| … | … | @@ -78,7 +78,8 @@ the `add-pair-stars` API, which, given the above, adds decorations |
| 78 | 78 | that almost all otherlayers above need. These are: |
| 79 | 79 | |
| 80 | 80 | 7) Lists of the left and right basis elements. The left basis is the |
| 81 | | set. `{x | (x,y) exists in the atomspace for some y}` |
| 81 | set. `{x | (x,y) exists in the atomspace for some y}` (Provided by |
| 82 | the `'left-basis` and `'right-basis` methods.) |
| 82 | 83 | 8) Size of the left and right basis |
| 83 | 84 | 9) True/false membership predicates: does an Atom appear in the left |
| 84 | 85 | or right basis set? (These are provided by the `'in-left-basis?` |
| … | … | @@ -232,4 +233,20 @@ Item (8): Use SizeOfLink |
| 232 | 233 | |
| 233 | 234 | Item (9): Does not appear to be used anywhere in the code. |
| 234 | 235 | |
| 235 | | Item (10) |
| 236 | Item (10): ... Almost unused; the `learn` code uses this in one place |
| 237 | for set subtraction, and another place for agglomerative MI. |
| 238 | |
| 239 | ### Conclusion |
| 240 | Simply using the `QueryLink` and caching the rewrite results is |
| 241 | sufficient for use to replace the parametric polymorphism base design. |
| 242 | Nothing more needs to be done, here. (Except possibly more testing, |
| 243 | and moving the search result keys to the rewrites from the query |
| 244 | itself.) |
| 245 | |
| 246 | Compute processing |
| 247 | ------------------ |
| 248 | The matrix code has two parts: first, the specification of the base |
| 249 | matrix, and second, the computation of the assorted similarity measures, |
| 250 | given the base matrix. |
| 251 | |
| 252 | The above deals with the base. What about the similarities? |