WIP Preference Learning

solidago/src/solidago/_pipeline/_preference_learning/generalized_bradley_terry.py

@@ -129,7 +129,7 @@ def init_scores(self,
         entity_name2index: dict[str, int],
         init_multiscores: MultiScore, # key_names == "entity_name"
     ) -> npt.NDArray:
-        scores = np.zeros(len(entity_name2index))
+        scores = np.zeros(len(entity_name2index), dtype=np.float64)
         for entity, init_score in init_multiscores:
             if not init_score.isnan():
                 scores[entity_name2index[str(entity)]] = init_score.value
@@ -171,10 +171,10 @@ def compute_uncertainties(self,
         rights: npt.NDArray
             rights[i] is the right uncertainty on scores[i]
         """
-        compared_entity_indices = comparisons.compared_entity_indices(entity_name2index)
-        indices = { loc: np.array(compared_entity_indices[loc]) for loc in ("left", "right") }
+        indices = comparisons.compared_entity_indices(entity_name2index, self.last_comparison_only)
+        indices = { loc: np.array(indices[loc]) for loc in ("left", "right") }
         score_diffs = scores[indices["left"]] - scores[indices["right"]]
-        normalized_comparisons = comparisons.normalized_comparisons()
+        normalized_comparisons = comparisons.normalized_comparisons(self.last_comparison_only)
         score_negative_log_likelihood = self.negative_log_likelihood(score_diffs, normalized_comparisons)
 
         kwargs = dict(

solidago/src/solidago/_pipeline/_preference_learning/lbfgs_generalized_bradley_terry.py

@@ -28,6 +28,7 @@ def __init__(self,
         convergence_error: float=1e-5,
         max_iter: int=100,
         device: torch.device=default_device,
+        last_comparison_only: bool=True,
     ):
         """ Generalized Bradley Terry is a class of porbability models of comparisons,
         introduced in the paper "Generalized Bradley-Terry Models for Score Estimation 
@@ -58,6 +59,7 @@ def __init__(self,
             prior_std_dev=prior_std_dev,
             uncertainty_nll_increase=uncertainty_nll_increase,
             max_uncertainty=max_uncertainty,
+            last_comparison_only=last_comparison_only,
         )
         self.convergence_error = convergence_error
         self.max_iter = max_iter
@@ -124,9 +126,9 @@ def negative_log_posterior(self,
         comparisons: Comparisons,
     ) -> torch.Tensor:
         """ Negative log posterior """
-        entity_indices = comparisons.compared_entity_indices(entity_name2index)
-        score_diffs = scores[entity_indices["left"]] - scores[entity_indices["right"]]
-        normalized_comparisons = comparisons.normalized_comparisons()
+        indices = comparisons.compared_entity_indices(entity_name2index, self.last_comparison_only)
+        score_diffs = scores[indices["left"]] - scores[indices["right"]]
+        normalized_comparisons = comparisons.normalized_comparisons(self.last_comparison_only)
         loss = self.cumulant_generating_function(score_diffs).sum()
         loss -= (score_diffs * torch.tensor(normalized_comparisons)).sum()
         return loss + (scores**2).sum() / (2 * self.prior_std_dev**2)
@@ -140,6 +142,7 @@ def __init__(self,
         convergence_error: float=1e-5,
         max_iter: int=100,
         device: torch.device=default_device,
+        last_comparison_only: bool=True,
     ):
         """
         Parameters (TODO)
@@ -152,6 +155,7 @@ def __init__(self,
             convergence_error=convergence_error,
             max_iter=max_iter,
             device=device,
+            last_comparison_only=last_comparison_only
         )
 
     def cumulant_generating_function(self, score_diffs: torch.Tensor) -> torch.Tensor:

solidago/src/solidago/_pipeline/_preference_learning/numba_generalized_bradley_terry.py

@@ -18,6 +18,7 @@ def __init__(self,
         uncertainty_nll_increase: float=1.0,
         max_uncertainty: float=1e3,
         convergence_error: float=1e-5,
+        last_comparison_only: bool=True,
     ):
         """ Generalized Bradley Terry is a class of porbability models of comparisons,
         introduced in the paper "Generalized Bradley-Terry Models for Score Estimation 
@@ -45,6 +46,7 @@ def __init__(self,
             prior_std_dev=prior_std_dev,
             uncertainty_nll_increase=uncertainty_nll_increase,
             max_uncertainty=max_uncertainty,
+            last_comparison_only=last_comparison_only,
         )
         self.convergence_error = convergence_error
 
@@ -79,10 +81,10 @@ def compute_scores(self,
         entity_ordered_comparisons = comparisons.order_by_entities()
         def get_partial_derivative_args(entity_index: int, scores: np.ndarray) -> tuple:
             entity_name = entities.iloc[entity_index].name
-            df = entity_ordered_comparisons[entity_name].to_df()
-            normalized_comparisons = np.array(df["comparison"] / df["comparison_max"])
+            normalized_comparisons = comparisons.normalized_comparisons(self.last_comparison_only)
+            df = entity_ordered_comparisons[entity_name].to_df(last_row_only=self.last_comparison_only)
             indices = df["other_name"].map(entity_name2index)
-            return scores[indices], normalized_comparisons
+            return scores[indices], np.array(normalized_comparisons)
 
         return coordinate_descent(
             self.partial_derivative,
@@ -123,6 +125,7 @@ def __init__(self,
         uncertainty_nll_increase: float=1.0,
         max_uncertainty: float=1e3,
         convergence_error: float=1e-5,
+        last_comparison_only: bool=True,
     ):
         """
 
@@ -134,6 +137,7 @@ def __init__(self,
             uncertainty_nll_increase=uncertainty_nll_increase,
             max_uncertainty=max_uncertainty,
             convergence_error=convergence_error,
+            last_comparison_only=last_comparison_only,
         )
 
     @cached_property

solidago/src/solidago/_state/_comparisons/base.py

@@ -72,16 +72,16 @@ def order_by_entities(self) -> "Comparisons": # key_names == ["entity_name", "ot
     def compared_entity_indices(self, 
         entity_name2index: dict[str, int], 
         last_comparison_only: bool=True,
-        returns: Literal["rows", "row_list", "last_row"]="rows"
     ) -> dict[str, list[int]]:
         key_indices = { loc: self.key_names.index(f"{loc}_name") for loc in ("left", "right") }
+        returns = "last_row" if last_comparison_only else "rows"
         return {
             location: [ 
                 entity_name2index[keys[key_indices[location]]] 
                 for keys, _ in self.iter(returns)
             ] for location in ("left", "right")
         }
 
-    def normalized_comparisons(self) -> Series:
-        df = self.to_df()
+    def normalized_comparisons(self, last_comparison_only: bool) -> Series:
+        df = self.to_df(last_row_only=last_comparison_only)
         return df["comparison"] / df["comparison_max"]

solidago/src/solidago/primitives/datastructure/nested_dict_of_row_lists.py

@@ -58,14 +58,18 @@ def __len__(self) -> int:
             return sum([len(row_list) for _, row_list in self._dict.items()])
         return sum([ len(sub_dicts) for sub_dicts in self._dict.values() ])
 
-    def to_rows(self, row_kwargs: Optional[dict]) -> list[dict]:
+    def to_rows(self, row_kwargs: Optional[dict]=None, last_row_only: bool=False) -> list[dict]:
         if row_kwargs is None:
             row_kwargs = dict()
+        returns = "last_row" if last_row_only else "rows"
         return [
             dict(zip(self.key_names, keys)) | row_kwargs | row
-            for keys, row in self.iter(returns="rows", value_process=False, key_process=False)
+            for keys, row in self.iter(returns=returns, value_process=False, key_process=False)
         ]
 
+    def to_df(self, row_kwargs: Optional[dict]=None, last_row_only: bool=False) -> DataFrame:
+        return DataFrame(self.to_rows(row_kwargs, last_row_only))
+
     def iter(self, 
         returns: Literal["rows", "row_list", "last_row"]="rows", 
         value_process: bool=True,

solidago/tests/load_save/generated_state/assessments.csv

@@ -1,99 +1,85 @@
 username,criterion,entity_name,assessment,assessment_min,assessment_max
-user_0,default,entity_16,-2.0749057611785386,-inf,inf
-user_0,default,entity_15,-2.39370352625087,-inf,inf
-user_0,default,entity_11,-1.5992836978625538,-inf,inf
-user_0,default,entity_18,-2.619031854481638,-inf,inf
-user_0,default,entity_6,-1.2481876740748055,-inf,inf
-user_0,default,entity_17,-0.8195927297894479,-inf,inf
-user_0,default,entity_14,-1.2667207574469999,-inf,inf
-user_0,default,entity_19,0.8509795934379412,-inf,inf
-user_0,default,entity_4,3.053938668091461,-inf,inf
-user_1,default,entity_10,1.666205471029322,-inf,inf
-user_1,default,entity_15,1.3982102010500272,-inf,inf
-user_1,default,entity_7,0.31743364012724,-inf,inf
-user_1,default,entity_2,2.0757722679531048,-inf,inf
-user_1,default,entity_1,0.24777798376771804,-inf,inf
-user_1,default,entity_17,1.2172607977951482,-inf,inf
-user_1,default,entity_14,0.7587810143743123,-inf,inf
-user_1,default,entity_16,-0.45322198998446633,-inf,inf
-user_1,default,entity_3,-0.6293299841653897,-inf,inf
-user_1,default,entity_12,-2.053760441767217,-inf,inf
-user_1,default,entity_18,-0.9601193237500203,-inf,inf
-user_1,default,entity_11,-0.7764255417543154,-inf,inf
-user_2,default,entity_8,-3.1840313925258865,-inf,inf
-user_2,default,entity_12,-1.3582673358306874,-inf,inf
-user_2,default,entity_4,-0.3920343878647034,-inf,inf
-user_2,default,entity_13,-1.0307685022464559,-inf,inf
-user_2,default,entity_19,1.2140338299536473,-inf,inf
-user_2,default,entity_1,0.34488131370003594,-inf,inf
-user_2,default,entity_5,0.4963366907015018,-inf,inf
-user_2,default,entity_16,1.3239011724971825,-inf,inf
-user_3,default,entity_1,1.7159560105752791,-inf,inf
-user_3,default,entity_2,0.900609937541631,-inf,inf
-user_3,default,entity_10,0.010155796202459877,-inf,inf
-user_3,default,entity_19,1.9939222430641497,-inf,inf
-user_3,default,entity_18,-0.2487610576189404,-inf,inf
-user_3,default,entity_6,-1.201379294506745,-inf,inf
-user_3,default,entity_8,-1.989535014457278,-inf,inf
-user_3,default,entity_14,-1.421354527295856,-inf,inf
-user_3,default,entity_5,-0.32127542248226026,-inf,inf
-user_4,default,entity_8,0.28435161869848846,-inf,inf
-user_4,default,entity_6,2.4003825078321817,-inf,inf
-user_4,default,entity_19,0.42259228898088025,-inf,inf
-user_4,default,entity_11,0.055986495346031506,-inf,inf
-user_4,default,entity_17,0.2099984004735524,-inf,inf
-user_4,default,entity_13,0.4161872688546774,-inf,inf
-user_4,default,entity_9,1.6036019683037805,-inf,inf
-user_4,default,entity_4,-0.18158399657965602,-inf,inf
-user_4,default,entity_2,2.0454369805034163,-inf,inf
-user_4,default,entity_0,0.3854155927801495,-inf,inf
-user_4,default,entity_15,-0.14632039422818038,-inf,inf
-user_4,default,entity_5,-2.022605231132655,-inf,inf
-user_4,default,entity_18,-0.872532836583685,-inf,inf
-user_4,default,entity_3,-0.3827672112907701,-inf,inf
-user_4,default,entity_12,0.6017254822778608,-inf,inf
-user_4,default,entity_7,-2.9180266496509044,-inf,inf
-user_4,default,entity_16,-0.7599310550355487,-inf,inf
-user_5,default,entity_3,-2.1101274264627663,-inf,inf
-user_5,default,entity_14,-2.762781599519231,-inf,inf
-user_5,default,entity_2,-0.5100542990101968,-inf,inf
-user_5,default,entity_17,0.05710431263487662,-inf,inf
-user_6,default,entity_16,-0.06305869415402687,-inf,inf
-user_6,default,entity_6,-0.7677539107375391,-inf,inf
-user_6,default,entity_3,-2.8643609493870237,-inf,inf
-user_6,default,entity_11,-2.1819602163698275,-inf,inf
-user_6,default,entity_4,-0.43687483459641485,-inf,inf
-user_6,default,entity_9,-0.7354562201117167,-inf,inf
-user_6,default,entity_13,-1.6808275477504329,-inf,inf
-user_6,default,entity_8,0.9663711729296698,-inf,inf
-user_6,default,entity_14,0.621626369112255,-inf,inf
-user_6,default,entity_2,-0.35611410923840825,-inf,inf
-user_6,default,entity_10,1.408450182481708,-inf,inf
-user_6,default,entity_5,0.5043003970870001,-inf,inf
-user_6,default,entity_7,2.0966207883047416,-inf,inf
-user_7,default,entity_15,-0.5655431776581341,-inf,inf
-user_7,default,entity_12,-0.18604211953544048,-inf,inf
-user_7,default,entity_8,-0.719456200755322,-inf,inf
-user_7,default,entity_10,-0.824253089052783,-inf,inf
-user_7,default,entity_1,-1.1268342583835582,-inf,inf
-user_7,default,entity_3,1.6072561062359942,-inf,inf
-user_7,default,entity_18,0.43807952480223084,-inf,inf
-user_7,default,entity_2,1.1942940555235904,-inf,inf
-user_7,default,entity_4,-0.4542680865093558,-inf,inf
-user_7,default,entity_14,2.5219389650833213,-inf,inf
-user_7,default,entity_19,-0.6650799718202678,-inf,inf
-user_7,default,entity_9,-0.13580933330195982,-inf,inf
-user_7,default,entity_11,-0.4812107308028669,-inf,inf
-user_7,default,entity_17,1.96090138537457,-inf,inf
-user_8,default,entity_8,2.1127574448260513,-inf,inf
-user_8,default,entity_4,2.642406825318763,-inf,inf
-user_8,default,entity_9,0.5444086059726662,-inf,inf
-user_8,default,entity_11,0.5846999636827297,-inf,inf
-user_8,default,entity_2,0.7698291989015807,-inf,inf
-user_8,default,entity_18,1.3419798902453495,-inf,inf
-user_8,default,entity_5,1.7797980695330238,-inf,inf
-user_9,default,entity_3,-0.01253707482118327,-inf,inf
-user_9,default,entity_13,1.3216410202684816,-inf,inf
-user_9,default,entity_11,1.694370071188104,-inf,inf
-user_9,default,entity_14,1.5236251305880073,-inf,inf
-user_9,default,entity_9,2.4498918215182184,-inf,inf
+user_0,default,entity_0,-3.0840321464854243,-inf,inf
+user_0,default,entity_16,-4.0878539983787,-inf,inf
+user_0,default,entity_15,-1.4902020610943079,-inf,inf
+user_0,default,entity_18,-1.0038588539696902,-inf,inf
+user_0,default,entity_9,0.15862347293943008,-inf,inf
+user_0,default,entity_13,0.6825297084581011,-inf,inf
+user_0,default,entity_19,-1.1330286216672034,-inf,inf
+user_0,default,entity_5,1.527715830950634,-inf,inf
+user_0,default,entity_4,3.4585969415576345,-inf,inf
+user_1,default,entity_3,1.667515895613929,-inf,inf
+user_1,default,entity_14,1.006242789479527,-inf,inf
+user_1,default,entity_7,2.664075720463755,-inf,inf
+user_1,default,entity_18,-0.8301415440372613,-inf,inf
+user_1,default,entity_8,-0.17420902398789284,-inf,inf
+user_1,default,entity_10,-1.1835582791092611,-inf,inf
+user_1,default,entity_6,-1.1948480725179846,-inf,inf
+user_1,default,entity_16,-0.9539951632086848,-inf,inf
+user_1,default,entity_12,-2.180286196407139,-inf,inf
+user_1,default,entity_2,-0.42652666498132275,-inf,inf
+user_2,default,entity_11,-1.387325030216633,-inf,inf
+user_2,default,entity_9,-0.9284619449437295,-inf,inf
+user_2,default,entity_18,-0.4850896006442237,-inf,inf
+user_2,default,entity_14,-0.5877998959704127,-inf,inf
+user_2,default,entity_17,-0.8219599742190198,-inf,inf
+user_2,default,entity_0,-0.7750717258171586,-inf,inf
+user_2,default,entity_4,-0.7272800370864463,-inf,inf
+user_2,default,entity_13,-0.33567087330509393,-inf,inf
+user_2,default,entity_5,1.0737649403543752,-inf,inf
+user_2,default,entity_1,1.3235236743850214,-inf,inf
+user_2,default,entity_16,2.548852058606136,-inf,inf
+user_3,default,entity_19,0.945310811705114,-inf,inf
+user_3,default,entity_1,2.280503660781391,-inf,inf
+user_3,default,entity_13,-1.5769377112488974,-inf,inf
+user_3,default,entity_16,1.5165254267520498,-inf,inf
+user_3,default,entity_10,2.072081238298041,-inf,inf
+user_3,default,entity_3,-1.1848498093087512,-inf,inf
+user_3,default,entity_7,1.3271180042064845,-inf,inf
+user_3,default,entity_18,-0.25013042087067283,-inf,inf
+user_3,default,entity_12,-1.0216656841229712,-inf,inf
+user_3,default,entity_6,-1.0810028748325473,-inf,inf
+user_3,default,entity_2,1.203173894386199,-inf,inf
+user_4,default,entity_19,0.2639409528014471,-inf,inf
+user_4,default,entity_8,0.440512579577302,-inf,inf
+user_4,default,entity_2,-0.43449183897222454,-inf,inf
+user_4,default,entity_17,0.6740247220648662,-inf,inf
+user_4,default,entity_18,1.2408002798143847,-inf,inf
+user_4,default,entity_3,0.48288392306535105,-inf,inf
+user_4,default,entity_15,0.8328232652751022,-inf,inf
+user_4,default,entity_12,-0.49299657597255997,-inf,inf
+user_4,default,entity_14,-0.5966527881357853,-inf,inf
+user_4,default,entity_7,0.11951141508835739,-inf,inf
+user_4,default,entity_16,-0.5090007368392502,-inf,inf
+user_5,default,entity_6,-0.21462643170061715,-inf,inf
+user_5,default,entity_17,0.05710431263487674,-inf,inf
+user_6,default,entity_13,-0.6865530798790694,-inf,inf
+user_6,default,entity_16,0.27755782227896963,-inf,inf
+user_6,default,entity_4,-0.6587271432487041,-inf,inf
+user_6,default,entity_8,-0.5583548331157635,-inf,inf
+user_6,default,entity_18,1.9506928061461883,-inf,inf
+user_6,default,entity_1,0.49050679907782496,-inf,inf
+user_6,default,entity_2,-1.3252639653101213,-inf,inf
+user_6,default,entity_0,-0.3233871818836944,-inf,inf
+user_7,default,entity_4,0.18732339474126747,-inf,inf
+user_7,default,entity_8,1.9008743051155528,-inf,inf
+user_7,default,entity_19,0.0966132005912641,-inf,inf
+user_7,default,entity_3,0.9342862554614455,-inf,inf
+user_7,default,entity_11,1.6668385201561917,-inf,inf
+user_7,default,entity_5,0.7660559749152163,-inf,inf
+user_7,default,entity_10,-1.394815185465011,-inf,inf
+user_7,default,entity_2,-0.5664965551290375,-inf,inf
+user_7,default,entity_14,3.0454263676652222,-inf,inf
+user_8,default,entity_8,2.4404515835847844,-inf,inf
+user_8,default,entity_4,2.1409310166269293,-inf,inf
+user_8,default,entity_11,0.8428911156251948,-inf,inf
+user_8,default,entity_14,0.11540882613427506,-inf,inf
+user_8,default,entity_6,0.45813136118948833,-inf,inf
+user_8,default,entity_2,0.31277617960683546,-inf,inf
+user_9,default,entity_3,-1.962426062799881,-inf,inf
+user_9,default,entity_12,-0.6855884162192276,-inf,inf
+user_9,default,entity_2,-0.885308032694007,-inf,inf
+user_9,default,entity_14,0.5609584765999138,-inf,inf
+user_9,default,entity_9,-0.023301797645628702,-inf,inf
+user_9,default,entity_5,-0.7796727684945641,-inf,inf
+user_9,default,entity_1,-2.172493689682739,-inf,inf