Table 3 Sampling range for the Heston parameters. LHS means Latin Hypercube Sampling, COS stands for the COS method (see the Appendix) and Brent for the root-finding iteration
From: A neural network-based framework for financial model calibration
ANN | Parameters | Value range | Generating method |
|---|---|---|---|
ANN Input | Moneyness, \(m=S_{0}/K\) | [0.6,1.4] | LHS |
Time to maturity, τ | [0.05,3.0] (year) | LHS | |
Risk free rate, r | [0.0%,5%] | LHS | |
Correlation, ρ | [−0.90,0.0] | LHS | |
Reversion speed, κ | (0,3.0] | LHS | |
Volatility of volatility, γ | (0.01,0.8] | LHS | |
Long average variance, ν̄ | (0.01,0.5] | LHS | |
Initial variance, \(\nu _{0}\) | (0.05,0.5] | LHS | |
– | European put price, V | (0,0.6) | COS |
ANN Output | Black–Scholes IV, σ | (0,0.76) | Brent |