Loss function - Githubissues

adolfocorreia / DGM

Deep Galerkin Method

16 stars 5 forks source link

Loss function #1

Closed Plemeur closed 5 years ago

Plemeur commented 5 years ago

Hi, according to this paper :https://arxiv.org/pdf/1811.08782.pdf the loss function in line 55 of merton.py should be :

f = -0.5*lambd*2V_x2 + (V_t + rx1 V* )V_xx

Have you been using V_x instead of V for a reason ?

adolfocorreia commented 5 years ago

Unfortunately, the equation in the paper (page 57) is incorrect. There should be no $-\frac{\lambda^2}{2 \sigma^2}$ term inside the parentheses and the other term should be $r x \partial_xH$ . The code in merton.py is correct.