Mathematics & Computation

NTK만kL2 errorH1 error1 1.718e-031.694e-0221.563e-031.503e-0234.942e-035.287e-024 5.266e-011.986e+005 3.085e-01 1.747e+0061.811e-011.171e+00 71.137e-01 8.212e-0188.470e-02 6.915e-01 9 6.481e-025.693e-01104.688e-024.376e-01 NTK + curriculum kL2 errorH1 error11.712e-03 1.695e-022 6.404e-047.420e-033 5.349e-047.210e-0342.361e-031.132e-0252.143e-031.396e-026 1.219e-03 1.048e-0271.748e-03 1.513e-028..

model = PINN([2, 50, 50, 50, 1]).to(device)step = 20a1, a2 = 1,1 LAMBDA_BC = 1.0LAMBDA_PDE = 1.0LAMBDA_IC = 5.0 # anchor를 쓰는 경우 # 손실 총 스케일(항상 고정): 예시 60.0SUM_W = 60.0 1.K_SCHEDULE = [ (0, 1.0), (5000, 3.0), (10000, 6.0), (15000, 10.0),] L2 abs: 1.566e-02 | L2 rel: 1.566e-02 H1 abs: 1.632e-01 | H1 rel: 3.531e-02 L_inf abs:2.810e-02 2.K_SCHEDULE = [ (0, 1.0), (10000, 10.0),] L2 abs: 2.551e-02 |..

epoch = 20000model = PINN([3, 50, 50, 50,50, 1]).to(device)a1, a2 = 1, 1 k=1k=2k=3k=4k=5k=6step = 10Relative L2 error4.8791e-026.7382e-02 1.0356e-01 9.9476e-01 9.9869e-01 4.8525e-01Relative H1 error 5.2810e-024.2188e-029.2454e-029.9470e-01 9.9657e-01 5.2326e-01Time 2m2m2m3m3m3mstep = 20Relative L2 error1.0876e-01 4.5514e-02 2.4261e-01 9.0612e-01 6.0647e-01Relative H1 error 1.1269e-014.0538e..

#Try1: Max error: 2.1042e+00 Relative Error: 1.4976e+00 #Try2: #Try3:L-BGFS추가파일