major changes: refactored params and converted model_opt to TPE (GA d…

…eprecated)

conda-recipe/meta.yaml

@@ -21,10 +21,12 @@ requirements:
   - pip
   - perl
   - pytest
+  - hyperopt
   - numpy
   - pandas>2
   - matplotlib
   - scikit-learn
+  - lightgbm
   - seaborn
   - geopandas
   - sortedcontainers

scripts/ensembleResistnet.py

@@ -7,7 +7,7 @@
 import glob
 import pandas as pd
 from resistnet.resistance_network import ResistanceNetwork
-from resistnet.model_optimisation import ModelRunner
+from resistnet.model_optimisation import ModelRunnerGA
 from resistnet.hall_of_fame import HallOfFame
 
 # Set environment variable
@@ -95,7 +95,7 @@ def main():
     )
 
     # Step 3: Set up ModelRunner
-    runner = ModelRunner(
+    runner = ModelRunnerGA(
         resistance_network=network,
         seed=params.seed,
         verbose=True

scripts/runResistnet.py

@@ -5,7 +5,7 @@
 
 from resistnet.params import parseArgs
 from resistnet.resistance_network import ResistanceNetwork
-from resistnet.model_optimisation import ModelRunner
+from resistnet.model_optimisation import ModelRunnerTPE, ModelRunnerGA
 
 
 def main():
@@ -27,7 +27,6 @@ def main():
         coords=params.coords,
         variables=params.variables,
         agg_opts=params.agg_opts,
-        pop_agg=params.pop_agg,
         inmat=params.inmat,
         reachid_col=params.reachid_col,
         length_col=params.length_col,
@@ -39,8 +38,7 @@ def main():
     if params.shapefile:
         network.write_geodataframe(params.out, params.output_driver)
 
-    # Step 2: Initialise ModelRunner
-    runner = ModelRunner(
+    runner = ModelRunnerTPE(
         resistance_network=network,
         seed=params.seed,
         verbose=True
@@ -50,41 +48,73 @@ def main():
     if params.gridSearch:
         runner.optimise_univariate(
             fitmetric=params.fitmetric,
-            threads=params.GA_procs,
+            threads=params.procs,
             max_shape=params.max_shape,
             out=params.out,
             plot=True,
             verbose=True
         )
 
-    # Step 3: Run GA optimisation
-    runner.run_ga(
-        maxgens=params.maxGens,
+    # Step 3: Run optimisation
+    runner.run_tpe(
+        max_evals=params.max_iter,
         fitmetric=params.fitmetric,
-        threads=params.GA_procs,
-        burnin=params.burnin,
-        deltaB=params.deltaB,
-        deltaB_perc=params.deltaB_perc,
-        indpb=params.indpb,
-        mutpb=params.mutpb,
-        cxpb=params.cxpb,
+        threads=params.procs,
         nFail=params.nfail,
-        popsize=params.popsize,
-        maxpopsize=params.maxpopsize,
         fixWeight=params.fixWeight,
         fixShape=params.fixShape,
         min_weight=params.min_weight,
         max_shape=params.max_shape,
         max_hof_size=params.max_hof_size,
         awsum=params.awsum,
-        only_keep=params.only_keep,
+        only_keep=True,
         use_full=params.use_full,
         fixed_params=params.transFile,
         out=params.out,
         plot=True,
-        verbose=True
+        verbose=True,
+        reps=params.reps,
+        n_startup=params.nstart,
+        n_candidates=params.ncand,
+        gamma=params.gamma
     )
 
+    # # Step 2: Initialise ModelRunner
+    # runner = ModelRunnerGA(
+    #     resistance_network=network,
+    #     seed=params.seed,
+    #     verbose=True
+    # )
+
+
+    # # Step 3: Run GA optimisation
+    # runner.run_ga(
+    #     maxgens=params.maxGens,
+    #     fitmetric=params.fitmetric,
+    #     threads=params.GA_procs,
+    #     burnin=params.burnin,
+    #     deltaB=params.deltaB,
+    #     deltaB_perc=params.deltaB_perc,
+    #     indpb=params.indpb,
+    #     mutpb=params.mutpb,
+    #     cxpb=params.cxpb,
+    #     nFail=params.nfail,
+    #     popsize=params.popsize,
+    #     maxpopsize=params.maxpopsize,
+    #     fixWeight=params.fixWeight,
+    #     fixShape=params.fixShape,
+    #     min_weight=params.min_weight,
+    #     max_shape=params.max_shape,
+    #     max_hof_size=params.max_hof_size,
+    #     awsum=params.awsum,
+    #     only_keep=params.only_keep,
+    #     use_full=params.use_full,
+    #     fixed_params=params.transFile,
+    #     out=params.out,
+    #     plot=True,
+    #     verbose=True
+    # )
+
 
 # Call main function
 if __name__ == '__main__':

src/resistnet/hall_of_fame.py

@@ -44,7 +44,7 @@ def __init__(self, variables, max_size, use_full=False, init_pop=None):
         cols = ["fitness"]
         for v in variables:
             cols.extend([
-                str(v), f"{v}_weight", f"{v}_trans", f"{v}_shape", f"{v}_asym"
+                str(v), f"{v}_weight", f"{v}_trans", f"{v}_shape"
             ])
         cols.extend(["loglik", "r2m", "aic", "delta_aic_null"])
         self.data = pd.DataFrame(columns=cols)
@@ -350,7 +350,6 @@ def get_best_model(self):
                 "Weight": best_model[f"{var}_weight"],
                 "Shape": best_model[f"{var}_shape"],
                 "Transformation": best_model[f"{var}_trans"],
-                "Asymmetry": best_model[f"{var}_asym"],
             }
             for var in self.variables
         ]
@@ -433,7 +432,7 @@ def cleanHOF(self):
 
         for v in self.variables:
             mask = ret[v] == 0
-            for attr in ["_weight", "_trans", "_shape", "_asym"]:
+            for attr in ["_weight", "_trans", "_shape"]:
                 ret.loc[mask, f"{v}{attr}"] = np.nan
 
         if ret.iloc[0]["fitness"] == (ret.iloc[0]["aic"] * -1):
@@ -732,7 +731,7 @@ def from_dataframe(cls, df, max_size=None):
                         the DataFrame.
         """
         # Extract variable names from the DataFrame column names
-        variable_prefixes = ["_weight", "_trans", "_shape", "_asym"]
+        variable_prefixes = ["_weight", "_trans", "_shape"]
         variables = set()
         for col in df.columns:
             for prefix in variable_prefixes:
@@ -760,5 +759,6 @@ def from_dataframe(cls, df, max_size=None):
         instance.data = df.copy()
         instance.data = instance.drop_active_duplicates(instance.data,
                                                         instance.variables)
+        instance._update_data_frame()
 
         return instance