vortex_equation/fixed_alhpa_no_lasso.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import sympy\n",
    "import numpy as np\n",
    "import torch\n",
    "torch.manual_seed(12)\n",
    "torch.cuda.manual_seed(12)\n",
    "torch.cuda.manual_seed_all\n",
    "\n",
    "dtype = torch.float\n",
    "device = torch.device(\"cuda:0\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "sys.path.append('/u/wangnisn/devel/machine_learning_kinetics/ml_kinetics')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from mlk import learning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Set up model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "vortex = {'eqn_type':'vortex',\n",
    "        'fcn':'exp(-1/2*(x-cos(t))**2-1/2*(y-sin(t))**2)',\n",
    "        'domain':{'x':[-2,2],'y':[-2,2],'t':[0,2]},\n",
    "        \n",
    "        'err_vec':[0,0,0,1,-1,0,0]}\n",
    "\n",
    "u_t,u_x,u_y,u,x,y = sympy.symbols('u_t u_x u_y u x y')\n",
    "\n",
    "dictionary = (u_x,u_y,x*u_x,y*u_x,x*u_y,y*u_y,u)\n",
    "vortex['dictionary'] = dictionary\n",
    "pde = vortex\n",
    "domain =pde['domain']\n",
    "fcn = pde['fcn']\n",
    "n_points = 50000\n",
    "x = sympy.symbols([x for x in domain.keys()])\n",
    "u_exact = sympy.lambdify(x,sympy.sympify(fcn),'numpy')\n",
    "inputs = {}\n",
    "for key in domain.keys():\n",
    "    min_d = domain[key][0]\n",
    "    max_d = domain[key][1]\n",
    "    #iputs[key] = ((max_d-min_d)*torch.rand((n_points,1),device=device,dtype=dtype,requires_grad=True)+min_d)\n",
    "    tt = np.random.uniform(min_d,max_d,(n_points,1))\n",
    "    inputs[key] = torch.tensor(tt,device=device,dtype=dtype,requires_grad=True)\n",
    "u_exact = sympy.lambdify(x,sympy.sympify(fcn),'numpy')\n",
    "u = u_exact(*[i.detach().cpu() for i in inputs.values()])\n",
    "pde = {}\n",
    "u = u.type(dtype).to(device)\n",
    "pde['inputs'] = inputs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch:    0,  Loss u: 1.376e-01, Loss pde: 9.807e-04, Loss_norm: 7.488e+00, Loss tot: 1.376e-01\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:   50,  Loss u: 1.573e-02, Loss pde: 3.447e-01, Loss_norm: 7.488e+00, Loss tot: 1.573e-02\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:  100,  Loss u: 2.357e-03, Loss pde: 9.132e-01, Loss_norm: 7.488e+00, Loss tot: 2.357e-03\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:  150,  Loss u: 1.290e-03, Loss pde: 9.376e-01, Loss_norm: 7.488e+00, Loss tot: 1.290e-03\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:  200,  Loss u: 1.007e-03, Loss pde: 9.301e-01, Loss_norm: 7.488e+00, Loss tot: 1.007e-03\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:  250,  Loss u: 6.218e-04, Loss pde: 9.504e-01, Loss_norm: 7.488e+00, Loss tot: 6.218e-04\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:  300,  Loss u: 4.770e-04, Loss pde: 9.501e-01, Loss_norm: 7.488e+00, Loss tot: 4.770e-04\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:  350,  Loss u: 8.324e-04, Loss pde: 9.547e-01, Loss_norm: 7.488e+00, Loss tot: 8.324e-04\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:  400,  Loss u: 2.854e-04, Loss pde: 9.511e-01, Loss_norm: 7.488e+00, Loss tot: 2.854e-04\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:  450,  Loss u: 1.209e-03, Loss pde: 9.353e-01, Loss_norm: 7.488e+00, Loss tot: 1.209e-03\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:  500,  Loss u: 2.073e-04, Loss pde: 9.531e-01, Loss_norm: 7.488e+00, Loss tot: 9.533e-01\n",
      "  coefs: -5.6304e-01   -2.8970e+00    1.4370e-01   -6.2471e-01   -1.4866e+00    1.3500e+00    4.2288e-01  \n",
      "Epoch:  550,  Loss u: 3.918e-02, Loss pde: 7.921e-03, Loss_norm: 6.605e+00, Loss tot: 4.717e-02\n",
      "  coefs: -3.3730e-01   -2.7277e+00    1.3197e-01   -4.5002e-01   -1.3331e+00    1.1569e+00    4.6847e-01  \n",
      "Epoch:  600,  Loss u: 2.848e-02, Loss pde: 5.568e-03, Loss_norm: 6.046e+00, Loss tot: 3.411e-02\n",
      "  coefs: -1.9465e-01   -2.6554e+00   -1.5429e-01   -3.5305e-01   -1.3374e+00    8.0776e-01    5.4395e-01  \n",
      "Epoch:  650,  Loss u: 2.115e-02, Loss pde: 4.153e-03, Loss_norm: 5.465e+00, Loss tot: 2.536e-02\n",
      "  coefs: -8.6139e-02   -2.5589e+00   -3.7214e-01   -2.1121e-01   -1.3218e+00    3.7058e-01    5.4446e-01  \n",
      "Epoch:  700,  Loss u: 1.784e-02, Loss pde: 2.668e-03, Loss_norm: 4.964e+00, Loss tot: 2.056e-02\n",
      "  coefs: -4.5261e-02   -2.4372e+00   -4.3785e-01   -1.5265e-01   -1.2789e+00    2.4549e-02    5.8765e-01  \n",
      "Epoch:  750,  Loss u: 1.591e-02, Loss pde: 2.337e-03, Loss_norm: 5.007e+00, Loss tot: 1.829e-02\n",
      "  coefs: -1.1953e-02   -2.3061e+00   -4.4937e-01   -1.5075e-01   -1.2184e+00   -3.0107e-01    5.6920e-01  \n",
      "Epoch:  800,  Loss u: 1.423e-02, Loss pde: 1.553e-03, Loss_norm: 5.082e+00, Loss tot: 1.583e-02\n",
      "  coefs:  3.9444e-02   -2.1627e+00   -4.3898e-01   -1.4902e-01   -1.1555e+00   -5.6041e-01    5.7609e-01  \n",
      "Epoch:  850,  Loss u: 1.280e-02, Loss pde: 1.306e-03, Loss_norm: 5.022e+00, Loss tot: 1.416e-02\n",
      "  coefs:  8.8915e-02   -2.0239e+00   -4.0521e-01   -1.2309e-01   -1.0835e+00   -7.5334e-01    5.4428e-01  \n",
      "Epoch:  900,  Loss u: 1.158e-02, Loss pde: 1.255e-03, Loss_norm: 4.809e+00, Loss tot: 1.289e-02\n",
      "  coefs:  1.3550e-01   -1.8878e+00   -3.6168e-01   -7.0599e-02   -1.0090e+00   -8.5402e-01    4.9053e-01  \n",
      "Epoch:  950,  Loss u: 1.052e-02, Loss pde: 1.220e-03, Loss_norm: 4.477e+00, Loss tot: 1.179e-02\n",
      "  coefs:  1.9224e-01   -1.7405e+00   -3.2813e-01   -1.7805e-02   -9.2578e-01   -8.6016e-01    4.1200e-01  \n",
      "Epoch: 1000,  Loss u: 9.305e-03, Loss pde: 1.170e-03, Loss_norm: 4.146e+00, Loss tot: 1.052e-02\n",
      "  coefs:  2.4179e-01   -1.5682e+00   -2.6833e-01    3.3304e-02   -8.5186e-01   -8.3593e-01    3.4688e-01  \n",
      "Epoch: 1050,  Loss u: 7.949e-03, Loss pde: 1.099e-03, Loss_norm: 3.754e+00, Loss tot: 9.086e-03\n",
      "  coefs:  2.9448e-01   -1.3675e+00   -2.0533e-01    8.3891e-02   -7.7433e-01   -7.6314e-01    2.6554e-01  \n",
      "Epoch: 1100,  Loss u: 6.291e-03, Loss pde: 9.721e-04, Loss_norm: 3.258e+00, Loss tot: 7.295e-03\n",
      "  coefs:  3.5002e-01   -1.1288e+00   -1.4001e-01    1.3914e-01   -6.8429e-01   -6.4303e-01    1.7250e-01  \n",
      "Epoch: 1150,  Loss u: 4.210e-03, Loss pde: 7.239e-04, Loss_norm: 2.646e+00, Loss tot: 4.960e-03\n",
      "  coefs:  4.0986e-01   -8.5189e-01   -7.6077e-02    2.0283e-01   -5.7285e-01   -4.6005e-01    7.2645e-02  \n",
      "Epoch: 1200,  Loss u: 2.261e-03, Loss pde: 4.014e-04, Loss_norm: 2.039e+00, Loss tot: 2.682e-03\n",
      "  coefs:  4.7993e-01   -5.8813e-01   -3.4395e-02    2.5800e-01   -4.4105e-01   -2.3433e-01    2.8807e-03  \n",
      "Epoch: 1250,  Loss u: 9.882e-04, Loss pde: 2.623e-04, Loss_norm: 1.759e+00, Loss tot: 1.268e-03\n",
      "  coefs:  5.0062e-01   -4.1829e-01   -1.7337e-02    3.1080e-01   -3.7602e-01   -1.1063e-01   -2.5410e-02  \n",
      "Epoch: 1300,  Loss u: 7.247e-04, Loss pde: 2.131e-04, Loss_norm: 1.682e+00, Loss tot: 9.546e-04\n",
      "  coefs:  5.0589e-01   -3.5794e-01   -1.6256e-02    3.3634e-01   -3.7537e-01   -6.6322e-02   -2.3629e-02  \n",
      "Epoch: 1350,  Loss u: 6.205e-04, Loss pde: 1.818e-04, Loss_norm: 1.667e+00, Loss tot: 8.190e-04\n",
      "  coefs:  4.9902e-01   -3.3706e-01   -1.4220e-02    3.5769e-01   -3.9567e-01   -4.5205e-02   -1.8312e-02  \n",
      "Epoch: 1400,  Loss u: 5.468e-04, Loss pde: 1.576e-04, Loss_norm: 1.668e+00, Loss tot: 7.210e-04\n",
      "  coefs:  4.8600e-01   -3.2214e-01   -1.1538e-02    3.8246e-01   -4.1992e-01   -3.2129e-02   -1.3780e-02  \n",
      "Epoch: 1450,  Loss u: 5.088e-04, Loss pde: 1.531e-04, Loss_norm: 1.662e+00, Loss tot: 6.785e-04\n",
      "  coefs:  4.7993e-01   -3.1178e-01   -4.3835e-03    3.9463e-01   -4.3246e-01   -2.7545e-02   -1.1536e-02  \n",
      "Epoch: 1500,  Loss u: 4.511e-04, Loss pde: 1.308e-04, Loss_norm: 1.672e+00, Loss tot: 5.986e-04\n",
      "  coefs:  4.5882e-01   -2.9881e-01   -5.2832e-03    4.2566e-01   -4.5901e-01   -1.7619e-02   -6.7956e-03  \n",
      "Epoch: 1550,  Loss u: 4.033e-04, Loss pde: 1.237e-04, Loss_norm: 1.683e+00, Loss tot: 5.439e-04\n",
      "  coefs:  4.4094e-01   -2.8383e-01   -4.8343e-03    4.5309e-01   -4.8436e-01   -1.1516e-02   -4.8789e-03  \n",
      "Epoch: 1600,  Loss u: 3.626e-04, Loss pde: 1.195e-04, Loss_norm: 1.694e+00, Loss tot: 4.991e-04\n",
      "  coefs:  4.2306e-01   -2.7001e-01   -3.8054e-03    4.7890e-01   -5.0813e-01   -6.9208e-03   -3.0241e-03  \n",
      "Epoch: 1650,  Loss u: 3.274e-04, Loss pde: 1.170e-04, Loss_norm: 1.704e+00, Loss tot: 4.615e-04\n",
      "  coefs:  4.0546e-01   -2.5716e-01   -2.9039e-03    5.0322e-01   -5.3043e-01   -3.6701e-03   -1.4997e-03  \n",
      "Epoch: 1700,  Loss u: 4.000e-04, Loss pde: 1.372e-04, Loss_norm: 1.692e+00, Loss tot: 5.541e-04\n",
      "  coefs:  4.1113e-01   -2.5176e-01    5.2342e-04    4.9045e-01   -5.2278e-01   -1.1184e-02   -4.5778e-03  \n",
      "Epoch: 1750,  Loss u: 2.993e-04, Loss pde: 1.103e-04, Loss_norm: 1.712e+00, Loss tot: 4.267e-04\n",
      "  coefs:  3.8882e-01   -2.4641e-01   -8.5759e-04    5.2348e-01   -5.4694e-01   -4.7953e-03   -9.0372e-04  \n",
      "Epoch: 1800,  Loss u: 2.691e-04, Loss pde: 1.092e-04, Loss_norm: 1.726e+00, Loss tot: 3.955e-04\n",
      "  coefs:  3.7041e-01   -2.3306e-01   -1.3325e-03    5.4847e-01   -5.6997e-01   -2.6311e-03   -5.6509e-04  \n",
      "Epoch: 1850,  Loss u: 2.440e-04, Loss pde: 1.087e-04, Loss_norm: 1.739e+00, Loss tot: 3.701e-04\n",
      "  coefs:  3.5413e-01   -2.2193e-01   -1.4305e-03    5.7003e-01   -5.9038e-01   -1.2175e-03   -1.9011e-04  \n",
      "Epoch: 1900,  Loss u: 2.222e-04, Loss pde: 1.076e-04, Loss_norm: 1.751e+00, Loss tot: 3.473e-04\n",
      "  coefs:  3.3896e-01   -2.1198e-01   -1.4820e-03    5.8961e-01   -6.0891e-01   -4.2647e-04    1.3189e-06  \n",
      "Epoch: 1950,  Loss u: 2.028e-04, Loss pde: 1.060e-04, Loss_norm: 1.763e+00, Loss tot: 3.264e-04\n",
      "  coefs:  3.2471e-01   -2.0274e-01   -1.6697e-03    6.0779e-01   -6.2613e-01   -6.9125e-07    5.9273e-06  \n",
      "Epoch: 2000,  Loss u: 3.076e-04, Loss pde: 1.203e-04, Loss_norm: 1.773e+00, Loss tot: 4.456e-04\n",
      "  coefs:  3.2658e-01   -1.9416e-01    3.1446e-03    6.0777e-01   -6.3091e-01   -8.2734e-03    2.5354e-03  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 2050,  Loss u: 1.855e-04, Loss pde: 1.008e-04, Loss_norm: 1.773e+00, Loss tot: 3.041e-04\n",
      "  coefs:  3.1069e-01   -1.9500e-01   -6.2745e-04    6.2324e-01   -6.4102e-01   -2.0839e-03   -2.2181e-04  \n",
      "Epoch: 2100,  Loss u: 1.676e-04, Loss pde: 9.905e-05, Loss_norm: 1.785e+00, Loss tot: 2.845e-04\n",
      "  coefs:  2.9553e-01   -1.8463e-01   -1.8897e-03    6.4367e-01   -6.5899e-01   -1.8841e-04   -2.9216e-04  \n",
      "Epoch: 2150,  Loss u: 1.527e-04, Loss pde: 9.753e-05, Loss_norm: 1.796e+00, Loss tot: 2.681e-04\n",
      "  coefs:  2.8239e-01   -1.7609e-01   -1.8562e-03    6.6037e-01   -6.7495e-01    2.6040e-06   -2.0849e-04  \n",
      "Epoch: 2200,  Loss u: 1.395e-04, Loss pde: 9.563e-05, Loss_norm: 1.806e+00, Loss tot: 2.532e-04\n",
      "  coefs:  2.7022e-01   -1.6835e-01   -1.8442e-03    6.7554e-01   -6.8949e-01    1.4876e-04   -1.5638e-04  \n",
      "Epoch: 2250,  Loss u: 1.305e-03, Loss pde: 5.523e-04, Loss_norm: 1.816e+00, Loss tot: 1.876e-03\n",
      "  coefs:  2.5882e-01   -1.6169e-01   -1.7996e-03    6.8921e-01   -7.0321e-01    7.5164e-04   -3.1377e-04  \n",
      "Epoch: 2300,  Loss u: 1.751e-04, Loss pde: 9.281e-05, Loss_norm: 1.801e+00, Loss tot: 2.860e-04\n",
      "  coefs:  2.8008e-01   -1.7360e-01   -2.0726e-05    6.6164e-01   -6.7675e-01   -6.1350e-03   -2.7213e-03  \n",
      "Epoch: 2350,  Loss u: 1.319e-04, Loss pde: 8.698e-05, Loss_norm: 1.813e+00, Loss tot: 2.370e-04\n",
      "  coefs:  2.6060e-01   -1.6267e-01   -2.0460e-03    6.8647e-01   -6.9776e-01   -2.0115e-03   -1.7215e-03  \n",
      "Epoch: 2400,  Loss u: 1.177e-04, Loss pde: 8.621e-05, Loss_norm: 1.824e+00, Loss tot: 2.221e-04\n",
      "  coefs:  2.4656e-01   -1.5353e-01   -2.0303e-03    7.0483e-01   -7.1547e-01   -4.7996e-04   -1.0910e-03  \n",
      "Epoch: 2450,  Loss u: 1.066e-04, Loss pde: 8.512e-05, Loss_norm: 1.833e+00, Loss tot: 2.101e-04\n",
      "  coefs:  2.3452e-01   -1.4609e-01   -1.9064e-03    7.1984e-01   -7.3004e-01   -2.1278e-05   -7.2835e-04  \n",
      "Epoch: 2500,  Loss u: 9.732e-05, Loss pde: 8.377e-05, Loss_norm: 1.841e+00, Loss tot: 1.995e-04\n",
      "  coefs:  2.2371e-01   -1.3934e-01   -1.8497e-03    7.3298e-01   -7.4291e-01    2.4051e-05   -6.0552e-04  \n",
      "Epoch: 2550,  Loss u: 8.920e-05, Loss pde: 8.229e-05, Loss_norm: 1.849e+00, Loss tot: 1.900e-04\n",
      "  coefs:  2.1373e-01   -1.3319e-01   -1.8101e-03    7.4499e-01   -7.5462e-01    1.6861e-04   -5.0913e-04  \n",
      "Epoch: 2600,  Loss u: 8.217e-05, Loss pde: 8.075e-05, Loss_norm: 1.856e+00, Loss tot: 1.815e-04\n",
      "  coefs:  2.0437e-01   -1.2744e-01   -1.7323e-03    7.5612e-01   -7.6547e-01    2.5888e-04   -4.2276e-04  \n",
      "Epoch: 2650,  Loss u: 9.901e-05, Loss pde: 7.651e-05, Loss_norm: 1.849e+00, Loss tot: 1.940e-04\n",
      "  coefs:  2.1644e-01   -1.3355e-01    4.0056e-06    7.4037e-01   -7.5318e-01   -4.0433e-03   -1.3696e-03  \n",
      "Epoch: 2700,  Loss u: 8.081e-05, Loss pde: 7.625e-05, Loss_norm: 1.856e+00, Loss tot: 1.756e-04\n",
      "  coefs:  2.0167e-01   -1.2628e-01   -1.9572e-03    7.5861e-01   -7.6691e-01    1.0526e-04   -7.7667e-04  \n",
      "Epoch: 2750,  Loss u: 7.296e-05, Loss pde: 7.538e-05, Loss_norm: 1.865e+00, Loss tot: 1.670e-04\n",
      "  coefs:  1.9087e-01   -1.1938e-01   -1.6811e-03    7.7227e-01   -7.8013e-01    1.6783e-04   -5.2121e-04  \n",
      "Epoch: 2800,  Loss u: 6.682e-05, Loss pde: 7.420e-05, Loss_norm: 1.872e+00, Loss tot: 1.597e-04\n",
      "  coefs:  1.8181e-01   -1.1374e-01   -1.4540e-03    7.8325e-01   -7.9084e-01    3.4803e-04   -3.1055e-04  \n",
      "Epoch: 2850,  Loss u: 6.163e-05, Loss pde: 7.278e-05, Loss_norm: 1.877e+00, Loss tot: 1.532e-04\n",
      "  coefs:  1.7374e-01   -1.0873e-01   -1.2906e-03    7.9286e-01   -8.0025e-01    4.0877e-04   -1.9422e-04  \n",
      "Epoch: 2900,  Loss u: 1.744e-04, Loss pde: 1.554e-04, Loss_norm: 1.868e+00, Loss tot: 3.485e-04\n",
      "  coefs:  1.9916e-01   -1.0894e-01    7.5318e-04    7.5988e-01   -7.8456e-01   -1.2516e-02   -2.4371e-03  \n",
      "Epoch: 2950,  Loss u: 7.361e-05, Loss pde: 6.848e-05, Loss_norm: 1.870e+00, Loss tot: 1.608e-04\n",
      "  coefs:  1.8540e-01   -1.1720e-01   -1.9030e-03    7.7736e-01   -7.8457e-01   -1.3643e-03   -1.8831e-03  \n",
      "Epoch: 3000,  Loss u: 6.222e-05, Loss pde: 6.663e-05, Loss_norm: 1.879e+00, Loss tot: 1.476e-04\n",
      "  coefs:  1.7254e-01   -1.0801e-01   -1.5811e-03    7.9479e-01   -8.0075e-01    2.5656e-06   -9.2817e-04  \n",
      "Epoch: 3050,  Loss u: 5.622e-05, Loss pde: 6.557e-05, Loss_norm: 1.885e+00, Loss tot: 1.406e-04\n",
      "  coefs:  1.6286e-01   -1.0191e-01   -1.1424e-03    8.0656e-01   -8.1239e-01    8.4616e-05   -5.3896e-04  \n",
      "Epoch: 3100,  Loss u: 5.155e-05, Loss pde: 6.437e-05, Loss_norm: 1.891e+00, Loss tot: 1.348e-04\n",
      "  coefs:  1.5489e-01   -9.6960e-02   -9.7002e-04    8.1604e-01   -8.2183e-01    2.6766e-04   -3.1745e-04  \n",
      "Epoch: 3150,  Loss u: 4.766e-05, Loss pde: 6.299e-05, Loss_norm: 1.896e+00, Loss tot: 1.296e-04\n",
      "  coefs:  1.4786e-01   -9.2602e-02   -7.8579e-04    8.2426e-01   -8.3001e-01    3.5704e-04   -1.5729e-04  \n",
      "Epoch: 3200,  Loss u: 4.427e-05, Loss pde: 6.150e-05, Loss_norm: 1.900e+00, Loss tot: 1.248e-04\n",
      "  coefs:  1.4144e-01   -8.8623e-02   -5.8609e-04    8.3171e-01   -8.3739e-01    3.8581e-04   -3.3575e-05  \n",
      "Epoch: 3250,  Loss u: 1.276e-04, Loss pde: 8.599e-05, Loss_norm: 1.888e+00, Loss tot: 2.325e-04\n",
      "  coefs:  1.5666e-01   -9.2769e-02    2.4792e-03    8.1108e-01   -8.1948e-01   -4.7751e-03    2.9602e-04  \n",
      "Epoch: 3300,  Loss u: 4.679e-05, Loss pde: 5.747e-05, Loss_norm: 1.896e+00, Loss tot: 1.232e-04\n",
      "  coefs:  1.4516e-01   -9.1016e-02   -8.5363e-04    8.2681e-01   -8.3215e-01    1.7481e-04   -2.3508e-04  \n",
      "Epoch: 3350,  Loss u: 4.201e-05, Loss pde: 5.650e-05, Loss_norm: 1.904e+00, Loss tot: 1.176e-04\n",
      "  coefs:  1.3613e-01   -8.5447e-02   -4.5840e-04    8.3814e-01   -8.4313e-01    2.7849e-04   -2.0524e-06  \n",
      "Epoch: 3400,  Loss u: 3.865e-05, Loss pde: 5.529e-05, Loss_norm: 1.909e+00, Loss tot: 1.130e-04\n",
      "  coefs:  1.2932e-01   -8.1105e-02   -3.3772e-04    8.4630e-01   -8.5122e-01    3.3826e-04    7.3789e-06  \n",
      "Epoch: 3450,  Loss u: 3.599e-05, Loss pde: 5.387e-05, Loss_norm: 1.913e+00, Loss tot: 1.090e-04\n",
      "  coefs:  1.2348e-01   -7.7391e-02   -2.0427e-04    8.5314e-01   -8.5803e-01    3.1554e-04    7.1667e-06  \n",
      "Epoch: 3500,  Loss u: 7.770e-05, Loss pde: 6.610e-05, Loss_norm: 1.905e+00, Loss tot: 1.629e-04\n",
      "  coefs:  1.4400e-01   -8.7643e-02    4.2576e-04    8.3003e-01   -8.3742e-01   -3.9096e-03   -1.4765e-03  \n",
      "Epoch: 3550,  Loss u: 3.995e-05, Loss pde: 5.012e-05, Loss_norm: 1.909e+00, Loss tot: 1.092e-04\n",
      "  coefs:  1.2983e-01   -8.1105e-02   -6.9795e-04    8.4607e-01   -8.5022e-01   -1.2892e-05   -6.1839e-04  \n",
      "Epoch: 3600,  Loss u: 3.559e-05, Loss pde: 4.912e-05, Loss_norm: 1.915e+00, Loss tot: 1.039e-04\n",
      "  coefs:  1.2118e-01   -7.5957e-02   -2.0195e-04    8.5651e-01   -8.6061e-01    3.6585e-04   -1.3920e-04  \n",
      "Epoch: 3650,  Loss u: 3.273e-05, Loss pde: 4.790e-05, Loss_norm: 1.919e+00, Loss tot: 9.982e-05\n",
      "  coefs:  1.1477e-01   -7.1872e-02   -4.0643e-05    8.6405e-01   -8.6823e-01    4.5607e-04    2.5040e-06  \n",
      "Epoch: 3700,  Loss u: 3.040e-05, Loss pde: 4.658e-05, Loss_norm: 1.923e+00, Loss tot: 9.621e-05\n",
      "  coefs:  1.0932e-01   -6.8353e-02   -1.1650e-05    8.7037e-01   -8.7463e-01    4.1906e-04    1.2036e-05  \n",
      "Epoch: 3750,  Loss u: 7.789e-05, Loss pde: 7.186e-05, Loss_norm: 1.913e+00, Loss tot: 1.689e-04\n",
      "  coefs:  1.1014e-01   -6.3480e-02    7.4251e-03    8.5722e-01   -8.6820e-01   -5.4159e-03    1.5793e-03  \n",
      "Epoch: 3800,  Loss u: 3.690e-05, Loss pde: 4.423e-05, Loss_norm: 1.916e+00, Loss tot: 1.003e-04\n",
      "  coefs:  1.1743e-01   -7.4177e-02   -5.9323e-04    8.5921e-01   -8.6417e-01   -5.0679e-05   -4.8181e-04  \n",
      "Epoch: 3850,  Loss u: 3.025e-05, Loss pde: 4.214e-05, Loss_norm: 1.924e+00, Loss tot: 9.164e-05\n",
      "  coefs:  1.0801e-01   -6.7629e-02   -6.3173e-06    8.7199e-01   -8.7597e-01    1.3579e-04   -2.8909e-05  \n",
      "Epoch: 3900,  Loss u: 2.761e-05, Loss pde: 4.113e-05, Loss_norm: 1.929e+00, Loss tot: 8.803e-05\n",
      "  coefs:  1.0152e-01   -6.3416e-02   -1.7757e-05    8.7980e-01   -8.8375e-01    2.6656e-04    9.5789e-06  \n",
      "Epoch: 3950,  Loss u: 2.560e-05, Loss pde: 3.999e-05, Loss_norm: 1.932e+00, Loss tot: 8.492e-05\n",
      "  coefs:  9.6300e-02   -6.0084e-02    2.8384e-05    8.8589e-01   -8.8984e-01    2.8748e-04   -1.9112e-06  \n",
      "Epoch: 4000,  Loss u: 4.159e-05, Loss pde: 4.214e-05, Loss_norm: 1.922e+00, Loss tot: 1.029e-04\n",
      "  coefs:  1.1448e-01   -6.9158e-02    5.6547e-04    8.6404e-01   -8.7044e-01   -2.4149e-03   -1.1130e-03  \n",
      "Epoch: 4050,  Loss u: 2.909e-05, Loss pde: 3.729e-05, Loss_norm: 1.928e+00, Loss tot: 8.567e-05\n",
      "  coefs:  1.0251e-01   -6.3846e-02   -1.4353e-04    8.7906e-01   -8.8222e-01    2.8135e-04   -4.2692e-04  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 4100,  Loss u: 2.575e-05, Loss pde: 3.624e-05, Loss_norm: 1.935e+00, Loss tot: 8.134e-05\n",
      "  coefs:  9.4771e-02   -5.9229e-02   -1.0433e-05    8.8831e-01   -8.9162e-01    6.2750e-04    6.1659e-08  \n",
      "Epoch: 4150,  Loss u: 2.368e-05, Loss pde: 3.518e-05, Loss_norm: 1.938e+00, Loss tot: 7.824e-05\n",
      "  coefs:  8.9166e-02   -5.5553e-02    1.7405e-05    8.9481e-01   -8.9830e-01    5.5334e-04    4.6082e-06  \n",
      "Epoch: 4200,  Loss u: 2.204e-05, Loss pde: 3.409e-05, Loss_norm: 1.941e+00, Loss tot: 7.554e-05\n",
      "  coefs:  8.4422e-02   -5.2486e-02   -9.4028e-06    9.0023e-01   -9.0381e-01    4.7067e-04   -9.8486e-06  \n",
      "Epoch: 4250,  Loss u: 3.208e-05, Loss pde: 3.595e-05, Loss_norm: 1.931e+00, Loss tot: 8.733e-05\n",
      "  coefs:  9.5288e-02   -5.4849e-02    3.8374e-03    8.8301e-01   -8.8909e-01   -3.3627e-03    1.1628e-03  \n",
      "Epoch: 4300,  Loss u: 2.369e-05, Loss pde: 3.154e-05, Loss_norm: 1.938e+00, Loss tot: 7.461e-05\n",
      "  coefs:  8.7908e-02   -5.4567e-02   -2.0400e-05    8.9600e-01   -8.9934e-01    2.1859e-05   -1.3178e-04  \n",
      "Epoch: 4350,  Loss u: 2.113e-05, Loss pde: 3.081e-05, Loss_norm: 1.944e+00, Loss tot: 7.138e-05\n",
      "  coefs:  8.0981e-02   -5.0320e-02    6.1928e-06    9.0439e-01   -9.0773e-01    2.3979e-04   -1.0493e-05  \n",
      "Epoch: 4400,  Loss u: 1.958e-05, Loss pde: 2.992e-05, Loss_norm: 1.947e+00, Loss tot: 6.897e-05\n",
      "  coefs:  7.6127e-02   -4.7227e-02   -1.3406e-06    9.1005e-01   -9.1340e-01    2.5206e-04   -1.2371e-05  \n",
      "Epoch: 4450,  Loss u: 6.028e-05, Loss pde: 6.153e-05, Loss_norm: 1.943e+00, Loss tot: 1.412e-04\n",
      "  coefs:  8.8876e-02   -4.7064e-02    4.0807e-03    8.9320e-01   -9.0191e-01   -6.8039e-03   -1.4957e-03  \n",
      "Epoch: 4500,  Loss u: 2.333e-05, Loss pde: 2.810e-05, Loss_norm: 1.942e+00, Loss tot: 7.085e-05\n",
      "  coefs:  8.3698e-02   -5.2202e-02    6.5941e-05    9.0166e-01   -9.0385e-01    2.7434e-04   -3.2332e-04  \n",
      "Epoch: 4550,  Loss u: 2.011e-05, Loss pde: 2.694e-05, Loss_norm: 1.948e+00, Loss tot: 6.654e-05\n",
      "  coefs:  7.5818e-02   -4.7142e-02    2.2633e-05    9.1103e-01   -9.1379e-01    6.0107e-04   -1.5150e-06  \n",
      "Epoch: 4600,  Loss u: 1.845e-05, Loss pde: 2.611e-05, Loss_norm: 1.952e+00, Loss tot: 6.408e-05\n",
      "  coefs:  7.0459e-02   -4.3622e-02    3.1012e-05    9.1715e-01   -9.2010e-01    5.0065e-04    1.5749e-05  \n",
      "Epoch: 4650,  Loss u: 3.645e-04, Loss pde: 2.092e-04, Loss_norm: 1.943e+00, Loss tot: 5.931e-04\n",
      "  coefs:  6.8992e-02   -3.9974e-02    4.4293e-03    9.0965e-01   -9.1595e-01   -3.5869e-03    3.7723e-04  \n",
      "Epoch: 4700,  Loss u: 2.068e-05, Loss pde: 2.507e-05, Loss_norm: 1.948e+00, Loss tot: 6.524e-05\n",
      "  coefs:  7.3513e-02   -4.5772e-02    2.0231e-04    9.1297e-01   -9.1574e-01    6.1598e-05    3.6255e-05  \n",
      "Epoch: 4750,  Loss u: 1.752e-05, Loss pde: 2.375e-05, Loss_norm: 1.954e+00, Loss tot: 6.081e-05\n",
      "  coefs:  6.6485e-02   -4.1099e-02    2.9621e-05    9.2183e-01   -9.2463e-01    2.3809e-04   -4.8321e-06  \n",
      "Epoch: 4800,  Loss u: 1.619e-05, Loss pde: 2.311e-05, Loss_norm: 1.957e+00, Loss tot: 5.887e-05\n",
      "  coefs:  6.1798e-02   -3.8119e-02    3.5623e-05    9.2720e-01   -9.3004e-01    2.1429e-04   -1.0067e-05  \n",
      "Epoch: 4850,  Loss u: 1.518e-05, Loss pde: 2.243e-05, Loss_norm: 1.960e+00, Loss tot: 5.721e-05\n",
      "  coefs:  5.7999e-02   -3.5697e-02    6.5558e-05    9.3153e-01   -9.3434e-01    1.5478e-04    1.4897e-05  \n",
      "Epoch: 4900,  Loss u: 5.893e-05, Loss pde: 4.384e-05, Loss_norm: 1.950e+00, Loss tot: 1.223e-04\n",
      "  coefs:  7.1172e-02   -4.1085e-02    2.1290e-03    9.1482e-01   -9.1878e-01   -1.8157e-03    6.9048e-05  \n",
      "Epoch: 4950,  Loss u: 1.709e-05, Loss pde: 2.088e-05, Loss_norm: 1.957e+00, Loss tot: 5.754e-05\n",
      "  coefs:  6.3221e-02   -3.9010e-02    5.2407e-05    9.2603e-01   -9.2812e-01    4.7651e-04   -1.2831e-04  \n",
      "Epoch: 5000,  Loss u: 1.541e-05, Loss pde: 2.016e-05, Loss_norm: 1.961e+00, Loss tot: 5.518e-05\n",
      "  coefs:  5.7310e-02   -3.5277e-02    7.4803e-06    9.3281e-01   -9.3526e-01    4.4399e-04   -1.0876e-05  \n",
      "Epoch: 5050,  Loss u: 7.575e-05, Loss pde: 4.093e-05, Loss_norm: 1.954e+00, Loss tot: 1.362e-04\n",
      "  coefs:  6.7105e-02   -3.9581e-02    6.2086e-04    9.2001e-01   -9.2397e-01   -1.6373e-03   -6.7990e-04  \n",
      "Epoch: 5100,  Loss u: 1.609e-05, Loss pde: 1.908e-05, Loss_norm: 1.959e+00, Loss tot: 5.476e-05\n",
      "  coefs:  5.9108e-02   -3.6470e-02    2.3464e-05    9.3071e-01   -9.3315e-01   -1.6661e-05   -1.7083e-05  \n",
      "Epoch: 5150,  Loss u: 1.459e-05, Loss pde: 1.861e-05, Loss_norm: 1.964e+00, Loss tot: 5.283e-05\n",
      "  coefs:  5.3301e-02   -3.2743e-02    2.8996e-05    9.3748e-01   -9.3991e-01    2.1470e-04    2.9819e-05  \n",
      "Epoch: 5200,  Loss u: 1.357e-05, Loss pde: 1.816e-05, Loss_norm: 1.966e+00, Loss tot: 5.139e-05\n",
      "  coefs:  4.9216e-02   -3.0147e-02    5.1438e-05    9.4212e-01   -9.4451e-01    1.7620e-04    8.9253e-06  \n",
      "Epoch: 5250,  Loss u: 1.278e-05, Loss pde: 1.769e-05, Loss_norm: 1.968e+00, Loss tot: 5.015e-05\n",
      "  coefs:  4.5887e-02   -2.8019e-02    6.2276e-05    9.4587e-01   -9.4823e-01    7.5016e-05    1.4968e-06  \n",
      "Epoch: 5300,  Loss u: 5.130e-05, Loss pde: 3.546e-05, Loss_norm: 1.963e+00, Loss tot: 1.064e-04\n",
      "  coefs:  5.4821e-02   -2.9401e-02    3.2622e-03    9.3172e-01   -9.3768e-01   -4.3155e-03   -2.0737e-03  \n",
      "Epoch: 5350,  Loss u: 1.534e-05, Loss pde: 1.669e-05, Loss_norm: 1.965e+00, Loss tot: 5.168e-05\n",
      "  coefs:  5.2154e-02   -3.1920e-02   -1.5789e-05    9.3937e-01   -9.4083e-01    2.9580e-04   -1.2555e-04  \n",
      "Epoch: 5400,  Loss u: 3.068e-05, Loss pde: 2.308e-05, Loss_norm: 1.968e+00, Loss tot: 7.345e-05\n",
      "  coefs:  4.6109e-02   -2.8151e-02   -3.5859e-05    9.4562e-01   -9.4772e-01    3.8542e-04    9.7115e-06  \n",
      "Epoch: 5450,  Loss u: 1.799e-05, Loss pde: 1.675e-05, Loss_norm: 1.965e+00, Loss tot: 5.439e-05\n",
      "  coefs:  5.1984e-02   -3.2909e-02   -2.3741e-05    9.3855e-01   -9.4098e-01    4.6821e-04   -2.5727e-04  \n",
      "Epoch: 5500,  Loss u: 1.309e-05, Loss pde: 1.532e-05, Loss_norm: 1.969e+00, Loss tot: 4.810e-05\n",
      "  coefs:  4.5572e-02   -2.7810e-02    2.1707e-05    9.4671e-01   -9.4871e-01    2.2055e-04   -2.6003e-05  \n",
      "Epoch: 5550,  Loss u: 1.212e-05, Loss pde: 1.502e-05, Loss_norm: 1.972e+00, Loss tot: 4.685e-05\n",
      "  coefs:  4.1329e-02   -2.5155e-02    5.0265e-05    9.5149e-01   -9.5356e-01    1.5088e-04   -1.8466e-05  \n",
      "Epoch: 5600,  Loss u: 1.141e-05, Loss pde: 1.470e-05, Loss_norm: 1.974e+00, Loss tot: 4.584e-05\n",
      "  coefs:  3.8110e-02   -2.3102e-02    7.1262e-05    9.5510e-01   -9.5714e-01    6.9874e-05    7.3833e-06  \n",
      "Epoch: 5650,  Loss u: 1.595e-04, Loss pde: 8.617e-05, Loss_norm: 1.975e+00, Loss tot: 2.654e-04\n",
      "  coefs:  4.2265e-02   -2.3141e-02    4.9415e-03    9.4771e-01   -9.5144e-01   -5.5448e-03   -2.7600e-04  \n",
      "Epoch: 5700,  Loss u: 1.365e-05, Loss pde: 1.434e-05, Loss_norm: 1.970e+00, Loss tot: 4.769e-05\n",
      "  coefs:  4.3158e-02   -2.6254e-02   -1.0868e-04    9.4957e-01   -9.5107e-01    8.5287e-05   -2.4347e-05  \n",
      "Epoch: 5750,  Loss u: 1.135e-05, Loss pde: 1.370e-05, Loss_norm: 1.974e+00, Loss tot: 4.479e-05\n",
      "  coefs:  3.7935e-02   -2.3042e-02   -2.4299e-05    9.5548e-01   -9.5738e-01    5.8070e-05   -3.8934e-05  \n",
      "Epoch: 5800,  Loss u: 7.634e-05, Loss pde: 4.491e-05, Loss_norm: 1.968e+00, Loss tot: 1.409e-04\n",
      "  coefs:  4.3381e-02   -2.2619e-02    3.0462e-03    9.4418e-01   -9.4999e-01   -4.0098e-03   -8.5720e-04  \n",
      "Epoch: 5850,  Loss u: 1.336e-05, Loss pde: 1.300e-05, Loss_norm: 1.971e+00, Loss tot: 4.606e-05\n",
      "  coefs:  4.4226e-02   -2.6793e-02   -3.6255e-05    9.4912e-01   -9.5008e-01    3.8132e-04   -1.2729e-04  \n",
      "Epoch: 5900,  Loss u: 1.169e-05, Loss pde: 1.263e-05, Loss_norm: 1.975e+00, Loss tot: 4.407e-05\n",
      "  coefs:  3.8051e-02   -2.3057e-02   -1.8948e-05    9.5566e-01   -9.5744e-01    3.8352e-04    1.5135e-05  \n",
      "Epoch: 5950,  Loss u: 1.081e-05, Loss pde: 1.242e-05, Loss_norm: 1.977e+00, Loss tot: 4.300e-05\n",
      "  coefs:  3.4080e-02   -2.0551e-02   -2.7294e-05    9.6010e-01   -9.6187e-01    2.1688e-04    2.2375e-05  \n",
      "Epoch: 6000,  Loss u: 2.597e-05, Loss pde: 1.824e-05, Loss_norm: 1.973e+00, Loss tot: 6.393e-05\n",
      "  coefs:  3.2210e-02   -1.8355e-02    3.3837e-03    9.5701e-01   -9.5975e-01   -2.0506e-03    3.2243e-04  \n",
      "Epoch: 6050,  Loss u: 1.317e-05, Loss pde: 1.272e-05, Loss_norm: 1.975e+00, Loss tot: 4.564e-05\n",
      "  coefs:  3.7593e-02   -2.2739e-02   -9.2354e-05    9.5643e-01   -9.5779e-01   -3.5848e-05   -8.4332e-05  \n",
      "Epoch: 6100,  Loss u: 1.042e-05, Loss pde: 1.183e-05, Loss_norm: 1.978e+00, Loss tot: 4.202e-05\n",
      "  coefs:  3.2446e-02   -1.9580e-02   -3.3820e-05    9.6194e-01   -9.6364e-01    1.1834e-04    3.0391e-05  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 6150,  Loss u: 9.763e-06, Loss pde: 1.164e-05, Loss_norm: 1.980e+00, Loss tot: 4.120e-05\n",
      "  coefs:  2.9367e-02   -1.7619e-02    7.1316e-06    9.6542e-01   -9.6706e-01    5.9376e-05   -1.8690e-05  \n",
      "Epoch: 6200,  Loss u: 4.945e-04, Loss pde: 3.176e-04, Loss_norm: 1.978e+00, Loss tot: 8.319e-04\n",
      "  coefs:  2.8442e-02   -1.4410e-02    1.7828e-04    9.6593e-01   -9.6692e-01   -4.8440e-04    1.7309e-03  \n",
      "Epoch: 6250,  Loss u: 1.285e-05, Loss pde: 1.246e-05, Loss_norm: 1.975e+00, Loss tot: 4.506e-05\n",
      "  coefs:  3.5855e-02   -2.1679e-02    9.1636e-05    9.5857e-01   -9.5909e-01   -5.1801e-05   -5.5453e-05  \n",
      "Epoch: 6300,  Loss u: 1.007e-05, Loss pde: 1.079e-05, Loss_norm: 1.979e+00, Loss tot: 4.065e-05\n",
      "  coefs:  3.0594e-02   -1.8345e-02   -1.1130e-05    9.6425e-01   -9.6581e-01    2.1419e-04    2.0141e-05  \n",
      "Epoch: 6350,  Loss u: 1.730e-04, Loss pde: 9.995e-05, Loss_norm: 1.970e+00, Loss tot: 2.927e-04\n",
      "  coefs:  3.4239e-02   -1.8565e-02    3.3770e-03    9.5347e-01   -9.5634e-01   -3.5363e-03   -5.2036e-04  \n",
      "Epoch: 6400,  Loss u: 1.080e-05, Loss pde: 1.057e-05, Loss_norm: 1.977e+00, Loss tot: 4.114e-05\n",
      "  coefs:  3.4060e-02   -2.0529e-02   -1.1558e-04    9.6034e-01   -9.6185e-01    1.8855e-04   -1.3233e-04  \n",
      "Epoch: 6450,  Loss u: 9.694e-06, Loss pde: 1.040e-05, Loss_norm: 1.980e+00, Loss tot: 3.989e-05\n",
      "  coefs:  2.9119e-02   -1.7519e-02   -3.3208e-05    9.6594e-01   -9.6747e-01    1.7690e-04   -1.0490e-06  \n",
      "Epoch: 6500,  Loss u: 9.055e-06, Loss pde: 1.026e-05, Loss_norm: 1.982e+00, Loss tot: 3.914e-05\n",
      "  coefs:  2.5980e-02   -1.5529e-02   -7.0279e-06    9.6943e-01   -9.7093e-01    1.9434e-05   -2.5390e-05  \n",
      "Epoch: 6550,  Loss u: 8.584e-06, Loss pde: 1.015e-05, Loss_norm: 1.983e+00, Loss tot: 3.857e-05\n",
      "  coefs:  2.3678e-02   -1.4066e-02    1.0510e-05    9.7200e-01   -9.7344e-01    9.0540e-05   -1.2973e-05  \n",
      "Epoch: 6600,  Loss u: 1.158e-05, Loss pde: 1.114e-05, Loss_norm: 1.980e+00, Loss tot: 4.252e-05\n",
      "  coefs:  3.1182e-02   -1.9192e-02   -1.5575e-04    9.6322e-01   -9.6523e-01   -1.6986e-04   -3.9138e-04  \n",
      "Epoch: 6650,  Loss u: 9.038e-06, Loss pde: 9.609e-06, Loss_norm: 1.982e+00, Loss tot: 3.846e-05\n",
      "  coefs:  2.6463e-02   -1.5795e-02   -3.8887e-06    9.6905e-01   -9.7039e-01    1.1346e-04   -1.0377e-06  \n",
      "Epoch: 6700,  Loss u: 1.140e-04, Loss pde: 7.070e-05, Loss_norm: 1.985e+00, Loss tot: 2.046e-04\n",
      "  coefs:  2.9511e-02   -1.5718e-02    4.1490e-03    9.6400e-01   -9.6594e-01   -4.9983e-03   -4.2846e-04  \n",
      "Epoch: 6750,  Loss u: 1.058e-05, Loss pde: 1.001e-05, Loss_norm: 1.980e+00, Loss tot: 4.039e-05\n",
      "  coefs:  2.9692e-02   -1.7791e-02   -2.9167e-04    9.6528e-01   -9.6659e-01    3.6701e-04   -1.5729e-04  \n",
      "Epoch: 6800,  Loss u: 8.696e-06, Loss pde: 9.265e-06, Loss_norm: 1.983e+00, Loss tot: 3.779e-05\n",
      "  coefs:  2.5226e-02   -1.5111e-02    4.2406e-05    9.7045e-01   -9.7182e-01    1.8373e-04    1.8133e-05  \n",
      "Epoch: 6850,  Loss u: 8.140e-06, Loss pde: 9.159e-06, Loss_norm: 1.984e+00, Loss tot: 3.714e-05\n",
      "  coefs:  2.2394e-02   -1.3306e-02   -4.5907e-07    9.7358e-01   -9.7492e-01    7.0451e-05    1.2590e-05  \n",
      "Epoch: 6900,  Loss u: 1.573e-05, Loss pde: 1.507e-05, Loss_norm: 1.980e+00, Loss tot: 5.060e-05\n",
      "  coefs:  2.6098e-02   -1.3490e-02    1.4778e-03    9.6721e-01   -9.6931e-01   -2.0436e-03   -5.3420e-04  \n",
      "Epoch: 6950,  Loss u: 8.892e-06, Loss pde: 8.789e-06, Loss_norm: 1.983e+00, Loss tot: 3.751e-05\n",
      "  coefs:  2.5841e-02   -1.5442e-02   -9.7926e-05    9.6996e-01   -9.7090e-01    2.1437e-04   -8.7789e-05  \n",
      "Epoch: 7000,  Loss u: 8.030e-06, Loss pde: 8.647e-06, Loss_norm: 1.985e+00, Loss tot: 3.652e-05\n",
      "  coefs:  2.2292e-02   -1.3202e-02    3.0778e-05    9.7383e-01   -9.7509e-01    1.4339e-04    2.5887e-05  \n",
      "Epoch: 7050,  Loss u: 5.673e-05, Loss pde: 4.316e-05, Loss_norm: 1.975e+00, Loss tot: 1.196e-04\n",
      "  coefs:  2.9398e-02   -1.6937e-02   -4.4400e-04    9.6310e-01   -9.6413e-01    2.5366e-04   -5.6413e-04  \n",
      "Epoch: 7100,  Loss u: 8.908e-06, Loss pde: 8.482e-06, Loss_norm: 1.982e+00, Loss tot: 3.721e-05\n",
      "  coefs:  2.6243e-02   -1.5778e-02   -1.2955e-04    9.6927e-01   -9.7062e-01    2.1948e-04   -7.7777e-05  \n",
      "Epoch: 7150,  Loss u: 7.928e-06, Loss pde: 8.390e-06, Loss_norm: 1.985e+00, Loss tot: 3.616e-05\n",
      "  coefs:  2.2409e-02   -1.3375e-02   -1.5636e-05    9.7373e-01   -9.7497e-01    1.2836e-04   -1.1375e-05  \n",
      "Epoch: 7200,  Loss u: 7.443e-06, Loss pde: 8.299e-06, Loss_norm: 1.986e+00, Loss tot: 3.560e-05\n",
      "  coefs:  1.9851e-02   -1.1753e-02   -6.5633e-05    9.7653e-01   -9.7776e-01    2.6834e-05    3.2732e-05  \n",
      "Epoch: 7250,  Loss u: 9.758e-06, Loss pde: 9.080e-06, Loss_norm: 1.983e+00, Loss tot: 3.866e-05\n",
      "  coefs:  2.6271e-02   -1.5446e-02   -7.0787e-04    9.6912e-01   -9.7043e-01    1.8094e-04   -4.3077e-04  \n",
      "Epoch: 7300,  Loss u: 8.083e-06, Loss pde: 8.013e-06, Loss_norm: 1.985e+00, Loss tot: 3.594e-05\n",
      "  coefs:  2.2581e-02   -1.3595e-02   -2.1549e-04    9.7344e-01   -9.7465e-01    1.8473e-04   -1.0768e-04  \n",
      "Epoch: 7350,  Loss u: 7.293e-06, Loss pde: 7.868e-06, Loss_norm: 1.986e+00, Loss tot: 3.502e-05\n",
      "  coefs:  1.9646e-02   -1.1613e-02   -7.2092e-05    9.7686e-01   -9.7802e-01    4.4673e-05   -2.1361e-05  \n",
      "Epoch: 7400,  Loss u: 5.264e-05, Loss pde: 3.447e-05, Loss_norm: 1.982e+00, Loss tot: 1.069e-04\n",
      "  coefs:  2.4044e-02   -1.3011e-02    3.0909e-03    9.6943e-01   -9.6972e-01   -2.6593e-03    5.2055e-04  \n",
      "Epoch: 7450,  Loss u: 7.887e-06, Loss pde: 7.803e-06, Loss_norm: 1.984e+00, Loss tot: 3.553e-05\n",
      "  coefs:  2.2463e-02   -1.3467e-02   -1.0779e-06    9.7363e-01   -9.7476e-01    3.4045e-05   -6.8509e-05  \n",
      "Epoch: 7500,  Loss u: 7.066e-06, Loss pde: 7.653e-06, Loss_norm: 1.987e+00, Loss tot: 3.459e-05\n",
      "  coefs:  1.9298e-02   -1.1452e-02   -6.8727e-05    9.7727e-01   -9.7834e-01    1.3734e-04   -1.2198e-05  \n",
      "Epoch: 7550,  Loss u: 9.507e-06, Loss pde: 9.869e-06, Loss_norm: 1.988e+00, Loss tot: 3.925e-05\n",
      "  coefs:  1.7274e-02   -1.0008e-02   -1.2081e-04    9.7959e-01   -9.8049e-01    1.3876e-04    1.0939e-04  \n",
      "Epoch: 7600,  Loss u: 1.138e-05, Loss pde: 1.003e-05, Loss_norm: 1.983e+00, Loss tot: 4.124e-05\n",
      "  coefs:  2.4136e-02   -1.5320e-02    3.5864e-06    9.7077e-01   -9.7204e-01    5.2536e-05   -5.0707e-04  \n",
      "Epoch: 7650,  Loss u: 7.144e-06, Loss pde: 7.270e-06, Loss_norm: 1.986e+00, Loss tot: 3.428e-05\n",
      "  coefs:  2.0425e-02   -1.2086e-02   -5.2808e-05    9.7611e-01   -9.7723e-01    1.7242e-04   -1.7873e-05  \n",
      "Epoch: 7700,  Loss u: 7.188e-06, Loss pde: 7.379e-06, Loss_norm: 1.988e+00, Loss tot: 3.444e-05\n",
      "  coefs:  1.7711e-02   -1.0444e-02    6.5030e-05    9.7915e-01   -9.8020e-01    4.6122e-05   -2.1655e-06  \n",
      "Epoch: 7750,  Loss u: 1.445e-05, Loss pde: 7.960e-06, Loss_norm: 1.985e+00, Loss tot: 4.227e-05\n",
      "  coefs:  2.3048e-02   -1.4748e-02   -2.1403e-04    9.7293e-01   -9.7375e-01    2.9317e-04   -3.4223e-04  \n",
      "Epoch: 7800,  Loss u: 6.844e-06, Loss pde: 7.100e-06, Loss_norm: 1.987e+00, Loss tot: 3.381e-05\n",
      "  coefs:  1.9315e-02   -1.1524e-02   -6.2105e-05    9.7736e-01   -9.7835e-01    1.5854e-04   -2.1422e-07  \n",
      "Epoch: 7850,  Loss u: 6.357e-06, Loss pde: 7.033e-06, Loss_norm: 1.988e+00, Loss tot: 3.327e-05\n",
      "  coefs:  1.6826e-02   -9.9589e-03   -2.1365e-05    9.8004e-01   -9.8105e-01    2.4348e-05   -1.7184e-05  \n",
      "Epoch: 7900,  Loss u: 6.052e-06, Loss pde: 6.962e-06, Loss_norm: 1.989e+00, Loss tot: 3.290e-05\n",
      "  coefs:  1.5146e-02   -8.9275e-03   -1.2989e-05    9.8190e-01   -9.8286e-01    1.2609e-04   -1.5855e-06  \n",
      "Epoch: 7950,  Loss u: 1.640e-05, Loss pde: 1.358e-05, Loss_norm: 1.985e+00, Loss tot: 4.982e-05\n",
      "  coefs:  2.4172e-02   -1.4934e-02   -1.0358e-03    9.7167e-01   -9.7277e-01   -8.9711e-05   -1.1192e-04  \n",
      "Epoch: 8000,  Loss u: 6.729e-06, Loss pde: 6.739e-06, Loss_norm: 1.987e+00, Loss tot: 3.334e-05\n",
      "  coefs:  1.9948e-02   -1.1740e-02   -2.7886e-04    9.7683e-01   -9.7787e-01    2.6776e-04   -4.9919e-05  \n",
      "Epoch: 8050,  Loss u: 2.772e-04, Loss pde: 1.024e-04, Loss_norm: 1.986e+00, Loss tot: 3.995e-04\n",
      "  coefs:  1.6068e-02   -9.5354e-03   -5.4599e-04    9.7860e-01   -9.7988e-01    1.0533e-03   -4.3100e-05  \n",
      "Epoch: 8100,  Loss u: 7.984e-06, Loss pde: 7.067e-06, Loss_norm: 1.986e+00, Loss tot: 3.491e-05\n",
      "  coefs:  2.1334e-02   -1.2674e-02   -2.4176e-04    9.7554e-01   -9.7625e-01   -3.5874e-05   -1.8339e-04  \n",
      "Epoch: 8150,  Loss u: 6.234e-06, Loss pde: 6.560e-06, Loss_norm: 1.988e+00, Loss tot: 3.267e-05\n",
      "  coefs:  1.7557e-02   -1.0490e-02   -6.1071e-05    9.7932e-01   -9.8028e-01    1.9291e-04   -6.0756e-05  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 8200,  Loss u: 5.828e-06, Loss pde: 6.505e-06, Loss_norm: 1.989e+00, Loss tot: 3.222e-05\n",
      "  coefs:  1.5362e-02   -9.0846e-03    1.4188e-05    9.8171e-01   -9.8264e-01    5.5722e-05    2.2549e-05  \n",
      "Epoch: 8250,  Loss u: 5.557e-06, Loss pde: 6.469e-06, Loss_norm: 1.990e+00, Loss tot: 3.192e-05\n",
      "  coefs:  1.3850e-02   -8.1336e-03   -8.2403e-05    9.8337e-01   -9.8425e-01    8.2924e-05   -5.2879e-05  \n",
      "Epoch: 8300,  Loss u: 7.856e-06, Loss pde: 7.828e-06, Loss_norm: 1.988e+00, Loss tot: 3.557e-05\n",
      "  coefs:  1.6852e-02   -1.0527e-02   -1.6969e-04    9.8003e-01   -9.8059e-01    1.1227e-04    7.8793e-06  \n",
      "Epoch: 8350,  Loss u: 5.507e-06, Loss pde: 6.282e-06, Loss_norm: 1.989e+00, Loss tot: 3.168e-05\n",
      "  coefs:  1.4766e-02   -8.7009e-03    4.8008e-05    9.8242e-01   -9.8327e-01    1.2784e-04   -4.8345e-05  \n",
      "Epoch: 8400,  Loss u: 1.829e-05, Loss pde: 1.530e-05, Loss_norm: 1.980e+00, Loss tot: 5.338e-05\n",
      "  coefs:  1.9166e-02   -1.0498e-02    7.3729e-04    9.7349e-01   -9.7494e-01   -7.2533e-05    8.2981e-04  \n",
      "Epoch: 8450,  Loss u: 6.215e-06, Loss pde: 6.060e-06, Loss_norm: 1.987e+00, Loss tot: 3.214e-05\n",
      "  coefs:  1.9397e-02   -1.1649e-02   -2.2785e-04    9.7728e-01   -9.7812e-01    1.0678e-04   -1.5663e-04  \n",
      "Epoch: 8500,  Loss u: 2.894e-04, Loss pde: 9.663e-05, Loss_norm: 1.993e+00, Loss tot: 4.059e-04\n",
      "  coefs:  1.7462e-02   -1.0182e-02    1.1502e-03    9.8103e-01   -9.8155e-01   -1.2507e-03    1.8868e-04  \n",
      "Epoch: 8550,  Loss u: 6.190e-06, Loss pde: 6.155e-06, Loss_norm: 1.988e+00, Loss tot: 3.223e-05\n",
      "  coefs:  1.8668e-02   -1.1022e-02   -3.0207e-04    9.7856e-01   -9.7927e-01    2.1119e-04   -1.6920e-05  \n",
      "Epoch: 8600,  Loss u: 5.452e-06, Loss pde: 5.959e-06, Loss_norm: 1.989e+00, Loss tot: 3.130e-05\n",
      "  coefs:  1.5518e-02   -9.2463e-03   -2.1923e-05    9.8162e-01   -9.8246e-01    1.1056e-04    3.5384e-05  \n",
      "Epoch: 8650,  Loss u: 5.127e-06, Loss pde: 5.916e-06, Loss_norm: 1.990e+00, Loss tot: 3.094e-05\n",
      "  coefs:  1.3659e-02   -8.0478e-03    4.5883e-06    9.8367e-01   -9.8447e-01    6.4698e-05   -3.1516e-05  \n",
      "Epoch: 8700,  Loss u: 7.346e-06, Loss pde: 7.113e-06, Loss_norm: 1.990e+00, Loss tot: 3.436e-05\n",
      "  coefs:  1.2254e-02   -7.2335e-03   -1.2869e-04    9.8496e-01   -9.8578e-01    1.1641e-04   -1.5217e-07  \n",
      "Epoch: 8750,  Loss u: 8.851e-06, Loss pde: 7.501e-06, Loss_norm: 1.987e+00, Loss tot: 3.623e-05\n",
      "  coefs:  1.7956e-02   -1.1191e-02   -1.3795e-04    9.7874e-01   -9.7925e-01    6.2963e-05    2.2025e-05  \n",
      "Epoch: 8800,  Loss u: 5.146e-06, Loss pde: 5.716e-06, Loss_norm: 1.989e+00, Loss tot: 3.076e-05\n",
      "  coefs:  1.5124e-02   -9.0369e-03   -7.3325e-05    9.8206e-01   -9.8287e-01    2.1974e-04   -2.3729e-05  \n",
      "Epoch: 8850,  Loss u: 3.186e-05, Loss pde: 2.442e-05, Loss_norm: 1.983e+00, Loss tot: 7.611e-05\n",
      "  coefs:  1.4272e-02   -7.3804e-03    7.2039e-04    9.7935e-01   -9.8012e-01   -6.0644e-04    3.3258e-04  \n",
      "Epoch: 8900,  Loss u: 5.634e-06, Loss pde: 5.693e-06, Loss_norm: 1.988e+00, Loss tot: 3.121e-05\n",
      "  coefs:  1.7100e-02   -1.0061e-02   -2.5254e-04    9.7991e-01   -9.8045e-01    1.0327e-04   -1.1138e-04  \n",
      "Epoch: 8950,  Loss u: 4.965e-06, Loss pde: 5.538e-06, Loss_norm: 1.990e+00, Loss tot: 3.040e-05\n",
      "  coefs:  1.4499e-02   -8.5898e-03   -7.3369e-05    9.8281e-01   -9.8358e-01    1.3043e-04   -1.0628e-04  \n",
      "Epoch: 9000,  Loss u: 1.462e-05, Loss pde: 7.202e-06, Loss_norm: 1.991e+00, Loss tot: 4.173e-05\n",
      "  coefs:  1.2641e-02   -7.3810e-03   -2.4258e-04    9.8470e-01   -9.8537e-01    3.1046e-04   -4.8450e-05  \n",
      "Epoch: 9050,  Loss u: 7.029e-06, Loss pde: 6.274e-06, Loss_norm: 1.989e+00, Loss tot: 3.319e-05\n",
      "  coefs:  1.4567e-02   -8.7034e-03   -6.0550e-05    9.8276e-01   -9.8284e-01    9.8656e-05   -1.3142e-04  \n",
      "Epoch: 9100,  Loss u: 4.584e-06, Loss pde: 5.464e-06, Loss_norm: 1.991e+00, Loss tot: 2.995e-05\n",
      "  coefs:  1.2758e-02   -7.5840e-03   -9.1262e-05    9.8468e-01   -9.8534e-01    9.3728e-05   -9.5665e-05  \n",
      "Epoch: 9150,  Loss u: 7.148e-05, Loss pde: 5.460e-05, Loss_norm: 1.992e+00, Loss tot: 1.460e-04\n",
      "  coefs:  1.2248e-02   -6.3958e-03   -3.0503e-04    9.8633e-01   -9.8590e-01   -4.3252e-05    7.4821e-04  \n",
      "Epoch: 9200,  Loss u: 4.924e-06, Loss pde: 5.472e-06, Loss_norm: 1.990e+00, Loss tot: 3.030e-05\n",
      "  coefs:  1.5058e-02   -8.7291e-03   -1.9606e-04    9.8260e-01   -9.8304e-01    1.9222e-04   -1.0941e-04  \n",
      "Epoch: 9250,  Loss u: 4.464e-06, Loss pde: 5.321e-06, Loss_norm: 1.991e+00, Loss tot: 2.969e-05\n",
      "  coefs:  1.2755e-02   -7.5131e-03   -1.2270e-04    9.8473e-01   -9.8538e-01    4.8737e-05   -6.8363e-06  \n",
      "Epoch: 9300,  Loss u: 1.016e-05, Loss pde: 7.167e-06, Loss_norm: 1.989e+00, Loss tot: 3.721e-05\n",
      "  coefs:  1.7722e-02   -1.1218e-02   -5.1183e-04    9.7888e-01   -9.7940e-01    7.2602e-04   -3.3139e-04  \n",
      "Epoch: 9350,  Loss u: 4.619e-06, Loss pde: 5.143e-06, Loss_norm: 1.990e+00, Loss tot: 2.966e-05\n",
      "  coefs:  1.4863e-02   -8.9010e-03   -4.7022e-05    9.8245e-01   -9.8320e-01    8.6401e-05    2.5362e-05  \n",
      "Epoch: 9400,  Loss u: 4.253e-06, Loss pde: 5.118e-06, Loss_norm: 1.991e+00, Loss tot: 2.928e-05\n",
      "  coefs:  1.2716e-02   -7.5607e-03   -1.3101e-05    9.8478e-01   -9.8547e-01    7.4148e-05   -3.9098e-06  \n",
      "Epoch: 9450,  Loss u: 1.951e-05, Loss pde: 1.372e-05, Loss_norm: 1.987e+00, Loss tot: 5.310e-05\n",
      "  coefs:  1.4891e-02   -9.0862e-03    2.9070e-04    9.8094e-01   -9.8134e-01    2.2121e-04    1.6315e-04  \n",
      "Epoch: 9500,  Loss u: 4.567e-06, Loss pde: 5.124e-06, Loss_norm: 1.990e+00, Loss tot: 2.959e-05\n",
      "  coefs:  1.4325e-02   -8.5511e-03   -2.0156e-04    9.8292e-01   -9.8353e-01    5.3085e-05   -9.9273e-05  \n",
      "Epoch: 9550,  Loss u: 4.139e-06, Loss pde: 4.992e-06, Loss_norm: 1.991e+00, Loss tot: 2.904e-05\n",
      "  coefs:  1.2382e-02   -7.3293e-03   -3.4467e-05    9.8518e-01   -9.8578e-01    5.3722e-05   -7.4807e-05  \n",
      "Epoch: 9600,  Loss u: 7.857e-05, Loss pde: 1.940e-05, Loss_norm: 1.991e+00, Loss tot: 1.179e-04\n",
      "  coefs:  1.2589e-02   -7.1132e-03    1.7044e-03    9.8356e-01   -9.8395e-01   -1.6884e-03   -1.5402e-04  \n",
      "Epoch: 9650,  Loss u: 4.598e-06, Loss pde: 4.972e-06, Loss_norm: 1.990e+00, Loss tot: 2.947e-05\n",
      "  coefs:  1.3456e-02   -7.9663e-03    2.6880e-06    9.8392e-01   -9.8437e-01    2.9148e-05   -6.2191e-05  \n",
      "Epoch: 9700,  Loss u: 3.904e-06, Loss pde: 4.909e-06, Loss_norm: 1.991e+00, Loss tot: 2.873e-05\n",
      "  coefs:  1.1739e-02   -6.9505e-03   -3.2399e-06    9.8591e-01   -9.8647e-01    9.2503e-05   -1.0182e-04  \n",
      "Epoch: 9750,  Loss u: 2.122e-05, Loss pde: 5.321e-06, Loss_norm: 1.989e+00, Loss tot: 4.643e-05\n",
      "  coefs:  1.2716e-02   -7.7434e-03   -9.9044e-05    9.8392e-01   -9.8402e-01    6.2865e-04    6.5773e-05  \n",
      "Epoch: 9800,  Loss u: 3.889e-06, Loss pde: 4.784e-06, Loss_norm: 1.991e+00, Loss tot: 2.858e-05\n",
      "  coefs:  1.2287e-02   -7.2787e-03   -3.2903e-05    9.8533e-01   -9.8585e-01    1.5885e-04   -3.5406e-05  \n",
      "Epoch: 9850,  Loss u: 4.087e-06, Loss pde: 5.004e-06, Loss_norm: 1.992e+00, Loss tot: 2.901e-05\n",
      "  coefs:  1.0867e-02   -6.4539e-03    1.1614e-04    9.8690e-01   -9.8749e-01    1.4883e-04   -1.6769e-04  \n",
      "Epoch: 9900,  Loss u: 7.389e-06, Loss pde: 5.349e-06, Loss_norm: 1.989e+00, Loss tot: 3.263e-05\n",
      "  coefs:  1.4879e-02   -9.2160e-03   -1.9759e-04    9.8212e-01   -9.8276e-01    4.3192e-05   -9.6355e-05  \n",
      "Epoch: 9950,  Loss u: 3.818e-06, Loss pde: 4.647e-06, Loss_norm: 1.991e+00, Loss tot: 2.837e-05\n",
      "  coefs:  1.2599e-02   -7.4557e-03   -7.9430e-05    9.8496e-01   -9.8553e-01    6.7297e-05   -2.6655e-06  \n"
     ]
    }
   ],
   "source": [
    "params = {'n_epochs':10000,\n",
    "          'alpha_pde_start':1,\n",
    "           'alpha_pde_end':1,\n",
    "           'alpha_l1':1e-5, \n",
    "          'warmup_nsteps':500,\n",
    "          'linearRegInterval':5,\n",
    "          'linearRegression':False,\n",
    "          'width':50,\n",
    "          'layers':8,\n",
    "          'lr':0.002,\n",
    "          'update_coef_in_dl':True,\n",
    "          'logfile':'fixed_alpha_no_lasso.txt'}\n",
    "\n",
    "model = learning(inputs=inputs, u=u, dictionary=dictionary,**params)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "log = pd.read_csv('fixed_alpha_no_lasso.txt')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(-1.0, 1.0)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(log['Epoch'].values, log['p3'].values)\n",
    "plt.plot([0,10000],[1,1])\n",
    "plt.ylim(-1,1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x2b6131f3de90>]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD8CAYAAACb4nSYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3deXxU1f3/8dcnCSQsARIIAcIuS0BQwajgWgE3ulCse93aWmyr/qxdtba2/X77bfVr26+1rVVK3aq2WhFRcam4IS5AQJFAQCBsCYFMEhKyr+f3x1xioAkJmUkmM/N+Ph7zyNw7J/ecOxfec3Pm3HPNOYeIiESHmFA3QEREuo5CX0Qkiij0RUSiiEJfRCSKKPRFRKKIQl9EJIoEHPpmlmxmr5vZVu9n0lHK9jOzXDP7U6D1iojIsQvGmf7twBvOufHAG95ya/4bWBGEOkVEpAOCEfrzgMe8548BX26pkJmdDKQC/w5CnSIi0gFxQdhGqnMu33u+D3+wH8bMYoDfAVcDc462MTNbACwA6NOnz8np6elBaGLo7SysoK7RMX5wXwDySqo4UFHLxCGJ9IjVVysiEhxr164tdM6ltPZ6u0LfzJYDQ1p46c7mC845Z2YtzevwHeBl51yumR21LufcQmAhQEZGhsvMzGxPE7u9M+5+k1NGJ3HfFdMAyD1Qybm/fZsvnjKCX315aohbJyKRwsx2He31doW+c67Vs3Mz229mQ51z+WY2FChoodhM4Cwz+w7QF+hpZuXOuaP1/0eMytp68kqquHLwiKZ1w5N6c2nGCJ5es4dvf24caQN6hbCFIhItgtGv8AJwnff8OmDpkQWcc191zo10zo0GfgA8Hi2BD5BfWg1AWtLhwX7TueMA+NOb27q8TSISnYIR+ncD55nZVvz99XcDmFmGmS0KwvbDnq+sBoDBiQmHrU8b0IvLMkbwj9W7+e1rW0LRNBGJMgGHvnOuyDk32zk33jk3xzlX7K3PdM7d0EL5R51zNwdabzgp8EI/JTH+P177ydxJnDC8P4tW5lBaVdfVTRORKKNhI12g4KC/e2dwC6HfJz6O31w8leq6Rp5Zs6ermyYiUUah3wV8ZTX0jIuhf68eLb5+/LD+nDYmmUff30l9Q2MXt05EoolCvwsUlNWQ0jeeow1X/fqZY8grqWJ59v4ubJmIRBuFfhcoKKtmcL//7Nppbs6kVEYk9+LhlTu7plEiEpUU+l3A553pH01sjHHdzNGs3lnMhtzSLmqZiEQbhX4XKK6oZVALX+Ie6bJTRpAYH8eDK7Z3QatEJBop9DtZQ6Pzh36fnm2W7ZfQg6tnjuKVDfnsKKzogtaJSLRR6HeykspaGh0MbKN755CvnzGGuNgYHnpHZ/siEnwK/U5WVFELwMC+bZ/pg/8CrssyhrN4XS77vOkbRESCRaHfyQrL/VfjDuzTvjN9gBvPPo5GB399N6ezmiUiUUqh38mKyv1n+oPaeaYPMCK5N/NOGsYTH+4ix1fOL1/cyJ7iys5qoohEEYV+Jys6dKbfzj79Q26bM4FG55j1u3d45L2d/PdLmzqjeSISZRT6nayoopYYgwGtTMHQmhHJvbn8lM/m31+x1df0ASIi0lEK/U5WWF5Lcp94YmKOfsewltwyazxpA3rxwwsmUl3XyGMfHPWGOCIibQrGPXLlKIrKa46pP7+51H4JvHf7LAA+3lPC4x/s5Mazx9InXodNRDpGZ/qdrKiitt3DNY/mO587jpLKOh59f2fgjRKRqKXQ72RF5TXHNFyzNdNGJjErfTALV+RwsFo3WxGRjlHod7Ki8uCc6QN877wJlFbV8bd3dwRleyISfQIKfTNLNrPXzWyr9zOplXIjzezfZpZtZpvMbHQg9YaL6roGymrqGXSMwzVbMyWtPxceP4SHV+6gpLI2KNsUkegS6Jn+7cAbzrnxwBveckseB+51zk0CTgUKAqw3LBR7UzAkt2Oytfa67bwJlNfW88c3twVtmyISPQIN/XnAY97zx4AvH1nAzCYDcc651wGcc+XOuai4vPTQDdFbujduR00cksgVp4zg0fd3smVfWdC2KyLRIdDQT3XO5XvP9wGpLZSZAJSY2XNm9pGZ3WtmsQHWGxY+uyF6QlC3+8ML0klMiOOupVk454K6bRGJbG2GvpktN7OsFh7zmpdz/vRpKYHigLOAHwCnAGOB649S3wIzyzSzTJ/Pdyz70u3s9870U9u4VeKxSu7Tkx9eMJFVO4p5/uO8oG5bRCJbm6HvnJvjnJvSwmMpsN/MhgJ4P1vqq88FPnbO5Tjn6oHngelHqW+hcy7DOZeRkpLSsb3qJgoOVhNjxz7vTntcccpIpo0cwC9f3ISvTNMziEj7BNq98wJwnff8OmBpC2XWAAPM7FCCzwKiYvaw/QerSUmMJ7YDUzC0JTbGuPeSE6isbeCupVk8/sFO/vK2brwiIkcXaOjfDZxnZluBOd4yZpZhZosAnHMN+Lt23jCzDYABfw2w3rCw/2BN0Pvzmxs3OJHvzhnPK1n7uGvpRu55dTO5B6LiO3IR6aCAQt85V+Scm+2cG+91AxV76zOdczc0K/e6c+4E59xU59z1zrmoGGReUFYT9P78Iy04ayzDk3o1Lf/qpWx9uSsirdLMXZ2o4GA100YO6NQ64mJjePW7Z7OvtJrl2fu5+5XNLNuQzxdOGNap9YpIeNI0DJ2ktr6RoopaUjuxe+eQvvFxjBvclxvOHMOJw/tz19KNTbdpFBFpTqHfSXzlnTNc82jiYmO499ITKa+u5+dLN3ZZvSISPhT6nWT/oQuzujD0ASakJnLrnPEs25DPsk/ym9a/tbmAW//5EfUNjV3aHhHpXhT6naSzrsZtjxvPHsvUtP7ctTSr6cPnN69ks/TjvSxel9vl7RGR7kOh30nySvxhmzagVxslgy8uNobfX3YiVXUNfPuJtdTWf3Z2/9t/f0p5TX2Xt0lEugeFfifJO1BFrx6xDOh9bDdED5bxqYnce8mJrNtdwl1Ls8jxVXDmuEH4ymr445tbQ9ImEQk9DdnsJHkllaQl9cIs+FfjttfnTxjKJ7ljeWhFDgCXnTKCYQMSeHjlDi7LGMFxKX1D1jYRCQ2d6XeSvJKqkHTtHOmHF0zkxOH9AZg8tB8/vCCdhLhY/uvFTbqISyQKKfQ7yd6SatKSQh/6cbExPPXNGTy9YAbjBvclJTGeW+eM551PfbyRHRX3shGRZhT6naCytp7iitpucaYP0Cc+jtPGDmxavu700Ywf3Jc7lmxoGmUkItFBod8J9pZUAaEZudMePWJj+ONV0yivrufmpz6iTmP3RaKGQr8T5B7wQr8bdO+0Jn1IP+7+ylRW7yzmp0t0By6RaKHRO51gbwjH6B+LeSelsb2gnPvf3MbAvj350YXpANQ3NOLw/0UgIpFF/6s7QV5JJbExRmq/rr8a91jddt4ErjptJA+8vZ1F7/qHdv7XS5s46Zf/pqq2IcStE5Fg05l+J9hTXMXQ/gmdcsesYDMz/nveFEoqa/nVsmwanePxD3YB8IsXNnLPJSeEuIUiEkw60+8Eu4oqGDOoT6ib0W6xMcZ9l09j7tQh/PrlzQAM6ZfA05l7WKobr4tEFIV+J9hZVMmogb1D3Yxj0jMuhvuvmMbF09LolxDHCzefQcaoJG5fvIGPdh84rGzmzmIufuA98rxRSiISPhT6QXagopbSqjpGDwyfM/1D4mJj+P3lJ7H6zjkM7pfAA1dPJyUxnq89uoYt+8qayr2atY91u0u48e+ZNDRq1I9IOAk49M0s2cxeN7Ot3s+kVsr9r5ltNLNsM7vfQjkpTSfaWVQBEJahf0hCj1jAPy30E984jZ6xMVz9t1Vk5ZUCn+1jVt5Bnlq1K2TtFJFjF4wz/duBN5xz44E3vOXDmNnpwBnACcAU4BTgnCDU3e00hf6g8Oreac3Igb158obT6BFjXP7QB7y1pYBtBeXMnTqEM8YN5J5Xt6ibRySMBCP05wGPec8fA77cQhkHJAA9gXigB7A/CHV3OzsLKzGDEcmREfrgn6Z5yU1nMHpQH772yBp2FlUybnAid1/sH9mz4PFMKjRHv0hYCEbopzrnDt2Xbx+QemQB59wHwFtAvvd4zTmX3dLGzGyBmWWaWabP5wtC87rWrqIKhvXvRXxcbKibElSp/RJ45saZnDV+EABThvVjRHJv/njVNLLzD/L1R9dQVl0X4laKSFvaFfpmttzMslp4zGtezvmv5f+Pb/bMbBwwCRgOpAGzzOyslupyzi10zmU45zJSUlKOeYdCbUdRZcR07RypT3wcj3/9VF665UzOm+z/bD934mD+7/KTWLvrAFf9dRX5pZ919VTXNfDnt7Y1zUUkIqHXrtB3zs1xzk1p4bEU2G9mQwG8ny3N1zsf+NA5V+6cKwdeAWYGaye6k11FFWH9JW5bzIwpaf0PuznMvJPSWHjtyWz3lTP3D+/y1hb/P4GVWwu597UtfOlP76n7R6SbCEb3zgvAdd7z64ClLZTZDZxjZnFm1gP/l7gtdu+Es5LKWkoqw3O4ZqBmpafy4i1nktovga89soZfvbSJd7f6u+cKy2u46al11NQfPq1DYXmN/goQ6WLBCP27gfPMbCswx1vGzDLMbJFX5llgO7ABWA+sd869GIS6u5WdRZUAjA6jq3GD6biUvjx/0xlcPWMki1bu4LEPdpExKonfXDyVt7f4+Nbf11Jd91nwf+PRNZx+95t8sL0ohK0WiS4Bh75zrsg5N9s5N97rBir21mc6527wnjc45250zk1yzk12zn0v0Hq7o11NY/Qjs0+/PRJ6xPKrL0/lH9+cwZS0fsyblsaVp47k1/On8tYWH5cv/JCdhRXU1jeyPtc/7v/mp9axr1Q3cxHpCroiN4h2FFZE3HDNjpp53EBeuuUsrpkxCoCrThvJg1dPZ4evnLn3v8svX9wIwG1zJlBV18CCv2dSUll72DZKKmt5dm0ujbrqVyRoFPpBtKPQP1zz0BWtcrgLpwzl1e+ezenHDeLJVbsBuHh6GvdfMY3N+WXMf+D9pqt+AR5akcMP/rWeH/xrvW7yIhIkCv0gyvFVMDYlOvvz22vYgF4sui6Dh6/P4PaL0hme1Is5k1N54obTqKyt58t/fo//enETpVV1TR8Az32Ux11LNx52xv/p/jJue/pjDlTUtlaViLRA8+kHiXOOHF85l2aMCHVTwsKs9FRmpX92Hd+pY5J55dazufe1zTzy/g6e/ziP4opaLssYTlLvnjy0IoeSqjp+d+mJ9IyL4eGVO1jyUR7rc0tYetMZJCb0COHeiIQPnekHSUFZDRW1DTrTD0Byn5785uITePHmM5vuR3DyqCTumDuJ2y9K58X1e7nkwff5aPcBdnkjpXYVVfKdJ9cdNipIRFqn0A+S7b5yAMYO6hviloS/KWn9efZbM3n9trO55GT/X07fOuc4/nzVdPaVVjP/gfdZtaOIa2aM4jfzp7JyWyHzH3j/sOmf//L2dk79n+XsKa78j+0XHNRIIYleCv0gyfH5h2vqTD84zIzxqYmH3XLy8ycM5c0ffI5bZ49nSL8Ezj8+lctOGcGiazMoOFjN3Pvf5fbFn7C3pIp7Xt1MQVkNVy36kG0Fn30YrNxayKm/foOfPr9BXw5LVFLoB8l2Xzm9esQyJAxuhh7O+sbHcdt5E3j/jtmcNd4/N9PsSam8dtvZXDNjFIvX5XL63W8CcOroZKpqG5n/wPss9oZ+rtrhvxDsiQ938z/Lsg8L/tU7itVNJBFPX+QGSY7Pf1/cmDC4GXokGtQ3nl986Xi+ceYYHnh7G6t2FPOnr06jtr6Rm576iO//az2LVu6grqGRiamJzDxuIItW7qC0qo6ffn4yB6vruOyhD+gZF8OqO2aT1KdnqHdJpFMo9IMkp7Cck0a0eNMw6UIjknvzG2+e/0OWfPt0XvxkL39YvpWcwgq+etpIfv7FyfTuGcuD72znjc0FFHtDP2vrG5n/wHv85eqTmTS0H+AfmXXtw6sB+PNXp9MvoQeVtfW8s8XH+ccPOawLSqS7U/dOEFTXNZB7oIqxUTrnTncXE2PMOymN5d87h399ayY/OH8iZsaPLkznhZvPZGpa/6ayi789k7Lqeube/y7ff2Y9m/YeZE9xFe9uLeTdrYVc/tCH5JdWseDxtXz7yXVc/8hqqmrVJSThw7rzl1kZGRkuMzMz1M1o05Z9ZVxw3wr+cMVJzDspLdTNkQ54f1shsTHGaWMHUlJZy5/f2sbjH+yipr6Rof0TyC+t5idz07lv+VZizSjzpoo2gxOGD+DeS05gQmpi0/ZezdrHX97ZzvfPm8DZE1J49L0d9I6P4zJdxyGdzMzWOucyWntd3TtBsKNQwzXD3enjBjU9H9C7J3d+fjI3nzuexetyWbwul77xcXzjzLGcP3kIv1qWzfLs/fzsC5MZ2j+BO5ds4IL7VnDepFS+cvJwzpmQwreeWAvA9Y+s5qZzx/HHN7cB/umkv33OcYfdj2BHYQU19Q2kD/F3J9U1NPLkh7u4YMoQhvbv1YXvgkQDnekHwcIV2/n1y5tZ//Pz6d9LV4ZGg32l1QxOjCcmxigqr+GR93by1OrdFFfU0rtnLJVel8+Fxw/h1Y37DvvdKWn9uGbGKL544jB694xj9O3LALhl1jhunjWOFz7eyw+f/QSAh645mQuOH9L0u6tyihiR3JthA9r+MHDOHfbhItGhrTN9hX4Q/Oz5LF5Yv5f1Pz8/1E2REKpraOSD7UW8sH4vz67N5cazx3L7Rem8mrWPTfkH+e6cCSxem8uilTl8ur+cxPg40ocmsmbngaZtjEzuTZ/4OLLzDzIxNZEt+8s4f3Iq158+mtGD+jQNR73nK1O59OQRTaPF5v1pJcWVtdx3+TROHpXEr17axKKVO/jbdRnMnvQft60+zIGKWvrEx9EzTl/xRQKFfhe47uHVFFXU8NItLd72V+QwzjnW7jrAP1bvYdmGvVTXNbL8e2eTV1LNH9/YSuauA5wzIYWF157MX1fk8NA7OU3fITQ3amBvLjh+CNNHJjV1J8UYXDRlKMs25DeV+9zEFL562ijOnZhCXGwM7271cc3fVrPo2gxmTxrMmDteBuDui6dyWcZnHyRvbykgO7+Mb50zFjPjkfd28MsXN/HPBTOYMXZgF7xT0hEK/S4w63dvMzE1kb9cfXKomyJhpqHRcaCylkF945vW7SisoF9CHAO9ddV1DSzP3s/zH+URHxfL/VdO46VP9rJ4XR4fbC+krsH/f/jOuZMorKjhsfd3Ul3XyFemD2dEci+eWrWbgrIahvRL4PRxA3luXV6r7RkzqA9fPGEop4xJ5pq/+YepThraj+tPH8WPF29oKnfK6CTmTxvO56cO5ZWsfG5/bgNfmT6cG84aQ/qQRL7/zHqy9pZy5akj+dKJwxjYN56/f7iLnz2fxfWnj+aWWeMY2Deef67ezStZ+/jZFyYzbvBn34k98eEufvp8FndclM43zxrLngOV3Pj3tfz12oyj3q+isdHxwvq9TB+ZxMiBvampb6BHTExUXT+j0O9kDY2OST97la+dOZo7LpoU6uZIlDlYXcfKrYXsLania2eMITbGqKip592thZybnkJ8XCx1DY28kV3Ako9yydx5gKKKWtKHJHLDWWP5+4e7WL+nhFU/mc2HOUU8uWo3mTuLaX7fmuNS+rDdm2Zkalp/Lpo6hMVrc9nuq6BnbAy1DY0AxMUY9Y2OvvFxlDf7yyTGIGNUMqt3Fjet6xFrTE3rz7rdJYB/FNSpo5M5bUwyU4cP4JuPf/b/fuygPuQUVjQtnzMhhVnpg5k8rB9jBvVh+ab95JVUMfO4gfwrM5clH/k/1M6ekMKKT30M6N2DeScO4/Rxg5g2cgCDExP45+rd3PXCRq46dSSz0gez7JN8ns7cw0VThnDFqSOZPnIAPWJjeCZzD+dOHMzwpF6YGet2H6C6roFpI5LI2lvKgsczuXbmaOZPS2PUwN7kHqjirP99iydvOI2TRyWR0COWkspa3t9exOxJg4mP899ro76hkdgYw8yorW/k5y9s5DufO44Ryb2prmvAjKayx6pTQ9/MLgV+AUwCTnXOtZjQZnYh8AcgFljknLu7PdsPh9DPK6nijLvf5Nfzp3LVaSND3RyRo3LOsbOokpHJvVu9qKy0qo41O4rJPVDJNTNHE2Owce9B1uwsbjprd86xIa+UZZ/k89CKHM6fnMqvL57Kaxv38dbmApZnF/Dst2aS0COW1zft5/VN+9mUf5Czxg/i51+czDOZubyzxcd2XzkPXn0yG/JKeX3TfjbvO9j0gXPxtDTOnpDCP9fs5sOcYkYN7M2Fxw/htY37mu5H3ZpvnjWG59blUVRRS7+EOKrrG6mt9384Jffp2XQxXvMPraMZ1DeexIQ4dngfPrExRsMRd3Tr36sHpVV1TctxMUb60ESy8g7664qLIX1IIs7BhrxSBvXtSfqQfqzcVtj0O4eGB48Z1IdXbj2rQzdk6uzQnwQ0Ag8BP2gp9M0sFvgUOA/IBdYAVzrnNrW1/XAI/fe3F3LVX1fx5A2ncUazYX8icrjqugbi42IOG1HU2OgO63qprK1n096D5PgqmD1pcFMXV219Iz1irel39xRXst1XTo6vgr0lVdx4znFk5ZXyzcczmTS0Hy/ecibOOQ5U1pHcpyfVdQ1s3FvKR7tLyCmsIL+kimtnjiZjdBKf5Jbyf69/yldnjOTC44eyZmcxG/JKeeLDXVTVNfD/Zo0nO/8gu4srqa5v5NoZo9hRWMGm/INcljGC44f1Y8VWH1l5B1m/p4TC8hp+8aXjycor5ZPcUlZuK2T84L6cMyGFLfvL2F1cya6iSi6ensa2gnI+8e4VfcdF6WTnH2Tt7gPMnzac7503oUPvc5d075jZ27Qe+jOBXzjnLvCW7wBwzv2mre12NPTvWX0Pm4s3H/PvdURBWQ05vnKmjUwiXqMfRCQI0pPT+fGpP+7Q77YV+l2RUmnAnmbLud66FpnZAjPLNLNMn8/X6Y0LVHVdAzFmGu4mImGhzStyzWw5MKSFl+50zi0NdoOccwuBheA/0+/INjr6CdkR33lyLa6ijEcv/FyX1Ski0lFthr5zbk6AdeQBzSccGe6tiwi7iioZObD1IWQiIt1JV/RJrAHGm9kYM+sJXAG80AX1djrnHLuLKhl1lHHDIiLdSUChb2bzzSwXmAksM7PXvPXDzOxlAOdcPXAz8BqQDTzjnNsYWLO7h+KKWspq6hk5UFMqi0h4CGiWTefcEmBJC+v3AnObLb8MvBxIXd3RoQtGdF9cEQkXGnISgByff0rl4zSlsoiECYV+AHJ8FfSMiyEtSXOei0h4UOgHYLuvgtEDW7+cXUSku1HoByCnsFx3yxKRsKLQ76C6hkZ2F1XqS1wRCSsK/Q7aU1xJfaNjbIrO9EUkfCj0OyjHp+GaIhJ+FPodlFOo4ZoiEn4U+h2U46tgYJ+e9O/dI9RNERFpN4V+B+X4KtS1IyJhR6HfQRquKSLhSKHfAaVVdRSW1+pMX0TCjkK/Aw7NuaPhmiISbhT6HbC1wBu5ozN9EQkzCv0O+HRfGfFxMYzSPPoiEmYU+h2wZX8Z41P7aqI1EQk7Cv0O+HR/GRNSE0PdDBGRY6bQP0YllbXsP1jDRIW+iIShQO+Re6mZbTSzRjPLaKXMCDN7y8w2eWVvDaTOUPt0v/9L3AlDFPoiEn4CPdPPAi4GVhylTD3wfefcZGAGcJOZTQ6w3pDZsr8MQGf6IhKWAr0xejaAWetfaDrn8oF873mZmWUDacCmQOoOla37y0iMj2No/4RQN0VE5Jh1aZ++mY0GpgGrjlJmgZllmlmmz+frqqa125Z9/pE7R/ugExHprtoMfTNbbmZZLTzmHUtFZtYXWAx81zl3sLVyzrmFzrkM51xGSkrKsVTRJbYWlGvkjoiErTa7d5xzcwKtxMx64A/8J51zzwW6vVApLK+huKKW8Qp9EQlTnd69Y/5+kL8B2c6533d2fZ3p0336EldEwlugQzbnm1kuMBNYZmaveeuHmdnLXrEzgGuAWWb2sfeYG1CrQ+RTb+TOhFRNtCYi4SnQ0TtLgCUtrN8LzPWerwQi4lvPTfkHSerdg5TE+FA3RUSkQ3RF7jFYv6eUE0cM0MgdEQlbCv12qqipZ2tBGScOHxDqpoiIdJhCv52y8kppdHDSCIW+iIQvhX47rc8tAeCE4f1D3BIRkY5T6LfT+j2ljEjuxcC++hJXRMKXQr+dPt5Tov58EQl7Cv128JXVkFdSpf58EQl7Cv12WLf7AAAnKvRFJMwp9Nvh3a0+eveMVfeOiIQ9hX4bnHO886mP048bSM84vV0iEt6UYm3Y7qtgT3EV50zoftM8i4gcq4Dm3olUb2TvZ9mGfEYk9aa+sRGA2ZNSQ9wqEZHAKfRb8OPFGygsr2laPnHEAIYN6BXCFomIBIdC/willXUUltfwk7npZIxO5sX1e7ny1JGhbpaISFAo9I+wt7QKgOFJvZk+MonpI5NC3CIRkeDRF7lHOFhVB0D/Xj1C3BIRkeBT6B/hYHU9AP0SFPoiEnkU+kcoq/af6ScmqOdLRCJPoPfIvdTMNppZo5lltFE21sw+MrOXAqmzsx3q3umn7h0RiUCBnulnARcDK9pR9lYgO8D6Ol1FbQMAvXvGhrglIiLBF1DoO+eynXNb2ipnZsOBzwOLAqmvK9Q1+C/G6hGrni8RiTxdlWz3AT8CGtsqaGYLzCzTzDJ9Pl/nt+wI9Q2OGIPYGN38XEQiT5uhb2bLzSyrhce89lRgZl8ACpxza9tT3jm30DmX4ZzLSEnp+vlu6hobidNZvohEqDaHqDjn5gRYxxnAl8xsLpAA9DOzJ5xzVwe43U5R3+DoobN8EYlQnX5K65y7wzk33Dk3GrgCeLO7Bj5AfYPO9EUkcgU6ZHO+meUCM4FlZvaat36Ymb0cjAZ2tbpGR49YnemLSGQK6Aok59wSYEkL6/cCc1tY/zbwdiB1drb6hkbiYnSmLyKRSel2hPoGR5zO9EUkQin0j+Dv3tHbIiKRSel2BH/3js70RSQyKfSPUNfgNHpHRCKW0u0I9Y2NGr0jIhErIkjHXQgAAAkRSURBVOcPvuGxNdTUtznjQ4s25JUydlCfILdIRKR7iMjQL6+p73DojxnUhy+cMCzILRIR6R4iMvT/uWBmqJsgItItqU9fRCSKKPRFRKKIQl9EJIoo9EVEoohCX0Qkiij0RUSiiEJfRCSKKPRFRKKIQl9EJIoo9EVEokig98i91Mw2mlmjmWUcpdwAM3vWzDabWbaZaZ4EEZEQCPRMPwu4GFjRRrk/AK8659KBE4HsAOsVEZEOCPTG6NkAZq3PP29m/YGzgeu936kFagOpV0REOqYr+vTHAD7gETP7yMwWmVmrE9ab2QIzyzSzTJ/P1wXNExGJHm2GvpktN7OsFh7z2llHHDAd+ItzbhpQAdzeWmHn3ELnXIZzLiMlJaWdVYiISHu02b3jnJsTYB25QK5zbpW3/CxHCX0REek8nd6945zbB+wxs4neqtnAps6uV0RE/lOgQzbnm1kuMBNYZmaveeuHmdnLzYreAjxpZp8AJwG/DqReERHpmEBH7ywBlrSwfi8wt9nyx0Cr4/hFRKRr6IpcEZEootAXEYkiCn0RkSii0BcRiSIKfRGRKKLQFxGJIgp9EZEootAXEYkiCn0RkSii0BcRiSIKfRGRKKLQFxGJIgp9EZEootAXEYkiCn0RkSii0BcRiSIKfRGRKKLQFxGJIoHeI/dSM9toZo1m1urtEM3sNq9clpn9w8wSAqlXREQ6JtAz/SzgYmBFawXMLA34f0CGc24KEAtcEWC9IiLSAYHeGD0bwMzaU08vM6sDegN7A6lXREQ6ptP79J1zecBvgd1APlDqnPt3Z9crIiL/qc3QN7PlXl/8kY957anAzJKAecAYYBjQx8yuPkr5BWaWaWaZPp+vvfshIiLt0Gb3jnNuToB1zAF2OOd8AGb2HHA68EQr9S0EFgJkZGS4AOsWEZFmumLI5m5ghpn1Nn/n/2wguwvqFRGRIwQ6ZHO+meUCM4FlZvaat36Ymb0M4JxbBTwLrAM2eHUuDKjVIiLSIeZc9+1BycjIcJmZmaFuhohI2DCztc65Vq+b0hW5IiJRRKEvIhJFFPoiIlFEoS8iEkUU+iIiUUShLyISRRT6IiJRRKEvIhJFFPoiIlFEoS8iEkUU+iIiUUShLyISRRT6IiJRRKEvIhJFFPoiIlFEoS8iEkUU+iIiUUShLyISRRT6IiJRJNAbo99rZpvN7BMzW2JmA1opd6GZbTGzbWZ2eyB1iohIxwV6pv86MMU5dwLwKXDHkQXMLBb4M3ARMBm40swmB1iviIh0QECh75z7t3Ou3lv8EBjeQrFTgW3OuRznXC3wT2BeIPWKiEjHxAVxW18Hnm5hfRqwp9lyLnBaaxsxswXAAm+x3My2dLA9g4DCDv5uuNI+R75o21/QPh+rUUd7sc3QN7PlwJAWXrrTObfUK3MnUA882ZEWNuecWwgsDHQ7ZpbpnMsIdDvhRPsc+aJtf0H7HGxthr5zbs7RXjez64EvALOdc66FInnAiGbLw711IiLSxQIdvXMh8CPgS865ylaKrQHGm9kYM+sJXAG8EEi9IiLSMYGO3vkTkAi8bmYfm9mDAGY2zMxeBvC+6L0ZeA3IBp5xzm0MsN72CLiLKAxpnyNftO0vaJ+DylrukRERkUikK3JFRKKIQl9EJIpEXOhH0pQPZjbCzN4ys01mttHMbvXWJ5vZ62a21fuZ5K03M7vf2/dPzGx6s21d55XfambXhWqf2sPMYs3sIzN7yVseY2arvP162hsQgJnFe8vbvNdHN9vGHd76LWZ2QWj2pP3MbICZPetNa5JtZjMj+Tib2W3ev+ksM/uHmSVE4nE2s4fNrMDMspqtC9pxNbOTzWyD9zv3m5m12SjnXMQ8gFhgOzAW6AmsByaHul0B7M9QYLr3PBH/VBeTgf8FbvfW3w7c4z2fC7wCGDADWOWtTwZyvJ9J3vOkUO/fUfb7e8BTwEve8jPAFd7zB4Fve8+/AzzoPb8CeNp7Ptk79vHAGO/fRGyo96uNfX4MuMF73hMYEKnHGf8FmzuAXs2O7/WReJyBs4HpQFazdUE7rsBqr6x5v3tRm20K9ZsS5Dd4JvBas+U7gDtC3a4g7t9S4DxgCzDUWzcU2OI9fwi4sln5Ld7rVwIPNVt/WLnu9MB/HccbwCzgJe8fcyEQd+Qxxj8ibKb3PM4rZ0ce9+bluuMD6O+FoB2xPiKPM59dpZ/sHbeXgAsi9TgDo48I/aAcV++1zc3WH1autUekde+0NOVDWojaElTen7TTgFVAqnMu33tpH5DqPW9t/8PpfbkP/7Ufjd7yQKDEfTbHU/O2N+2X93qpVz6c9hf8Z6k+4BGvW2uRmfUhQo+zcy4P+C2wG8jHf9zWEvnH+ZBgHdc07/mR648q0kI/IplZX2Ax8F3n3MHmrzn/R3xEjLs1sy8ABc65taFuSxeLw98F8Bfn3DSgAv+f/U0i7Dgn4Z90cQwwDOgDXBjSRoVIKI5rpIV+xE35YGY98Af+k86557zV+81sqPf6UKDAW9/a/ofL+3IG8CUz24l/NtZZwB+AAWZ2aMqQ5m1v2i/v9f5AEeGzv4fkArnOuVXe8rP4PwQi9TjPAXY453zOuTrgOfzHPtKP8yHBOq55HD6zcbv2P9JCP6KmfPC+if8bkO2c+32zl14ADn2Dfx3+vv5D66/1RgHMAEq9PyNfA843syTvLOt8b1234py7wzk33Dk3Gv+xe9M591XgLeASr9iR+3vofbjEK++89Vd4oz7GAOPxf+HVLTnn9gF7zGyit2o2sIkIPc74u3VmmFlv79/4of2N6OPcTFCOq/faQTOb4b2P1zbbVutC/SVHJ3xpMhf/KJft+GcCDXmbAtiXM/H/6fcJ8LH3mIu/P/MNYCuwHEj2yhv+G9ZsBzYAGc229XVgm/f4Wqj3rR37/jk+G70zFv9/5m3Av4B4b32Ct7zNe31ss9+/03sfttCOEQ2hfgAnAZnesX4e/yiNiD3OwC+BzUAW8Hf8I3Ai7jgD/8D/vUUd/r/ovhHM4wpkeO/hdvzT4lhbbdI0DCIiUSTSundEROQoFPoiIlFEoS8iEkUU+iIiUUShLyISRRT6IiJRRKEvIhJF/j9YZmhdWu9OeQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(log['Epoch'], log['p4'])\n",
    "plt.plot([0,10000],[1,1])\n",
    "plt.ylim(-2,-0.4)\n",
    "plt.plot([0,10000],[-1,-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1200.0, 3000.0)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.semilogy(log['Epoch'], log['Loss_u'], label='loss_u')\n",
    "plt.semilogy(log['Epoch'], log['Loss_pde'], label='loss_pde')\n",
    "plt.semilogy(log['Epoch'], log['Loss_l1'], label='loss_l1')\n",
    "plt.legend()\n",
    "plt.xlim(1200,3000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "t_mesh = dt*np.arange(nsteps)\n",
    "x_mesh = dx*np.arange(nx)\n",
    "y_mesh = dy*np.arange(ny)\n",
    "t_mesh, x_mesh,y_mesh = np.meshgrid(t_mesh, x_mesh, y_mesh,indexing='ij')\n",
    "t_mesh = np.reshape(t_mesh, (-1,1))\n",
    "x_mesh = np.reshape(x_mesh, (-1,1))\n",
    "y_mesh = np.reshape(y_mesh, (-1,1))\n",
    "x_mesh /= x_max\n",
    "y_mesh /= y_max\n",
    "t_mesh /= t_max\n",
    "prediction = model(torch.cat([torch.tensor(x_mesh), torch.tensor(y_mesh), torch.tensor(t_mesh)],dim=1).double().cuda(0))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}