diff --git a/docs/New Upernavik with Documentation.ipynb b/docs/New Upernavik with Documentation.ipynb
new file mode 100644
index 0000000..43634f6
--- /dev/null
+++ b/docs/New Upernavik with Documentation.ipynb
@@ -0,0 +1,434 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "id": "833e34b3",
+ "metadata": {},
+ "source": [
+ "**Import packages (PINNICLE, os, numpy, datetime, and deepxe)**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "ab2dea95",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "import pinnicle as pinn\n",
+ "import os\n",
+ "import numpy as np\n",
+ "from datetime import datetime\n",
+ "import deepxde as dde"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "cc230466",
+ "metadata": {},
+ "source": [
+ "**Set configuration for deepxde(dde):** default float type (‘float64’), disable xla_jit, set random seed (1234)\n",
+ "
\n",
+ "**deepxde examples:** https://deepxde.readthedocs.io/en/latest/"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "990d62c6",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "dde.config.set_default_float('float64')\n",
+ "dde.config.disable_xla_jit()\n",
+ "dde.config.set_random_seed(1234)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "e707f55d",
+ "metadata": {},
+ "source": [
+ "**Create experiments:** create variable “datestr” and set it as datetime.now().strftime(“%Y%m%d_%H%M%S”)"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "89ace198",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# create experiments\n",
+ "datestr = datetime.now().strftime(\"%Y%m%d_%H%M%S\")"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "7d291506",
+ "metadata": {},
+ "source": [
+ "**Create date file and path:**\n",
+ "
\n",
+ "Create the “inputFileName” varible and set it as “Upernavik.mat”.\n",
+ "
\n",
+ "Create the “expFileName” variable and set it as “Upernavik.exp”."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "6d7bceea",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# data file and path\n",
+ "inputFileName=\"Upernavik_ForSade.mat\"\n",
+ "expFileName = \"Upernavik-Copy1.exp\""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "195d9a25",
+ "metadata": {},
+ "source": [
+ "Create “repoPath” variable and set it as “./examples”.\n",
+ "
\n",
+ "Create “appDataPath” variable and set the path “os.path.join(repoPath, “dataset”).\n",
+ "
\n",
+ "Create “data_path” variable and set the path “os.path.join(appDataPath, inputFileName)”."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "ba969078",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "#inputFileName=\"Helheim_fastflow.mat\"\n",
+ "#expFileName = \"Helheim_Big.exp\"\n",
+ "repoPath = \"./examples/\"\n",
+ "appDataPath = os.path.join(repoPath, \"dataset\")\n",
+ "data_path = os.path.join(appDataPath, inputFileName)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "a030a0e6",
+ "metadata": {},
+ "source": [
+ "Check data path"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "5b545ee0",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "data_path"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "b9ddced6",
+ "metadata": {},
+ "source": [
+ "**Create a path for saving results and figures:**\n",
+ "
\n",
+ "Create “ouputFileName” variable and set this as “Upernavik_test”.\n",
+ "
\n",
+ "Create the variable for folder “modelFolder” and set the folder “./Models/” + outputFileName + “ “ + datestr + “/””."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "e5228d36",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# path for saving results and figures\n",
+ "outputFileName=\"Upernavik_test\"\n",
+ "#outputFileName=\"Helheim_test\"\n",
+ "modelFolder = \"./Models/\" + outputFileName + \"_\" + datestr + \"/\""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "56a86e6e",
+ "metadata": {},
+ "source": [
+ "**Set the general parameters:**\n",
+ "
\n",
+ "Create hyperparameters variable, “hp” and set this as { } \n",
+ "
\n",
+ "Set “hp[“epochs”] = 1000” Set “hp[“learning_rate”] = 0.001” \n",
+ "
\n",
+ "Set “hp[“save_path”] = modelFolder” \n",
+ "
\n",
+ "Set “hp[“is_save”] = True” \n",
+ "
\n",
+ "Set “hp[“is_plot”] = True”"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "397d17ee",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# General parameters\n",
+ "hp = {}\n",
+ "hp[\"epochs\"] = 1000\n",
+ "hp[\"learning_rate\"] = 0.001\n",
+ "hp[\"loss_function\"] = \"MSE\"\n",
+ "hp[\"save_path\"] = modelFolder\n",
+ "hp[\"is_save\"] = True\n",
+ "hp[\"is_plot\"] = True"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "97b92bf2",
+ "metadata": {},
+ "source": [
+ "**Set Neural Networks:**\n",
+ "
\n",
+ "Set “hp[“activation”] = “tanh”” \n",
+ "
\n",
+ "Set “hp[“initializer”] = “Glorot uniform”” \n",
+ "
\n",
+ "Set the number of neurons: “hp[“num_neurons”] = 20” \n",
+ "
\n",
+ "Set the number of layers” “hp[“num_layers”] = 6”"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "941b7552",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# NN\n",
+ "hp[\"activation\"] = \"tanh\"\n",
+ "hp[\"initializer\"] = \"Glorot uniform\"\n",
+ "hp[\"num_neurons\"] = 20\n",
+ "hp[\"num_layers\"] = 6"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "5be86553",
+ "metadata": {},
+ "source": []
+ },
+ {
+ "cell_type": "markdown",
+ "id": "7743f436",
+ "metadata": {},
+ "source": [
+ "**Set the Data:**\n",
+ "
\n",
+ "Set issm, “issm = { }”\n",
+ "
\n",
+ "Set the data size for each of the issm data variables, “issm[“data_size”] = {“u”:1000, “v”:1000, “s”:1000, “H”:1000, “C”:None}”\n",
+ "
\n",
+ "Set the data path, “issm[“data_path”] = data_path”\n",
+ "
\n",
+ "Set “hp[“data”] = {“ISSM”:issm}”"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "d2658e70",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# data\n",
+ "issm = {}\n",
+ "# issm[\"data_size\"] = {\"u\":1000, \"v\":1000, \"s\":1000, \"H\":1000, \"C\":None, \"vel\":1000}\n",
+ "issm[\"data_size\"] = {\"u\":1000, \"v\":1000, \"s\":1000, \"H\":1000, \"C\":None}\n",
+ "issm[\"data_path\"] = data_path\n",
+ "hp[\"data\"] = {\"ISSM\":issm}"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "0feba1b6",
+ "metadata": {},
+ "source": [
+ "**Create the domain:**\n",
+ "
\n",
+ "Set the shapefile path, “hp[“shapefile”] = os.path.join(repoPath, “dataset”, expFileName)”\n",
+ "
\n",
+ "Set the collocation points, “hp[“num_collocation_points”] = 5000”\n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "7b2361f2",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# domain\n",
+ "hp[\"shapefile\"] = os.path.join(repoPath, \"dataset\", expFileName)\n",
+ "hp[\"num_collocation_points\"] = 5000"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "c435bc72",
+ "metadata": {},
+ "source": [
+ "**Set the additional loss function:**\n",
+ "
\n",
+ "Create the variable for velocity loss function: “vel_loss = { }”\n",
+ "
\n",
+ "Set “vel_loss[‘name’] = “vel log””\n",
+ "
\n",
+ "Set “vel_loss[‘function’] = “VEL_LOG””\n",
+ "
\n",
+ "Set “vel_loss[‘weight’] = 1.0e-5”\n",
+ "
\n",
+ "Set “hp[“additional_loss”] = {“vel”:vel_loss}”"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "08bc9807",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# additional loss function\n",
+ "# vel_loss = {}\n",
+ "# vel_loss['name'] = \"vel log\"\n",
+ "# vel_loss['function'] = \"VEL_LOG\"\n",
+ "# vel_loss['weight'] = 1.0e-5\n",
+ "# hp[\"additional_loss\"] = {\"vel\":vel_loss}"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "7b7eb349",
+ "metadata": {},
+ "source": [
+ "**Add the physics components:**\n",
+ "
\n",
+ "Create the variable for SSA, “SSA = { }”\n",
+ "
\n",
+ "Set “SSA[“scalar_variables”] = {“B”:1.26802073401e+08}”"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "16ad20cd",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# physics\n",
+ "SSA = {}\n",
+ "SSA[\"scalar_variables\"] = {\"B\":1.26802073401e+08}\n",
+ "hp[\"equations\"] = {\"SSA\":SSA}"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "1595c0f3",
+ "metadata": {},
+ "source": [
+ "**Create the experiment:**\n",
+ "
\n",
+ "Create the experiment variable and set it, “experiment = pinn.PINN(hp)”\n",
+ "
\n",
+ "Print out the experiment parameters, “print(experiment.params)”\n",
+ "
\n",
+ "Compile the experiment, “experiement.compile( )”"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "f99f4d25",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# create experiment\n",
+ "experiment = pinn.PINN(hp)\n",
+ "print(experiment.params)\n",
+ "experiment.compile()"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "1e6ee2d3",
+ "metadata": {},
+ "source": [
+ "**Train the experiment\n",
+ "“experiment.train( )”**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "a917c1be",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# Train\n",
+ "experiment.train()"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "70673e1a",
+ "metadata": {},
+ "source": [
+ "**Show the results:**\n",
+ "
\n",
+ "“experiment.plot_predictions(X_ref=experiments.model_data.data[“ISSM”].X_dict, sol_ref=experiment.model_data.data[“ISSM”].data_dict)”"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "6e2bde9c",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# show results\n",
+ "experiment.plot_predictions(X_ref=experiment.model_data.data[\"ISSM\"].X_dict, \n",
+ " sol_ref=experiment.model_data.data[\"ISSM\"].data_dict)"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3 (ipykernel)",
+ "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": "3.8.8"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/docs/examples/PINN.MC.tutorial.ipynb b/docs/examples/PINN.MC.tutorial.ipynb
new file mode 100644
index 0000000..9128f63
--- /dev/null
+++ b/docs/examples/PINN.MC.tutorial.ipynb
@@ -0,0 +1,655 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "id": "85baec1d",
+ "metadata": {},
+ "source": [
+ "# A PINN example to estimate sliding from surface observations"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "a3017550",
+ "metadata": {},
+ "source": [
+ "## Set-Up of the Problem"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "3b1f1fbd",
+ "metadata": {},
+ "source": [
+ "The problem: We have observations of surface heights, surface speed (from satellites), and estimates of ice thickness, but we don't know how fast ice is sliding on its bed. We can use physics: momentum balance (i.e., stress balance equations) to infer basal sliding from these surface observations.\n",
+ "\n",
+ "This document is a tutorial for the implementation of a PINN solver for infering basal friction coefficients using Shelfy-Stream Approximation (SSA): "
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "4c4cc9bc",
+ "metadata": {},
+ "source": [
+ "***We use Shelfy-Stream Approximation (SSA): \n",
+ "
\n",
+ "$$\\nabla \\cdot \\sigma _{SSA} + \\tau_b = \\rho_igH\\nabla s,\\; in\\; \\Omega$$\n",
+ "
\n",
+ "where $\\tau_b$ is the basal shear stress, determined by a friction law:\n",
+ "
\n",
+ "$$\\tau_b = −C^2|u|^{m − 1}u$$\n",
+ "
\n",
+ "with the a calving front boundary condition:\n",
+ "
\n",
+ "$$ \\sigma\\cdot n = (\\bar{p}_i -\\bar{p}_w)n on \\; \\Gamma$$\n",
+ "
\n",
+ "We can infer the basal friction coefficient ($C$) from known observations of surface velocity ($\\hat{u}$), ice thickness ($\\hat{h}$), and surface elevation ($\\hat{s}$) using the PINN while learning its observational data***"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "7a091f07",
+ "metadata": {},
+ "source": [
+ "[click for deepxde examples](https://deepxde.readthedocs.io/en/latest/)\n",
+ "\n",
+ "DeepXDE is a library for scientific machine and physics-informed learning.\n",
+ "
\n",
+ "DeepXDE includes algorithms for:\n",
+ "
\n",
+ "$\\star$ Solving different problems (such as: ordinary/partial differential equations, integro-differential equations,etc.)\n",
+ "
\n",
+ "$\\star$ Improving PINN accuracy (such as: residual-based adaptive sampling, gradient-end PINN, PINN with multi-scale Fourier features, etc.)\n",
+ "
\n",
+ "$\\star$ Physics-informed deep operator network (DeepONet)(such as: DeepONet: learning operators, MIONet: learning multiple-input operators, reliable extrapolation, etc.) \n",
+ "
\n",
+ "$\\star$ Multifidelity Neural Network (MFNN) (such as learning from multifidelity data)\n"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "02c131b2",
+ "metadata": {},
+ "source": [
+ "## PINN Solver Code"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "3b7a7d4d",
+ "metadata": {},
+ "source": [
+ "**Import the necessary packages:**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 1,
+ "id": "7047e456",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "2024-05-26 23:00:21.526258: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 AVX512F AVX512_VNNI FMA\n",
+ "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
+ "2024-05-26 23:00:21.646901: I tensorflow/core/util/port.cc:104] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
+ "Using backend: tensorflow\n",
+ "Other supported backends: tensorflow.compat.v1, pytorch, jax, paddle.\n",
+ "paddle supports more examples now and is recommended.\n",
+ "Enable just-in-time compilation with XLA.\n",
+ "\n"
+ ]
+ }
+ ],
+ "source": [
+ "%reset -f\n",
+ "import pinnicle as pinn\n",
+ "import os\n",
+ "import numpy as np\n",
+ "from datetime import datetime\n",
+ "import deepxde as dde"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "636ed52b",
+ "metadata": {},
+ "source": [
+ "**Let's set configuration for DeepXDE(dde):**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 2,
+ "id": "fe81b291",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "Disable just-in-time compilation with XLA.\n",
+ "\n"
+ ]
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Set the default float type to float64\n"
+ ]
+ }
+ ],
+ "source": [
+ "dde.config.set_default_float('float64')\n",
+ "dde.config.disable_xla_jit()\n",
+ "dde.config.set_random_seed(1234)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "67bd6566",
+ "metadata": {},
+ "source": [
+ "**Create the Experiment**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 3,
+ "id": "a722589f",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# create experiments\n",
+ "datestr = datetime.now().strftime(\"%Y%m%d_%H%M%S\")"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "eafc518e",
+ "metadata": {},
+ "source": [
+ "**Create data file and path:**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 4,
+ "id": "9d109522",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# data file and path\n",
+ "inputFileName=\"Upernavik_ForSade.mat\"\n",
+ "expFileName = \"Upernavik-Copy1.exp\""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 5,
+ "id": "5aa2b50a",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "./examples/dataset/Upernavik_ForSade.mat\n"
+ ]
+ }
+ ],
+ "source": [
+ "#inputFileName=\"Helheim_fastflow.mat\"\n",
+ "#expFileName = \"Helheim_Big.exp\"\n",
+ "repoPath = \"./examples/\"\n",
+ "appDataPath = os.path.join(repoPath, \"dataset\")\n",
+ "data_path = os.path.join(appDataPath, inputFileName)\n",
+ "\n",
+ "# check\n",
+ "print(data_path)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "02302eac",
+ "metadata": {},
+ "source": [
+ "**Create a path for saving results and figures:**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 6,
+ "id": "f8741a9e",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# path for saving results and figures\n",
+ "outputFileName=\"Upernavik_test\"\n",
+ "#outputFileName=\"Helheim_test\"\n",
+ "modelFolder = \"./Models/\" + outputFileName + \"_\" + datestr + \"/\""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "5073f50d",
+ "metadata": {},
+ "source": [
+ "**Set the general parameters for training:**\n",
+ "
\n",
+ "PINNICLE has hyperparameters stored in a dictionary, ```hp```, with specific names. PINNICLE loads the parameters from ```hp``` according to the corresponding keys. \n",
+ "
\n",
+ "[Click here for more details on PINN's default and required parameters](https://pinnicle.readthedocs.io/en/latest/api/pinnicle.html#pinnicle.parameter.DataParameter)\n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 7,
+ "id": "caa2ea31",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# General parameters\n",
+ "hp = {}\n",
+ "hp[\"epochs\"] = 1000000\n",
+ "hp[\"learning_rate\"] = 0.001\n",
+ "hp[\"loss_function\"] = \"MSE\"\n",
+ "hp[\"save_path\"] = modelFolder\n",
+ "hp[\"is_save\"] = True\n",
+ "hp[\"is_plot\"] = True"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "7d70b7b7",
+ "metadata": {},
+ "source": [
+ "**Set Neural Network's Architecture within ```hp```:**\n",
+ "
\n",
+ " We are going to use Hyperbolic Tangent functions for activation functions, we will use 6 layers with 20 neurons each and a Glorot uniform initializer."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 8,
+ "id": "3fa4830b",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# NN\n",
+ "hp[\"activation\"] = \"tanh\"\n",
+ "hp[\"initializer\"] = \"Glorot uniform\"\n",
+ "hp[\"num_neurons\"] = 20\n",
+ "hp[\"num_layers\"] = 6"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "f7786fdb",
+ "metadata": {},
+ "source": [
+ "**Create the domain of the computation:**\n",
+ "
\n",
+ "We define the domain of computation, and number of collocation points used to evaluate the PDEs residual."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 9,
+ "id": "1ac51d3e",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# domain\n",
+ "hp[\"shapefile\"] = os.path.join(repoPath, \"dataset\", expFileName)\n",
+ "hp[\"num_collocation_points\"] = 5000"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "a4b55e77",
+ "metadata": {},
+ "source": [
+ "**Add the physics components, ``SSA``, to the PINN:**\n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 10,
+ "id": "3ea03812",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# physics\n",
+ "MC = {}\n",
+ "MC[\"scalar_variables\"] = {\"B\":1.26802073401e+08}\n",
+ "hp[\"equations\"] = {\"MC\":MC}"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "edc7ae26",
+ "metadata": {},
+ "source": [
+ "**There are default parameter settings in ```SSA``` such as:\n",
+ "
**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 11,
+ "id": "9863b3bd",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "def set_default(self):\n",
+ " self.input = ['x', 'y']\n",
+ " self.output = ['u', 'v', 's', 'H', 'C']\n",
+ " self.output_lb = [-1.0e4/self.yts, -1.0e4/self.yts, -1.0e3, 10.0, 0.01]\n",
+ " self.output_ub = [ 1.0e4/self.yts, 1.0e4/self.yts, 2.5e3, 2000.0, 1.0e4]\n",
+ " self.data_weights = [1.0e-8*self.yts**2.0, 1.0e-8*self.yts**2.0, 1.0e-6, 1.0e-6, 1.0e-8]\n",
+ " self.residuals = [\"fSSA1\", \"fSSA2\"]\n",
+ " self.pde_weights = [1.0e-10, 1.0e-10]\n",
+ "\n",
+ " # scalar variables: name:value\n",
+ " self.scalar_variables = {\n",
+ " 'n': 3.0, # exponent of Glen's flow law\n",
+ " 'B':1.26802073401e+08 # -8 degree C, cuffey\n",
+ " }"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "513010b0",
+ "metadata": {},
+ "source": [
+ "These default parameters are the names of the input and output variables in the PINN, scaling factors, weights in the loss function, and names of the residual and weights. \n",
+ "
\n",
+ "
\n",
+ "We can set the input and output names as any other names. But, the orders of the input and output lists matters. These input and output lists will be used to load the data in the next section:"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "93e204e7",
+ "metadata": {},
+ "source": [
+ "**Set the Data used for training in ```issm``` dictionary:**\n",
+ "
\n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 12,
+ "id": "b3e4e528",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# data\n",
+ "issm = {}\n",
+ "issm[\"data_size\"] = {\"u\":1000, \"v\":1000, \"a\":1000, \"H\":None}\n",
+ "issm[\"data_path\"] = data_path\n",
+ "hp[\"data\"] = {\"ISSM\":issm}"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "55afd41f",
+ "metadata": {},
+ "source": [
+ "**In ```data_size``` the variables are defined by pairs of ```key``` = ```value```. If keys are not redefined in ```name_map``` then they will be set as the default or what they are set as in the physics section.** \n",
+ "
\n",
+ "The value associated with each key gives the number of data points used for the training. When ```value``` is set as **None**, then only Dirichlet boundary condition around the domain boundary will be used for the corresponding ```key```. If the variable is included in the training, but not given in ```data_size```, then there will be no data used in the training for this variable. "
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "13e00c56",
+ "metadata": {},
+ "source": [
+ "**Create the Experiment and Compile:** Now, let's run the PINN model.\n",
+ "
"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 13,
+ "id": "ab00ea28",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "ERROR:root:ERROR: MATLAB type not supported: bamggeom, (uint32)\n",
+ "ERROR:root:ERROR: MATLAB type not supported: bamgmesh, (uint32)\n",
+ "ERROR:root:ERROR: MATLAB type not supported: qmustatistics, (uint32)\n",
+ "ERROR:root:ERROR: MATLAB type not supported: lovenumbers, (uint32)\n",
+ "ERROR:root:ERROR: MATLAB type not supported: rotational, (uint32)\n",
+ "ERROR:root:ERROR: MATLAB type not supported: solidearthsettings, (uint32)\n",
+ "2024-05-26 23:00:25.911779: W tensorflow/core/common_runtime/gpu/gpu_bfc_allocator.cc:42] Overriding orig_value setting because the TF_FORCE_GPU_ALLOW_GROWTH environment variable is set. Original config value was 0.\n",
+ "2024-05-26 23:00:25.911832: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1613] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 3445 MB memory: -> device: 0, name: Tesla V100-SXM2-32GB, pci bus id: 0000:86:00.0, compute capability: 7.0\n"
+ ]
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "add input transform with [ -323028.74554883 -1870413.49271111] and [ -230318.78353497 -1740472.47370953]\n",
+ "add output transform with [-3.17097920e-04 -3.17097920e-04 -1.58548960e-07 1.62696374e-01] and [3.1709792e-04 3.1709792e-04 1.5854896e-07 2.5000000e+03]\n",
+ "Parameters: \n",
+ "\tTrainingParameter: \n",
+ "\t\tepochs:\t1000000\n",
+ "\t\toptimizer:\tadam\n",
+ "\t\tloss_functions:\t['MSE', 'MSE', 'MSE', 'MSE', 'MSE']\n",
+ "\t\tadditional_loss:\t{}\n",
+ "\t\tlearning_rate:\t0.001\n",
+ "\t\tloss_weights:\t[1000000.0, 31536.0, 31536.0, 315360000000.0, 1e-06]\n",
+ "\t\thas_callbacks:\tFalse\n",
+ "\t\tmin_delta:\tNone\n",
+ "\t\tpatience:\tNone\n",
+ "\t\tperiod:\tNone\n",
+ "\t\tcheckpoint:\tFalse\n",
+ "\t\tsave_path:\t./Models/Upernavik_test_20240526_230023/\n",
+ "\t\tis_save:\tTrue\n",
+ "\t\tis_plot:\tTrue\n",
+ "\tDomainParameter: \n",
+ "\t\tshapefile:\t./examples/dataset/Upernavik-Copy1.exp\n",
+ "\t\tnum_collocation_points:\t5000\n",
+ "\tDataParameter: \n",
+ "\t\tISSM:\n",
+ "\t\t\tdata_path:\t./examples/dataset/Upernavik_ForSade.mat\n",
+ "\t\t\tdata_size:\t{'u': 1000, 'v': 1000, 'a': 1000, 'H': None}\n",
+ "\t\t\tname_map:\t{'u': 'u', 'v': 'v', 'a': 'a', 'H': 'H'}\n",
+ "\t\t\tsource:\tISSM\n",
+ "\n",
+ "\tNNParameter: \n",
+ "\t\tinput_variables:\t['x', 'y']\n",
+ "\t\toutput_variables:\t['u', 'v', 'a', 'H']\n",
+ "\t\tnum_neurons:\t20\n",
+ "\t\tnum_layers:\t6\n",
+ "\t\tactivation:\ttanh\n",
+ "\t\tinitializer:\tGlorot uniform\n",
+ "\t\tinput_lb:\t[ -323028.74554883 -1870413.49271111]\n",
+ "\t\tinput_ub:\t[ -230318.78353497 -1740472.47370953]\n",
+ "\t\toutput_lb:\t[-3.17097920e-04 -3.17097920e-04 -1.58548960e-07 1.62696374e-01]\n",
+ "\t\toutput_ub:\t[3.1709792e-04 3.1709792e-04 1.5854896e-07 2.5000000e+03]\n",
+ "\t\tinput_size:\t2\n",
+ "\t\toutput_size:\t4\n",
+ "\tPhysicsParameter: \n",
+ "\t\tMC:\n",
+ "\t\t\trhoi:\t917.0\n",
+ "\t\t\trhow:\t1023.0\n",
+ "\t\t\tg:\t9.81\n",
+ "\t\t\tyts:\t31536000.0\n",
+ "\t\t\tinput:\t['x', 'y']\n",
+ "\t\t\toutput:\t['u', 'v', 'a', 'H']\n",
+ "\t\t\toutput_lb:\t[-0.00031709791983764585, -0.00031709791983764585, -1.5854895991882294e-07, 10.0]\n",
+ "\t\t\toutput_ub:\t[0.00031709791983764585, 0.00031709791983764585, 1.5854895991882294e-07, 2500.0]\n",
+ "\t\t\tdata_weights:\t[31536.0, 31536.0, 315360000000.0, 1e-06]\n",
+ "\t\t\tresiduals:\t['fMC']\n",
+ "\t\t\tpde_weights:\t[1000000.0]\n",
+ "\t\t\tscalar_variables:\t{'n': 3.0, 'B': 126802073.401}\n",
+ "\n",
+ "\n",
+ "Compiling model...\n",
+ "'compile' took 0.004787 s\n",
+ "\n"
+ ]
+ }
+ ],
+ "source": [
+ "# create experiment\n",
+ "experiment = pinn.PINN(hp)\n",
+ "print(experiment.params)\n",
+ "experiment.compile()"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "cca0d47a",
+ "metadata": {},
+ "source": [
+ "**Train the experiment:**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "04f636bc",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Training model...\n",
+ "\n"
+ ]
+ },
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "/opt/conda/lib/python3.8/site-packages/keras/initializers/initializers_v2.py:120: UserWarning: The initializer GlorotUniform is unseeded and being called multiple times, which will return identical values each time (even if the initializer is unseeded). Please update your code to provide a seed to the initializer, or avoid using the same initalizer instance more than once.\n",
+ " warnings.warn(\n"
+ ]
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "WARNING:tensorflow:From /opt/conda/lib/python3.8/site-packages/tensorflow/python/autograph/pyct/static_analysis/liveness.py:83: Analyzer.lamba_check (from tensorflow.python.autograph.pyct.static_analysis.liveness) is deprecated and will be removed after 2023-09-23.\n",
+ "Instructions for updating:\n",
+ "Lambda fuctions will be no more assumed to be used in the statement where they are used, or at least in the same block. https://github.com/tensorflow/tensorflow/issues/56089\n"
+ ]
+ },
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "WARNING:tensorflow:From /opt/conda/lib/python3.8/site-packages/tensorflow/python/autograph/pyct/static_analysis/liveness.py:83: Analyzer.lamba_check (from tensorflow.python.autograph.pyct.static_analysis.liveness) is deprecated and will be removed after 2023-09-23.\n",
+ "Instructions for updating:\n",
+ "Lambda fuctions will be no more assumed to be used in the statement where they are used, or at least in the same block. https://github.com/tensorflow/tensorflow/issues/56089\n"
+ ]
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Step Train loss Test loss Test metric\n",
+ "0 [2.53e-07, 1.16e-04, 1.81e-05, 4.39e-04, 5.87e-01] [2.53e-07, 1.16e-04, 1.81e-05, 4.39e-04, 5.87e-01] [] \n",
+ "10000 [2.47e-07, 2.08e-06, 8.22e-07, 3.51e-05, 5.49e-03] [2.47e-07, 2.08e-06, 8.22e-07, 3.51e-05, 5.49e-03] [] \n",
+ "20000 [2.63e-07, 2.26e-06, 8.22e-07, 3.32e-05, 4.00e-03] [2.63e-07, 2.26e-06, 8.22e-07, 3.32e-05, 4.00e-03] [] \n",
+ "30000 [3.22e-07, 2.31e-06, 8.15e-07, 3.16e-05, 2.47e-03] [3.22e-07, 2.31e-06, 8.15e-07, 3.16e-05, 2.47e-03] [] \n"
+ ]
+ }
+ ],
+ "source": [
+ "# Train\n",
+ "experiment.train()"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "d4a588ad",
+ "metadata": {},
+ "source": [
+ "#### The results should look like this: \n",
+ ""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "912568b6",
+ "metadata": {},
+ "source": [
+ "**Show the results with ```experiment.plot_predictions()```:**\n",
+ "
\n"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "4656f20f",
+ "metadata": {},
+ "source": [
+ "### The results should look like: \n",
+ ""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "1ef29f24",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# show results\n",
+ "experiment.plot_predictions(X_ref=experiment.model_data.data[\"ISSM\"].X_dict, \n",
+ " sol_ref=experiment.model_data.data[\"ISSM\"].data_dict)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "e6db727b",
+ "metadata": {},
+ "source": [
+ "**Use ```plot_similarity``` to plot the difference between predicted and refrence values for $H$**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "a0b3ab3b",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "pinn.utils.plotting.plot_similarity(pinn=experiment,feature_name = 'H', cmap='terrain')"
+ ]
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3 (ipykernel)",
+ "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": "3.8.8"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/docs/examples/PINN.NewUpernavik.tutorial.ipynb b/docs/examples/PINN.NewUpernavik.tutorial.ipynb
new file mode 100644
index 0000000..9b337e5
--- /dev/null
+++ b/docs/examples/PINN.NewUpernavik.tutorial.ipynb
@@ -0,0 +1,690 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "id": "98b4171a",
+ "metadata": {},
+ "source": [
+ "# A PINN example to estimate sliding from surface observations"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "561d5c8b",
+ "metadata": {},
+ "source": [
+ "## Set-Up of the Problem"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "f2593391",
+ "metadata": {},
+ "source": [
+ "The problem: We have observations of surface heights, surface speed (from satellites), and estimates of ice thickness, but we don't know how fast ice is sliding on its bed. We can use physics: momentum balance (i.e., stress balance equations) to infer basal sliding from these surface observations.\n",
+ "\n",
+ "This document is a tutorial for the implementation of a PINN solver for infering basal friction coefficients using Shelfy-Stream Approximation (SSA): "
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "5db37325",
+ "metadata": {},
+ "source": [
+ "***We use Shelfy-Stream Approximation (SSA): \n",
+ "
\n",
+ "$$\\nabla \\cdot \\sigma _{SSA} + \\tau_b = \\rho_igH\\nabla s,\\; in\\; \\Omega$$\n",
+ "
\n",
+ "where $\\tau_b$ is the basal shear stress, determined by a friction law:\n",
+ "
\n",
+ "$$\\tau_b = −C^2|u|^{m − 1}u$$\n",
+ "
\n",
+ "with the a calving front boundary condition:\n",
+ "
\n",
+ "$$ \\sigma\\cdot n = (\\bar{p}_i -\\bar{p}_w)n on \\; \\Gamma$$\n",
+ "
\n",
+ "We can infer the basal friction coefficient ($C$) from known observations of surface velocity ($\\hat{u}$), ice thickness ($\\hat{h}$), and surface elevation ($\\hat{s}$) using the PINN while learning its observational data***"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "be8a3025",
+ "metadata": {},
+ "source": [
+ "[click for deepxde examples](https://deepxde.readthedocs.io/en/latest/)\n",
+ "\n",
+ "DeepXDE is a library for scientific machine and physics-informed learning.\n",
+ "
\n",
+ "DeepXDE includes algorithms for:\n",
+ "
\n",
+ "$\\star$ Solving different problems (such as: ordinary/partial differential equations, integro-differential equations,etc.)\n",
+ "
\n",
+ "$\\star$ Improving PINN accuracy (such as: residual-based adaptive sampling, gradient-end PINN, PINN with multi-scale Fourier features, etc.)\n",
+ "
\n",
+ "$\\star$ Physics-informed deep operator network (DeepONet)(such as: DeepONet: learning operators, MIONet: learning multiple-input operators, reliable extrapolation, etc.) \n",
+ "
\n",
+ "$\\star$ Multifidelity Neural Network (MFNN) (such as learning from multifidelity data)\n"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "8a395c79",
+ "metadata": {},
+ "source": [
+ "## PINN Solver Code"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "26a85514",
+ "metadata": {},
+ "source": [
+ "**Import the necessary packages:**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 1,
+ "id": "438b0c81",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "2024-05-26 22:59:57.218490: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 AVX512F AVX512_VNNI FMA\n",
+ "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
+ "2024-05-26 22:59:57.343437: I tensorflow/core/util/port.cc:104] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
+ "Using backend: tensorflow\n",
+ "Other supported backends: tensorflow.compat.v1, pytorch, jax, paddle.\n",
+ "paddle supports more examples now and is recommended.\n",
+ "Enable just-in-time compilation with XLA.\n",
+ "\n"
+ ]
+ }
+ ],
+ "source": [
+ "%reset -f\n",
+ "import pinnicle as pinn\n",
+ "import os\n",
+ "import numpy as np\n",
+ "from datetime import datetime\n",
+ "import deepxde as dde"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "be837ede",
+ "metadata": {},
+ "source": [
+ "**Let's set configuration for DeepXDE(dde):**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 2,
+ "id": "e854776f",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "Disable just-in-time compilation with XLA.\n",
+ "\n"
+ ]
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Set the default float type to float64\n"
+ ]
+ }
+ ],
+ "source": [
+ "dde.config.set_default_float('float64')\n",
+ "dde.config.disable_xla_jit()\n",
+ "dde.config.set_random_seed(1234)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "d3c2022a",
+ "metadata": {},
+ "source": [
+ "**Create the Experiment**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 3,
+ "id": "737ddf3a",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# create experiments\n",
+ "datestr = datetime.now().strftime(\"%Y%m%d_%H%M%S\")"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "49b8d76c",
+ "metadata": {},
+ "source": [
+ "**Create data file and path:**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 4,
+ "id": "74f9660a",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# data file and path\n",
+ "inputFileName=\"Upernavik_ForSade.mat\"\n",
+ "expFileName = \"Upernavik-Copy1.exp\""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 5,
+ "id": "5100f564",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "./examples/dataset/Upernavik_ForSade.mat\n"
+ ]
+ }
+ ],
+ "source": [
+ "#inputFileName=\"Helheim_fastflow.mat\"\n",
+ "#expFileName = \"Helheim_Big.exp\"\n",
+ "repoPath = \"./examples/\"\n",
+ "appDataPath = os.path.join(repoPath, \"dataset\")\n",
+ "data_path = os.path.join(appDataPath, inputFileName)\n",
+ "\n",
+ "# check\n",
+ "print(data_path)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "73575911",
+ "metadata": {},
+ "source": [
+ "**Create a path for saving results and figures:**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 6,
+ "id": "d949d64c",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# path for saving results and figures\n",
+ "outputFileName=\"Upernavik_test\"\n",
+ "#outputFileName=\"Helheim_test\"\n",
+ "modelFolder = \"./Models/\" + outputFileName + \"_\" + datestr + \"/\""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "4144be5f",
+ "metadata": {},
+ "source": [
+ "**Set the general parameters for training:**\n",
+ "
\n",
+ "PINNICLE has hyperparameters stored in a dictionary, ```hp```, with specific names. PINNICLE loads the parameters from ```hp``` according to the corresponding keys. \n",
+ "
\n",
+ "[Click here for more details on PINN's default and required parameters](https://pinnicle.readthedocs.io/en/latest/api/pinnicle.html#pinnicle.parameter.DataParameter)\n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 7,
+ "id": "9811ed70",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# General parameters\n",
+ "hp = {}\n",
+ "hp[\"epochs\"] = 1000000\n",
+ "hp[\"learning_rate\"] = 0.001\n",
+ "hp[\"loss_function\"] = \"MSE\"\n",
+ "hp[\"save_path\"] = modelFolder\n",
+ "hp[\"is_save\"] = True\n",
+ "hp[\"is_plot\"] = True"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "d3e6863e",
+ "metadata": {},
+ "source": [
+ "**Set Neural Network's Architecture within ```hp```:**\n",
+ "
\n",
+ " We are going to use Hyperbolic Tangent functions for activation functions, we will use 6 layers with 20 neurons each and a Glorot uniform initializer."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 8,
+ "id": "d1910223",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# NN\n",
+ "hp[\"activation\"] = \"tanh\"\n",
+ "hp[\"initializer\"] = \"Glorot uniform\"\n",
+ "hp[\"num_neurons\"] = 20\n",
+ "hp[\"num_layers\"] = 6"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "e68312f2",
+ "metadata": {},
+ "source": [
+ "**Create the domain of the computation:**\n",
+ "
\n",
+ "We define the domain of computation, and number of collocation points used to evaluate the PDEs residual."
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 9,
+ "id": "a67ef1a5",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# domain\n",
+ "hp[\"shapefile\"] = os.path.join(repoPath, \"dataset\", expFileName)\n",
+ "hp[\"num_collocation_points\"] = 5000"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "1770218e",
+ "metadata": {},
+ "source": [
+ "**Add the physics components, ``SSA``, to the PINN:**\n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 10,
+ "id": "7fd17e7e",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# physics\n",
+ "SSA = {}\n",
+ "SSA[\"scalar_variables\"] = {\"B\":1.26802073401e+08}\n",
+ "hp[\"equations\"] = {\"SSA\":SSA}"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "1bd7bdd7",
+ "metadata": {},
+ "source": [
+ "**There are default parameter settings in ```SSA``` such as:\n",
+ "
**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 11,
+ "id": "ab377f9b",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "def set_default(self):\n",
+ " self.input = ['x', 'y']\n",
+ " self.output = ['u', 'v', 's', 'H', 'C']\n",
+ " self.output_lb = [-1.0e4/self.yts, -1.0e4/self.yts, -1.0e3, 10.0, 0.01]\n",
+ " self.output_ub = [ 1.0e4/self.yts, 1.0e4/self.yts, 2.5e3, 2000.0, 1.0e4]\n",
+ " self.data_weights = [1.0e-8*self.yts**2.0, 1.0e-8*self.yts**2.0, 1.0e-6, 1.0e-6, 1.0e-8]\n",
+ " self.residuals = [\"fSSA1\", \"fSSA2\"]\n",
+ " self.pde_weights = [1.0e-10, 1.0e-10]\n",
+ "\n",
+ " # scalar variables: name:value\n",
+ " self.scalar_variables = {\n",
+ " 'n': 3.0, # exponent of Glen's flow law\n",
+ " 'B':1.26802073401e+08 # -8 degree C, cuffey\n",
+ " }"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "6826e5b4",
+ "metadata": {},
+ "source": [
+ "These default parameters are the names of the input and output variables in the PINN, scaling factors, weights in the loss function, and names of the residual and weights. \n",
+ "
\n",
+ "
\n",
+ "We can set the input and output names as any other names. But, the orders of the input and output lists matters. These input and output lists will be used to load the data in the next section:"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "f2fdb916",
+ "metadata": {},
+ "source": [
+ "**Set the Data used for training in ```issm``` dictionary:**\n",
+ "
\n"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 12,
+ "id": "3919cda9",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# data\n",
+ "issm = {}\n",
+ "issm[\"data_size\"] = {\"u\":1000, \"v\":1000, \"s\":1000, \"H\":1000, \"C\":None}\n",
+ "issm[\"data_path\"] = data_path\n",
+ "hp[\"data\"] = {\"ISSM\":issm}"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "84be5ff9",
+ "metadata": {},
+ "source": [
+ "**In ```data_size``` the variables are defined by pairs of ```key``` = ```value```. If keys are not redefined in ```name_map``` then they will be set as the default or what they are set as in the physics section.** \n",
+ "
\n",
+ "The value associated with each key gives the number of data points used for the training. When ```value``` is set as **None**, then only Dirichlet boundary condition around the domain boundary will be used for the corresponding ```key```. If the variable is included in the training, but not given in ```data_size```, then there will be no data used in the training for this variable. "
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "3a850bf5",
+ "metadata": {},
+ "source": [
+ "**Set the additional loss function:**\n",
+ "
\n",
+ "We add the additional loss function to balance the contributions between the fast flow and slow moving regions:"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 13,
+ "id": "96460a74",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ " #additional loss function\n",
+ "vel_loss = {}\n",
+ "vel_loss['name'] = \"vel log\"\n",
+ "vel_loss['function'] = \"VEL_LOG\"\n",
+ "vel_loss['weight'] = 1.0e-5\n",
+ "hp[\"additional_loss\"] = {\"vel\":vel_loss}"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "c6582173",
+ "metadata": {},
+ "source": [
+ "**Create the Experiment and Compile:** Now, let's run the PINN model.\n",
+ "
"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 14,
+ "id": "0b193514",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "ERROR:root:ERROR: MATLAB type not supported: bamggeom, (uint32)\n",
+ "ERROR:root:ERROR: MATLAB type not supported: bamgmesh, (uint32)\n",
+ "ERROR:root:ERROR: MATLAB type not supported: qmustatistics, (uint32)\n",
+ "ERROR:root:ERROR: MATLAB type not supported: lovenumbers, (uint32)\n",
+ "ERROR:root:ERROR: MATLAB type not supported: rotational, (uint32)\n",
+ "ERROR:root:ERROR: MATLAB type not supported: solidearthsettings, (uint32)\n",
+ "2024-05-26 23:00:00.868567: W tensorflow/core/common_runtime/gpu/gpu_bfc_allocator.cc:42] Overriding orig_value setting because the TF_FORCE_GPU_ALLOW_GROWTH environment variable is set. Original config value was 0.\n",
+ "2024-05-26 23:00:00.868624: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1613] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 3829 MB memory: -> device: 0, name: Tesla V100-SXM2-32GB, pci bus id: 0000:86:00.0, compute capability: 7.0\n"
+ ]
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "add input transform with [ -323028.74554883 -1870413.49271111] and [ -230318.78353497 -1740472.47370953]\n",
+ "add output transform with [-3.1709792e-04 -3.1709792e-04 -1.0000000e+03 1.0000000e+00\n",
+ " 1.0000000e-02] and [3.17097920e-04 3.17097920e-04 2.50000000e+03 2.00000000e+03\n",
+ " 1.94839119e+05]\n",
+ "Parameters: \n",
+ "\tTrainingParameter: \n",
+ "\t\tepochs:\t1000000\n",
+ "\t\toptimizer:\tadam\n",
+ "\t\tloss_functions:\t['MSE', 'MSE', 'MSE', 'MSE', 'MSE', 'MSE', 'MSE']\n",
+ "\t\tadditional_loss:\t{'vel': }\n",
+ "\t\tlearning_rate:\t0.001\n",
+ "\t\tloss_weights:\t[1e-10, 1e-10, 9945192.96, 9945192.96, 1e-06, 1e-06, 1e-08]\n",
+ "\t\thas_callbacks:\tFalse\n",
+ "\t\tmin_delta:\tNone\n",
+ "\t\tpatience:\tNone\n",
+ "\t\tperiod:\tNone\n",
+ "\t\tcheckpoint:\tFalse\n",
+ "\t\tsave_path:\t./Models/Upernavik_test_20240526_225959/\n",
+ "\t\tis_save:\tTrue\n",
+ "\t\tis_plot:\tTrue\n",
+ "\tDomainParameter: \n",
+ "\t\tshapefile:\t./examples/dataset/Upernavik-Copy1.exp\n",
+ "\t\tnum_collocation_points:\t5000\n",
+ "\tDataParameter: \n",
+ "\t\tISSM:\n",
+ "\t\t\tdata_path:\t./examples/dataset/Upernavik_ForSade.mat\n",
+ "\t\t\tdata_size:\t{'u': 1000, 'v': 1000, 's': 1000, 'H': 1000, 'C': None}\n",
+ "\t\t\tname_map:\t{'u': 'u', 'v': 'v', 's': 's', 'H': 'H', 'C': 'C'}\n",
+ "\t\t\tsource:\tISSM\n",
+ "\n",
+ "\tNNParameter: \n",
+ "\t\tinput_variables:\t['x', 'y']\n",
+ "\t\toutput_variables:\t['u', 'v', 's', 'H', 'C']\n",
+ "\t\tnum_neurons:\t20\n",
+ "\t\tnum_layers:\t6\n",
+ "\t\tactivation:\ttanh\n",
+ "\t\tinitializer:\tGlorot uniform\n",
+ "\t\tinput_lb:\t[ -323028.74554883 -1870413.49271111]\n",
+ "\t\tinput_ub:\t[ -230318.78353497 -1740472.47370953]\n",
+ "\t\toutput_lb:\t[-3.1709792e-04 -3.1709792e-04 -1.0000000e+03 1.0000000e+00\n",
+ " 1.0000000e-02]\n",
+ "\t\toutput_ub:\t[3.17097920e-04 3.17097920e-04 2.50000000e+03 2.00000000e+03\n",
+ " 1.94839119e+05]\n",
+ "\t\tinput_size:\t2\n",
+ "\t\toutput_size:\t5\n",
+ "\tPhysicsParameter: \n",
+ "\t\tSSA:\n",
+ "\t\t\trhoi:\t917.0\n",
+ "\t\t\trhow:\t1023.0\n",
+ "\t\t\tg:\t9.81\n",
+ "\t\t\tyts:\t31536000.0\n",
+ "\t\t\tinput:\t['x', 'y']\n",
+ "\t\t\toutput:\t['u', 'v', 's', 'H', 'C']\n",
+ "\t\t\toutput_lb:\t[-0.00031709791983764585, -0.00031709791983764585, -1000.0, 10.0, 0.01]\n",
+ "\t\t\toutput_ub:\t[0.00031709791983764585, 0.00031709791983764585, 2500.0, 2000.0, 10000.0]\n",
+ "\t\t\tdata_weights:\t[9945192.96, 9945192.96, 1e-06, 1e-06, 1e-08]\n",
+ "\t\t\tresiduals:\t['fSSA1', 'fSSA2']\n",
+ "\t\t\tpde_weights:\t[1e-10, 1e-10]\n",
+ "\t\t\tscalar_variables:\t{'n': 3.0, 'B': 126802073.401}\n",
+ "\n",
+ "\n",
+ "Compiling model...\n",
+ "'compile' took 0.002921 s\n",
+ "\n"
+ ]
+ }
+ ],
+ "source": [
+ "# create experiment\n",
+ "experiment = pinn.PINN(hp)\n",
+ "print(experiment.params)\n",
+ "experiment.compile()"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "285abe0f",
+ "metadata": {},
+ "source": [
+ "**Train the experiment:**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "2063f171",
+ "metadata": {},
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Training model...\n",
+ "\n"
+ ]
+ },
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "/opt/conda/lib/python3.8/site-packages/keras/initializers/initializers_v2.py:120: UserWarning: The initializer GlorotUniform is unseeded and being called multiple times, which will return identical values each time (even if the initializer is unseeded). Please update your code to provide a seed to the initializer, or avoid using the same initalizer instance more than once.\n",
+ " warnings.warn(\n"
+ ]
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "WARNING:tensorflow:From /opt/conda/lib/python3.8/site-packages/tensorflow/python/autograph/pyct/static_analysis/liveness.py:83: Analyzer.lamba_check (from tensorflow.python.autograph.pyct.static_analysis.liveness) is deprecated and will be removed after 2023-09-23.\n",
+ "Instructions for updating:\n",
+ "Lambda fuctions will be no more assumed to be used in the statement where they are used, or at least in the same block. https://github.com/tensorflow/tensorflow/issues/56089\n"
+ ]
+ },
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "WARNING:tensorflow:From /opt/conda/lib/python3.8/site-packages/tensorflow/python/autograph/pyct/static_analysis/liveness.py:83: Analyzer.lamba_check (from tensorflow.python.autograph.pyct.static_analysis.liveness) is deprecated and will be removed after 2023-09-23.\n",
+ "Instructions for updating:\n",
+ "Lambda fuctions will be no more assumed to be used in the statement where they are used, or at least in the same block. https://github.com/tensorflow/tensorflow/issues/56089\n"
+ ]
+ },
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Step Train loss Test loss Test metric\n",
+ "0 [1.98e+06, 9.41e+06, 3.83e-03, 1.37e-02, 3.37e-01, 1.87e-01, 4.47e+01] [1.98e+06, 9.41e+06, 3.83e-03, 1.37e-02, 3.37e-01, 1.87e-01, 4.47e+01] [] \n",
+ "10000 [1.71e+00, 2.66e+00, 1.33e-03, 3.69e-04, 1.68e-01, 7.91e-02, 1.72e+02] [1.71e+00, 2.66e+00, 1.33e-03, 3.69e-04, 1.68e-01, 7.91e-02, 1.72e+02] [] \n"
+ ]
+ }
+ ],
+ "source": [
+ "# Train\n",
+ "experiment.train()"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "8a5962a3",
+ "metadata": {},
+ "source": [
+ "#### The results should look like this: \n",
+ ""
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "564afc06",
+ "metadata": {},
+ "source": [
+ "**Show the results with ```experiment.plot_predictions()```:**\n",
+ "
\n"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "ad0395b6",
+ "metadata": {},
+ "source": [
+ "### The results should look like: \n",
+ ""
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "e98c68f2",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "# show results\n",
+ "experiment.plot_predictions(X_ref=experiment.model_data.data[\"ISSM\"].X_dict, \n",
+ " sol_ref=experiment.model_data.data[\"ISSM\"].data_dict)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "id": "61862ab8",
+ "metadata": {},
+ "source": [
+ "**Use ```plot_similarity``` to plot the difference between predicted and refrence values for $C$**"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "96dcdd69",
+ "metadata": {},
+ "outputs": [],
+ "source": [
+ "pinn.utils.plotting.plot_similarity(pinn=experiment,feature_name = 'C', cmap='cubehelix')"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "id": "deefe01f",
+ "metadata": {},
+ "outputs": [],
+ "source": []
+ }
+ ],
+ "metadata": {
+ "kernelspec": {
+ "display_name": "Python 3 (ipykernel)",
+ "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": "3.8.8"
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/pinnicle/modeldata/issm_data.py b/pinnicle/modeldata/issm_data.py
index c6f3428..0542a8a 100644
--- a/pinnicle/modeldata/issm_data.py
+++ b/pinnicle/modeldata/issm_data.py
@@ -53,7 +53,17 @@ def load_data(self):
self.data_dict['s'] = md['geometry']['surface']
self.data_dict['a'] = (md['smb']['mass_balance'] - md['balancethickness']['thickening_rate'])/self.yts
self.data_dict['H'] = md['geometry']['thickness']
- self.data_dict['C'] = md['friction']['C']
+ try:
+ self.data_dict['C'] = md['friction']['C'] # Weertman
+ except:
+ C_b = md['friction']['coefficient'] # Budd
+ rho_ice = md['materials']['rho_ice']
+ rho_w = md['materials']['rho_water']
+ g = md['constants']['g']
+ base = md['geometry']['base']
+ N = rho_ice*g*self.data_dict['H'] + rho_w*g*base
+ N[np.where(N <= 0, True, False)] = 1
+ self.data_dict['C'] = C_b*np.sqrt(N)
self.data_dict['B'] = md['materials']['rheology_B']
self.data_dict['vel'] = np.sqrt(self.data_dict['u']**2.0+self.data_dict['v']**2.0)
# clean up is any of the keys are empty