Add basic benchmarks

Author: Frankie Robertson

benchmark/benchmarks.jl

@@ -0,0 +1,69 @@
+using BenchmarkTools
+using AirspeedVelocity
+using ComputerAdaptiveTesting
+using Random: Xoshiro, sample
+using StatsBase: sample
+using FittedItemBanks
+using FittedItemBanks.DummyData: dummy_full, SimpleItemBankSpec, StdModel4PL
+using ComputerAdaptiveTesting.Aggregators
+using PsychometricsBazaarBase.Optimizers
+using PsychometricsBazaarBase.Integrators: even_grid
+using ComputerAdaptiveTesting.NextItemRules: mirtcat_quadpts
+using ComputerAdaptiveTesting.NextItemRules: DistributionExpectationBasedItemCriterion,
+                                             PointExpectationBasedItemCriterion
+using ComputerAdaptiveTesting.NextItemRules
+using ComputerAdaptiveTesting.Responses
+
+const SUITE = BenchmarkGroup()
+
+SUITE["next_item_rules"] = BenchmarkGroup()
+
+#prepare_0(ability_estimator)
+function prepare_4pls(group)
+    rng = Xoshiro(42)
+    (; item_bank, abilities, responses) = dummy_full(
+        rng,
+        SimpleItemBankSpec(StdModel4PL(), OneDimContinuousDomain(), BooleanResponse());
+        num_questions = 20,
+        num_testees = 1
+    )
+    integrator = even_grid(-6.0, 6.0, mirtcat_quadpts(1))
+    optimizer = AbilityOptimizer(OneDimOptimOptimizer(-6.0, 6.0, NelderMead()))
+
+    dist_ability_estimator = PriorAbilityEstimator()
+    ability_estimators = [
+        ("mean", MeanAbilityEstimator(dist_ability_estimator, integrator)),
+        ("mode", ModeAbilityEstimator(dist_ability_estimator, optimizer))
+    ]
+    response_idxs = sample(rng, 1:20, 10)
+
+    for (est_nick, ability_estimator) in ability_estimators
+        next_item_rule = ItemStrategyNextItemRule(
+            ExhaustiveSearch1Ply(),
+            PointExpectationBasedItemCriterion(ability_estimator,
+                AbilityVarianceStateCriterion(
+                    integrator, distribution_estimator(ability_estimator)))
+        )
+        next_item_rule = preallocate(next_item_rule)
+        tracked_responses = TrackedResponses(BareResponses(ResponseType(item_bank)),
+            item_bank,
+            NullAbilityTracker())
+        group["$(est_nick)_point_mepv_bare"] = @benchmark ($next_item_rule)(
+            $tracked_responses, $item_bank)
+        bare_responses = BareResponses(
+            ResponseType(item_bank),
+            response_idxs,
+            # XXX: Not sure why this is needed (seems to not be needed else)
+            collect(responses[response_idxs, 1])
+        )
+        tracked_responses = TrackedResponses(
+            bare_responses,
+            item_bank,
+            NullAbilityTracker())
+        group["$(est_nick)_point_mepv_10"] = @benchmark ($next_item_rule)(
+            $tracked_responses, $item_bank)
+    end
+    return group
+end
+
+SUITE["next_item_rules"]["4pl"] = prepare_4pls(BenchmarkGroup())