Saving and loading populations#
Another feature Brush implements is the ability to save and load entire populations. We use JSON notation to store the population into a file that is human readable. The same way, we can feed an estimator a previous population file to serve as starting point for the evolution.
In this notebook, we will walk through how to use the save_population and load_population parameters.
We start by getting a sample dataset and splitting it into X and y:
import pandas as pd
from pybrush import BrushRegressor
# load data
df = pd.read_csv('../examples/datasets/d_enc.csv')
X = df.drop(columns='label')
y = df['label']
To save the population after finishing the evolution, you nee to set save_population parameter to a value different than an empty string. Then, the final population is going to be stored in that specific file.
In this example, we create a temporary file.
import pickle
import os, tempfile
pop_file = os.path.join(tempfile.mkdtemp(), 'population.json')
# set verbosity==2 to see the full report
est = BrushRegressor(
functions=['SplitBest','Add','Mul','Sin','Cos','Exp','Logabs'],
max_gens=10,
save_population=pop_file,
verbosity=2
)
est.fit(X,y)
y_pred = est.predict(X)
print('score:', est.score(X,y))
Generation 1/10 [////// ]
Train Loss (Med): 11.75939 (74.37032)
Val Loss (Med): 11.75939 (74.37032)
Median Size (Max): 3 (19)
Median complexity (Max): 9 (432)
Time (s): 0.12205
Generation 2/10 [/////////// ]
Train Loss (Med): 11.58283 (17.94969)
Val Loss (Med): 11.58283 (17.94969)
Median Size (Max): 3 (19)
Median complexity (Max): 9 (368)
Time (s): 0.27800
Generation 3/10 [//////////////// ]
Train Loss (Med): 11.15674 (17.94969)
Val Loss (Med): 11.15674 (17.94969)
Median Size (Max): 3 (20)
Median complexity (Max): 10 (915)
Time (s): 0.41845
Generation 4/10 [///////////////////// ]
Train Loss (Med): 10.62121 (17.94969)
Val Loss (Med): 10.62121 (17.94969)
Median Size (Max): 3 (20)
Median complexity (Max): 9 (381)
Time (s): 0.56585
Generation 5/10 [////////////////////////// ]
Train Loss (Med): 10.51181 (17.94969)
Val Loss (Med): 10.51181 (17.94969)
Median Size (Max): 3 (20)
Median complexity (Max): 9 (412)
Time (s): 0.73561
Generation 6/10 [/////////////////////////////// ]
Train Loss (Med): 10.51181 (17.94969)
Val Loss (Med): 10.51181 (17.94969)
Median Size (Max): 3 (20)
Median complexity (Max): 9 (412)
Time (s): 0.89526
Generation 7/10 [//////////////////////////////////// ]
Train Loss (Med): 10.51181 (17.94969)
Val Loss (Med): 10.51181 (17.94969)
Median Size (Max): 3 (20)
Median complexity (Max): 9 (412)
Time (s): 1.03213
Generation 8/10 [///////////////////////////////////////// ]
Train Loss (Med): 10.43982 (17.94969)
Val Loss (Med): 10.43982 (17.94969)
Median Size (Max): 3 (20)
Median complexity (Max): 9 (412)
Time (s): 1.19282
Generation 9/10 [////////////////////////////////////////////// ]
Train Loss (Med): 10.33524 (17.94969)
Val Loss (Med): 10.33524 (17.94969)
Median Size (Max): 3 (20)
Median complexity (Max): 9 (368)
Time (s): 1.33781
Generation 10/10 [//////////////////////////////////////////////////]
Train Loss (Med): 10.33524 (17.94969)
Val Loss (Med): 10.33524 (17.94969)
Median Size (Max): 3 (20)
Median complexity (Max): 9 (368)
Time (s): 1.50192
Saved population to file /tmp/tmpw7jkwa5m/population.json
score: 0.8856532915521027
Loading a previous population is done providing load_population a string value corresponding to a JSON file generated by Brush. In our case, we will use the same file from the previous code block.
After loading the population, we run the evolution for 10 more generations, and we can see that the first generation started from the previous population. This means that the population was successfully saved and loaded.
est = BrushRegressor(
functions=['SplitBest','Add','Mul','Sin','Cos','Exp','Logabs'],
load_population=pop_file,
max_gens=10,
verbosity=2
)
est.fit(X,y)
y_pred = est.predict(X)
print('score:', est.score(X,y))
Loaded population from /tmp/tmpw7jkwa5m/population.json of size = 200
Generation 1/10 [////// ]
Train Loss (Med): 10.33524 (17.94969)
Val Loss (Med): 10.33524 (17.94969)
Median Size (Max): 3 (20)
Median complexity (Max): 9 (368)
Time (s): 0.16596
Generation 2/10 [/////////// ]
Train Loss (Med): 10.33524 (17.94969)
Val Loss (Med): 10.33524 (17.94969)
Median Size (Max): 3 (18)
Median complexity (Max): 9 (240)
Time (s): 0.31669
Generation 3/10 [//////////////// ]
Train Loss (Med): 10.26326 (17.94969)
Val Loss (Med): 10.26326 (17.94969)
Median Size (Max): 3 (20)
Median complexity (Max): 9 (368)
Time (s): 0.45045
Generation 4/10 [///////////////////// ]
Train Loss (Med): 10.26326 (17.94969)
Val Loss (Med): 10.26326 (17.94969)
Median Size (Max): 3 (19)
Median complexity (Max): 9 (368)
Time (s): 0.63331
Generation 5/10 [////////////////////////// ]
Train Loss (Med): 10.26326 (16.41696)
Val Loss (Med): 10.26326 (16.41696)
Median Size (Max): 5 (17)
Median complexity (Max): 33 (330)
Time (s): 0.78002
Generation 6/10 [/////////////////////////////// ]
Train Loss (Med): 9.70269 (17.94969)
Val Loss (Med): 9.70269 (17.94969)
Median Size (Max): 3 (19)
Median complexity (Max): 9 (330)
Time (s): 0.91656
Generation 7/10 [//////////////////////////////////// ]
Train Loss (Med): 9.67577 (17.94969)
Val Loss (Med): 9.67577 (17.94969)
Median Size (Max): 3 (19)
Median complexity (Max): 9 (330)
Time (s): 1.10225
Generation 8/10 [///////////////////////////////////////// ]
Train Loss (Med): 9.67577 (16.41696)
Val Loss (Med): 9.67577 (16.41696)
Median Size (Max): 5 (19)
Median complexity (Max): 33 (330)
Time (s): 1.30773
Generation 9/10 [////////////////////////////////////////////// ]
Train Loss (Med): 9.67577 (16.41696)
Val Loss (Med): 9.67577 (16.41696)
Median Size (Max): 5 (19)
Median complexity (Max): 33 (330)
Time (s): 1.44840
Generation 10/10 [//////////////////////////////////////////////////]
Train Loss (Med): 9.67577 (15.67545)
Val Loss (Med): 9.67577 (15.67545)
Median Size (Max): 6 (19)
Median complexity (Max): 36 (723)
Time (s): 1.65144
score: 0.892949582824199
You can open the serialized file and change individuals’ programs manually.
This also allow us to have checkpoints in the execution.