Correcting sampling of log_price for heston

Author: aleCombi

src/distributions/heston.jl

@@ -110,6 +110,13 @@ struct LogHestonDistribution{TS0,TV0,Tκ,Tθ,Tσ,Tρ,Tr,T} <: ContinuousMultivar
     T::T
 end
 
+
+import Distributions: length
+
+function Base.length(::LogHestonDistribution)
+    return 2 # The Heston model is a 2-dimensional process (Asset Price, Variance)
+end
+
 """
     sample_V_T(rng, d::LogHestonDistribution)
 
@@ -251,6 +258,23 @@ function rand(rng::AbstractRNG, d::LogHestonDistribution; antithetic = false, kw
     return [log_S_T, V_T]
 end
 
+function Distributions.rand!(
+    rng::AbstractRNG,
+    d::LogHestonDistribution,
+    x::AbstractVector{<:Real};
+    antithetic = false,
+    kwargs...
+)
+    # Call your allocating rand function
+    sample = rand(rng, d; antithetic=antithetic, kwargs...)
+    
+    # Copy the result into the pre-allocated vector x
+    x[1] = sample[1]
+    x[2] = sample[2]
+    
+    return x
+end
+
 function sample_log_S_T(
     V_T,
     integral_V,

src/pricing_methods/montecarlo.jl

@@ -401,16 +401,15 @@ end
 
 Draws a specified number of samples from a given probability distribution (`law`).
 This is primarily used by the `ExactSimulation` strategy.
+It assumes that the first component of the multi-dimensional law correspond with the log-price.
 """
 function log_sample(rng, law::ContinuousUnivariateDistribution, trajectories)
-    log_sample = Distributions.rand(rng, law, trajectories)
-    return log_sample
+    return Distributions.rand(rng, law, trajectories)
 end
 
 function log_sample(rng, law::ContinuousMultivariateDistribution, trajectories)
-    print((rng, law, trajectories))
-    log_sample, _ = Distributions.rand(rng, law, trajectories)
-    return log_sample
+    X = Distributions.rand(rng, law, trajectories)  # shape: (n, trajectories)
+    return view(X, 1, :)  # returns a lightweight view of the first row (log-price)
 end
 
 
@@ -479,9 +478,10 @@ function solve(
 
     # Get samples using the dispatched helper
     sample_at_expiry = get_final_samples(prob, method)
-
+    @show typeof(sample_at_expiry)
     # Common logic for pricing
     payoffs = reduce_payoffs(sample_at_expiry, prob.payoff, config.variance_reduction)
+    @show typeof(payoffs)
     discount = df(prob.market_inputs.rate, prob.payoff.expiry)
     price = discount * mean(payoffs)