0 | 0 | 1 | One step in +Z (next layer) |
| -1 | 0 | 0 | One step in -X |
Add multiple rows to union several regular offsets (for example both +X and -X).
This is not a list of source-to-destination coordinate pairs. For filters
such as "only X 1 through 98" use a pattern rule
with N..M, not a vector list.
When to use vectors
- Uniform topographic shift
- Identity (
[0,0,0], same as coreblock_to_block) - A small set of regular neighbor offsets
Do not enumerate hundreds of unique channel offsets that are really one
pattern. Prefer ["1..98", "*", "*"] -> ["?", "?", 0] for a contiguous
source-X block onto dest Z=0.
Bounds
If src + offset leaves the destination area, that source neuron simply gets
no synapse for that vector. No wrap-around.
Related
- Pattern Connectivity - wildcards,
N..M,?+N - Connectivity Rules - rule types and manager
- Mapping Connections - applying a rule to a mapping