Spaces:
Sleeping
Sleeping
Commit
•
a65054b
1
Parent(s):
d67af40
Rearrange operators section
Browse files- docs/options.md +26 -26
docs/options.md
CHANGED
|
@@ -3,11 +3,11 @@
|
|
| 3 |
You likely don't need to tune the hyperparameters yourself,
|
| 4 |
but if you would like, you can use `hyperparamopt.py` as an example.
|
| 5 |
|
| 6 |
-
Common options
|
|
|
|
| 7 |
- `niterations`
|
| 8 |
- `procs`
|
| 9 |
- `populations`
|
| 10 |
-
- `binary_operators`, `unary_operators`
|
| 11 |
- `weights`
|
| 12 |
- `maxsize`, `maxdepth`
|
| 13 |
- `batching`, `batchSize`
|
|
@@ -22,29 +22,7 @@ at the end of every iteration,
|
|
| 22 |
which is `hall_of_fame.csv` by default. It also prints the
|
| 23 |
equations to stdout.
|
| 24 |
|
| 25 |
-
##
|
| 26 |
-
|
| 27 |
-
This is the total number of generations that `pysr` will run for.
|
| 28 |
-
I usually set this to a large number, and exit when I am satisfied
|
| 29 |
-
with the equations.
|
| 30 |
-
|
| 31 |
-
## Processors
|
| 32 |
-
|
| 33 |
-
One can adjust the number of workers used by Julia with the
|
| 34 |
-
`procs` option. You should set this equal to the number of cores
|
| 35 |
-
you want `pysr` to use. This will also run `procs` number of
|
| 36 |
-
populations simultaneously by default.
|
| 37 |
-
|
| 38 |
-
## Populations
|
| 39 |
-
|
| 40 |
-
By default, `populations=procs`, but you can set a different
|
| 41 |
-
number of populations with this option. More populations may increase
|
| 42 |
-
the diversity of equations discovered, though will take longer to train.
|
| 43 |
-
However, it may be more efficient to have `populations>procs`,
|
| 44 |
-
as there are multiple populations running
|
| 45 |
-
on each core.
|
| 46 |
-
|
| 47 |
-
## Custom operators
|
| 48 |
|
| 49 |
A list of operators can be found on the operators page.
|
| 50 |
One can define custom operators in Julia by passing a string:
|
|
@@ -68,7 +46,29 @@ so that the SymPy code can understand the output equation from Julia,
|
|
| 68 |
when constructing a useable function. This step is optional, but
|
| 69 |
is necessary for the `lambda_format` to work.
|
| 70 |
|
| 71 |
-
One can also edit `operators.jl`.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 72 |
|
| 73 |
## Weighted data
|
| 74 |
|
|
|
|
| 3 |
You likely don't need to tune the hyperparameters yourself,
|
| 4 |
but if you would like, you can use `hyperparamopt.py` as an example.
|
| 5 |
|
| 6 |
+
Common options to `PySR` include:
|
| 7 |
+
- `binary_operators`, `unary_operators`
|
| 8 |
- `niterations`
|
| 9 |
- `procs`
|
| 10 |
- `populations`
|
|
|
|
| 11 |
- `weights`
|
| 12 |
- `maxsize`, `maxdepth`
|
| 13 |
- `batching`, `batchSize`
|
|
|
|
| 22 |
which is `hall_of_fame.csv` by default. It also prints the
|
| 23 |
equations to stdout.
|
| 24 |
|
| 25 |
+
## Operators
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
|
| 27 |
A list of operators can be found on the operators page.
|
| 28 |
One can define custom operators in Julia by passing a string:
|
|
|
|
| 46 |
when constructing a useable function. This step is optional, but
|
| 47 |
is necessary for the `lambda_format` to work.
|
| 48 |
|
| 49 |
+
One can also edit `operators.jl`.
|
| 50 |
+
|
| 51 |
+
## Iterations
|
| 52 |
+
|
| 53 |
+
This is the total number of generations that `pysr` will run for.
|
| 54 |
+
I usually set this to a large number, and exit when I am satisfied
|
| 55 |
+
with the equations.
|
| 56 |
+
|
| 57 |
+
## Processors
|
| 58 |
+
|
| 59 |
+
One can adjust the number of workers used by Julia with the
|
| 60 |
+
`procs` option. You should set this equal to the number of cores
|
| 61 |
+
you want `pysr` to use. This will also run `procs` number of
|
| 62 |
+
populations simultaneously by default.
|
| 63 |
+
|
| 64 |
+
## Populations
|
| 65 |
+
|
| 66 |
+
By default, `populations=procs`, but you can set a different
|
| 67 |
+
number of populations with this option. More populations may increase
|
| 68 |
+
the diversity of equations discovered, though will take longer to train.
|
| 69 |
+
However, it may be more efficient to have `populations>procs`,
|
| 70 |
+
as there are multiple populations running
|
| 71 |
+
on each core.
|
| 72 |
|
| 73 |
## Weighted data
|
| 74 |
|