Fixes for learning ICN with CBLS (#19)

* Compat, CI, and Parallelism: release (#18)

* Cleaning and CI updates (#13)

* Cleaning and CI updates

* Fix CI for Julia 1.8

* Compat ci (#17)

* Cleaning and CI updates (#13) (#14)

* Cleaning and CI updates

* Fix CI for Julia 1.8

* Compat and CI

* Fixes for learning ICN aith CBLS

src/icn/cbls.jl

@@ -91,8 +91,13 @@ function CompositionalNetworks.optimize!(
     @debug "pool" s.pool best_values(s.pool) best_values(s) s.pool.configurations
 
     # Return best values
-    best = BitVector(collect(best_values(s)))
-    weights!(icn, best)
+
+    if has_solution(s)
+        weights!(icn, BitVector(collect(best_values(s))))
+    else
+        CompositionalNetworks.generate_weights(icn)
+    end
+    best = weights(icn)
 
     return best, Dictionary{BitVector, Int}([best], [1])
 end