More documentation

README.md

@@ -2,7 +2,7 @@
 
 For now, just modify the script to your liking:
 
-`julia paralleleureqa.jl`
+`JULIA_NUM_THREADS=8 julia paralleleureqa.jl`
 
 ## Modification
 
@@ -11,6 +11,10 @@ You can change the binary and unary operators in `eureqa.jl` here:
 const binops = [plus, mult]
 const unaops = [sin, cos, exp];
 ```
+E.g., you can add another binary function with:
+```
+const binops = [plus, mult, (x, y)->x^2*y]
+```
 
 You can change the dataset here:
 ```
@@ -19,12 +23,34 @@ const X = rand(100, nvar);
 # Here is the function we want to learn (x2^2 + cos(x3) + 5)
 const y = ((cx,)->cx^2).(X[:, 2]) + cos.(X[:, 3]) .+ 5.0;
 ```
-
-The default number of processes is 10; this is set with
-`addprocs(10)` in `paralleleureqa.jl`.
+by either loading in a dataset, or modifying the definition of `y`.
 
 ### Hyperparameters
 
 Turn on annealing by setting the following in `paralleleureqa.jl`:
 
 `const annealing = true`
+
+Annealing allows each evolutionary cycle to turn down the exploration
+rate over time: at the end (temperature 0), it will only select solutions
+better than existing solutions.
+
+The following parameter, parsimony, is how much to punish complex solutions:
+`
+const parsimony = 0.01
+`
+
+Finally, the following
+determins how much to scale temperature by (T between 0 and 1).
+`
+const alpha = 10.0
+`
+Larger alpha means more exploration.
+
+One can also adjust the relative probabilities of each mutation here:
+```
+weights = [8, 1, 1, 1]
+```
+(for: 1. perturb constant, 2. mutate operator,
+3. append a node, 4. delete a subtree).
+

paralleleureqa.jl

@@ -29,7 +29,7 @@ for k=1:niterations
     # Migration
     for j=1:nthreads
         for k in rand(1:npop, 50)
-            # Copy in case one gets copied twice
+            # Copy in case one gets used twice
             allPops[j].members[k] = deepcopy(bestPops.members[rand(1:size(bestPops.members)[1])])
         end
     end