Changeset 1241 for cpp/frams/genetics/f4

Timestamp:

05/18/23 03:43:42 (18 months ago)

Author:

Maciej Komosinski

Message:

No longer sort modifiers and cancel out antagonistic modifiers in f1 and f4; simplifying modifier sequences is now much less intrusive to allow for 2^{N distinct values of properties instead of only 2*N that resulted from earlier forced ordering (N is the number of same-letter upper- and lower-case characters in a modifier sequence)}

Location:

cpp/frams/genetics/f4

Files:

• f4_general.cpp (modified) (3 diffs)
• f4_oper.cpp (modified) (1 diff)

cpp/frams/genetics/f4/f4_general.cpp

@@ -1363 +1363 @@
                         // in the future this could be generalized to all neuron properties, for example N:|:power:0.6:range:1.4, or can even use '=' or ',' instead of ':' if no ambiguity
                         char prop_dir, prop_symbol, prop_end[2]; // prop_end is only to ensure that neuron parameter definition is completed
-                        if (sscanf(genot + pos_inout, ":%c%c%1[:]", &prop_dir, &prop_symbol, &prop_end) != 3)
+                        if (sscanf(genot + pos_inout, ":%c%c%1[:]", &prop_dir, &prop_symbol, prop_end) != 3)
                                 // error: incorrect format
                                 return pos_inout + 1 + 1;
@@ -1402 +1402 @@
 #ifdef F4_SIMPLIFY_MODIFIERS
                         char *ptr = (char*)(genot + pos_inout);
-
-#ifdef __BORLANDC__ // "[bcc32c Error] cannot compile this non-trivial TLS destruction yet" (C++B 10.4u2)
-                        static
-#else
-                        thread_local
-#endif
-                                vector<int> modifs_counts(strlen(all_modifiers_no_comma)); ///<an array with a known constant size storing counters of each modifier symbol from all_modifiers_no_comma, created once to avoid reallocation every time when modifier genes are simplified during parsing. Initialization of required size; it will never be resized.
-                        std::fill(modifs_counts.begin(), modifs_counts.end(), 0); //zeroing only needed if we encountered a char from all_modifiers_no_comma and enter the 'while' loop below
-
-                        while (char *m = GenoOperators::strchrn0(all_modifiers_no_comma, *ptr)) //only processes a section of chars known in all_modifiers_no_comma, other characters will exit the loop
-                        {
-                                modifs_counts[m - all_modifiers_no_comma]++;
+                        string original = "";
+                        while (GenoOperators::strchrn0(all_modifiers_no_comma, *ptr)) //only processes a section of chars known in all_modifiers_no_comma, other characters will exit the loop
+                        {
+                                original += *ptr;
                                 GenoOperators::skipWS(++ptr); //advance and ignore whitespace
                         }
@@ -1419 +1411 @@
                         if (advanced > 0) //found modifiers
                         {
-                                string simplified = GenoOperators::simplifiedModifiers(all_modifiers_no_comma, modifs_counts);
+                                string simplified = GenoOperators::simplifiedModifiers(original);
                                 // add a node for each char in "simplified"
                                 for (size_t i = 0; i < simplified.length(); i++)

cpp/frams/genetics/f4/f4_oper.cpp

@@ -18 +18 @@
 // TODO add support for properties of (any class of) neurons - not just sigmoid/force/intertia (':' syntax) for N
 // TODO add mapping genotype character ranges for neural [connections]
-// TODO for some genotypes, #defining/undefining F4_SIMPLIFY_MODIFIERS produces significantly different phenotypes (e.g. length of some Joint changes from 1.25 to 1.499, coordinates of Parts change, friction of some part changes from 1.28 to 0.32). Comparing f4_Node trees, the simplification works as intended, there are no huge changes apart from removing contradicting modifiers like 'R' and 'r' or 'L' and 'l', and dispersing the modifiers (changed order). There is no reason for such a significant influence of this. A hypothesis is that something may be wrong with calculating the influence of individual modifiers, e.g. some strong nonlinearity is introduced where it should not be, or some compensation between modifiers that should not influence each other (like L and R), or some modifier f4_Nodes are skipped/ignored when applying? Investigate. Example genotype that displays this issue: /*4*/,i<qlM,C<X>N:*#1>>,r<MRF<Xcm>N:Gpart>#5#1#2MLL#1>#1>>>>#5ML#2L#1>>>Lf,r<#1>rM<CqmLlCfqiFLqXFfl><F,<<XI>iN:|[-1:4.346]><XF><<XrRQ>N:G#3>>QiXFMR>fXM#2MfcR>R#3>>X
 // TODO The f0 genotypes for /*4*/<<RX>X>X> and RX(X,X) are identical, but if you replace R with Q or C, there are small differences - check why and perhaps unify?
+// TODO F4_SIMPLIFY_MODIFIERS in f4_general.cpp: currently it works while parsing (which is a bit "cheating": we get a phenotype that is a processed version of the genotype, thus some changes in modifiers in the genotype have no effect on its phenotype). Another (likely better) option, instead of simplifying while parsing, would be during mutations (like it is done in f1): when mutations add/modify/remove a modifier node, they could "clean" the tree by simplifying modifiers on the same subpath just as GenoOperators::simplifiedModifiers() does. This way, simplifying would be only performed when we actually modify a part of a genotype, not each time we interpret it, and there would be no hidden mechanism: all visible genes would have an expected effect on the phenotype.