Spaces:
Sleeping
Sleeping
Commit
·
5eeffc0
1
Parent(s):
37f71ff
Fix syntax highlighting
Browse files
README.md
CHANGED
|
@@ -35,13 +35,13 @@ const binops = [plus, mult]
|
|
| 35 |
const unaops = [sin, cos, exp];
|
| 36 |
```
|
| 37 |
E.g., you can add the function for powers with:
|
| 38 |
-
```
|
| 39 |
pow(x::Float32, y::Float32)::Float32 = sign(x)*abs(x)^y
|
| 40 |
const binops = [plus, mult, pow]
|
| 41 |
```
|
| 42 |
|
| 43 |
You can change the dataset here:
|
| 44 |
-
```
|
| 45 |
const X = convert(Array{Float32, 2}, randn(100, 5)*2)
|
| 46 |
# Here is the function we want to learn (x2^2 + cos(x3))
|
| 47 |
const y = convert(Array{Float32, 1}, ((cx,)->cx^2).(X[:, 2]) + cos.(X[:, 3]))
|
|
@@ -55,19 +55,19 @@ rate over time: at the end (temperature 0), it will only select solutions
|
|
| 55 |
better than existing solutions.
|
| 56 |
|
| 57 |
The following parameter, parsimony, is how much to punish complex solutions:
|
| 58 |
-
|
| 59 |
const parsimony = 0.01
|
| 60 |
-
|
| 61 |
|
| 62 |
Finally, the following
|
| 63 |
determins how much to scale temperature by (T between 0 and 1).
|
| 64 |
-
|
| 65 |
const alpha = 10.0
|
| 66 |
-
|
| 67 |
Larger alpha means more exploration.
|
| 68 |
|
| 69 |
One can also adjust the relative probabilities of each mutation here:
|
| 70 |
-
```
|
| 71 |
weights = [8, 1, 1, 1, 2]
|
| 72 |
```
|
| 73 |
(for: 1. perturb constant, 2. mutate operator,
|
|
|
|
| 35 |
const unaops = [sin, cos, exp];
|
| 36 |
```
|
| 37 |
E.g., you can add the function for powers with:
|
| 38 |
+
```julia
|
| 39 |
pow(x::Float32, y::Float32)::Float32 = sign(x)*abs(x)^y
|
| 40 |
const binops = [plus, mult, pow]
|
| 41 |
```
|
| 42 |
|
| 43 |
You can change the dataset here:
|
| 44 |
+
```julia
|
| 45 |
const X = convert(Array{Float32, 2}, randn(100, 5)*2)
|
| 46 |
# Here is the function we want to learn (x2^2 + cos(x3))
|
| 47 |
const y = convert(Array{Float32, 1}, ((cx,)->cx^2).(X[:, 2]) + cos.(X[:, 3]))
|
|
|
|
| 55 |
better than existing solutions.
|
| 56 |
|
| 57 |
The following parameter, parsimony, is how much to punish complex solutions:
|
| 58 |
+
```julia
|
| 59 |
const parsimony = 0.01
|
| 60 |
+
```
|
| 61 |
|
| 62 |
Finally, the following
|
| 63 |
determins how much to scale temperature by (T between 0 and 1).
|
| 64 |
+
```julia
|
| 65 |
const alpha = 10.0
|
| 66 |
+
```
|
| 67 |
Larger alpha means more exploration.
|
| 68 |
|
| 69 |
One can also adjust the relative probabilities of each mutation here:
|
| 70 |
+
```julia
|
| 71 |
weights = [8, 1, 1, 1, 2]
|
| 72 |
```
|
| 73 |
(for: 1. perturb constant, 2. mutate operator,
|