{ "cells": [ { "cell_type": "code", "execution_count": 1, "id": "22e44d2d-6987-4daa-b6a2-c5422d27b0d2", "metadata": {}, "outputs": [], "source": [ "import mlatom as ml" ] }, { "cell_type": "code", "execution_count": 2, "id": "4b74579b-ea47-45e4-8174-6da305c6f664", "metadata": {}, "outputs": [], "source": [ "# get initial structure\n", "h2 = ml.molecule.from_xyz_string('''2\n", "\n", "H 0 0 0\n", "H 0 0 0.7\n", "''')" ] }, { "cell_type": "code", "execution_count": 3, "id": "97b70cef-3541-43a3-9d97-6b93db33536a", "metadata": {}, "outputs": [], "source": [ "# define the reference method\n", "dft = ml.models.methods(method='B3LYP/6-31G*')" ] }, { "cell_type": "code", "execution_count": 4, "id": "d2c7267e-4c48-427e-a1d4-14c42eb04c69", "metadata": {}, "outputs": [], "source": [ "# optimize geometry and calculate frequency\n", "'''optmol = ml.optimize_geometry(\n", " model=dft,\n", " initial_molecule=h2\n", " ).optimized_molecule\n", "freq = ml.freq(model=dft,\n", "molecule=optmol)\n", "'''\n", "optmol = ml.molecule.load('materials/h2_freq.json')" ] }, { "cell_type": "code", "execution_count": 5, "id": "bf23832a-7b71-4abd-892d-3303f0ff12a2", "metadata": {}, "outputs": [], "source": [ "# settings for sampling points during AL loop\n", "sampler_kwargs = {\n", " 'initcond_sampler':'wigner',\n", " 'initcond_sampler_kwargs':{\n", " 'molecule':optmol,\n", " 'number_of_initial_conditions':100,\n", " 'initial_temperature':300},\n", " 'maximum_propagation_time':100.0,\n", " 'time_step':0.5,\n", " 'uq_threshold':0.000001,\n", "}" ] }, { "cell_type": "code", "execution_count": null, "id": "2db3efbc-603d-43a8-b93c-4b1b51e238aa", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ " Initial points sampling: Use cross validation\n", " Initial points sampling: Number of CV splits = 5\n", "Start initial points sampling...\n", " Initial points sampling: initial points RMSE threshold not found, fit learning curve instead\n", " Initial points sampling: Train on energies only\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=6.038879764503849e-10, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.35355339059327373\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.35355339059327373\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.35355339059327373\n", " Number of points: 50\n", " eRMSE = 1.0850793908605848e-06 Hartree\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.35355339059327373\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.35355339059327373\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.35355339059327373\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.10511205190671434\n", " Number of points: 100\n", " eRMSE = 2.577756354077729e-07 Hartree\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.35355339059327373\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.10511205190671434\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.35355339059327373\n", " Number of points: 150\n", " eRMSE = 7.148229877763e-08 Hartree\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.10511205190671434\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.10511205190671434\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.10511205190671434\n", " Number of points: 200\n", " eRMSE = 3.1661348155695476e-07 Hartree\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.35355339059327373\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.03125000000000001\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.10511205190671434\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.10511205190671434\n", " Number of points: 250\n", " eRMSE = 1.7222338387532526e-07 Hartree\n", " Linear regression: log(e) = -9.959073131303462 + -1.072119399985807 log(Ntr)\n", " Linear regression: Pearson correlation coefficient = -0.6886462379669256\n", " [e(Ntr)-e(Ntr+50)]/e(Ntr) = 0.17755237655667533\n", " Improvement is large, continue sampling\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.10511205190671434\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=1.189207115002721\n", "Training the main auxiliary model\n", "Optimized hyperparameters for energy aux model: lambda=2.9103830456733704e-11, sigma=0.10511205190671434\n", " Number of points: 300\n", " eRMSE = 1.096682985216811e-06 Hartree\n", " Linear regression: log(e) = -13.295918146719133 + -0.3276593524230377 log(Ntr)\n", " Linear regression: Pearson correlation coefficient = -0.20364644630691536\n", " [e(Ntr)-e(Ntr+50)]/e(Ntr) = 0.04925454434025745\n", " Improvement is small, initial points sampling done\n", " Number of initial points: 300\n", "Initial points sampling time: 42.54678368102759 s\n", "Active learning iteration 0\n", "Points to label: 300\n", "New labeled points: 300\n", "0 points are abandoned due to failed calculation\n", "Number of points in the labeled data set: 300\n", "Optimized hyperparameters for main model: lambda=2.9103830456733704e-11, sigma=0.10511205190671434\n", "Optimized hyperparameters for aux model: lambda=2.9103830456733704e-11, sigma=0.03125000000000001\n", "New threshold: 1.0227106919180605e-08\n", " Number of training points: 300\n", " Number of subtraining points: 270\n", " Number of validation points: 30\n", " Main model\n", " Subtraining set:\n", " RMSE of values = 5.881435953785552e-09\n", " Correlation coefficient = 0.9999999999998377\n", " RMSE of gradients = 2.490201827173034e-09\n", " Correlation coefficient = 0.9999999999999997\n", " Validation set:\n", " RMSE of values = 0.0\n", " Correlation coefficient = 0.9999999999999999\n", " RMSE of gradients = 0.0\n", " Correlation coefficient = 0.9999999999999998\n", " Auxiliary model\n", " Subtraining set:\n", " RMSE of values = 9.34233727021504e-10\n", " Correlation coefficient = 0.9999999999999958\n", " Validation set:\n", " RMSE of values = 5.469203647668958e-09\n", " Correlation coefficient = 0.9999999999999115\n", "UQ threshold is provided in sampler, overwriting saved threshold\n", "Current threshold: 1e-06\n", "Trajectory 1 number of steps: 201\n", "Trajectory 2 number of steps: 201\n", "Trajectory 3 number of steps: 201\n", "Trajectory 4 number of steps: 201\n", "Trajectory 5 number of steps: 201\n", "Trajectory 6 number of steps: 201\n", "Trajectory 7 number of steps: 201\n", "Trajectory 8 number of steps: 201\n", "Trajectory 9 number of steps: 201\n", "Trajectory 10 number of steps: 9\n", "Adding molecule from trajectory 10 at time 4.00 fs\n", "Trajectory 11 number of steps: 201\n", "Trajectory 12 number of steps: 201\n", "Trajectory 13 number of steps: 1\n", "Adding molecule from trajectory 13 at time 0.00 fs\n", "Trajectory 14 number of steps: 6\n", "Adding molecule from trajectory 14 at time 2.50 fs\n", "Trajectory 15 number of steps: 201\n", "Trajectory 16 number of steps: 201\n", "Trajectory 17 number of steps: 201\n", "Trajectory 18 number of steps: 201\n", "Trajectory 19 number of steps: 201\n", "Trajectory 20 number of steps: 201\n", "Trajectory 21 number of steps: 201\n", "Trajectory 22 number of steps: 4\n", "Adding molecule from trajectory 22 at time 1.50 fs\n", "Trajectory 23 number of steps: 201\n", "Trajectory 24 number of steps: 1\n", "Adding molecule from trajectory 24 at time 0.00 fs\n", "Trajectory 25 number of steps: 201\n", "Trajectory 26 number of steps: 12\n", "Adding molecule from trajectory 26 at time 5.50 fs\n", "Trajectory 27 number of steps: 201\n", "Trajectory 28 number of steps: 201\n", "Trajectory 29 number of steps: 201\n", "Trajectory 30 number of steps: 201\n", "Trajectory 31 number of steps: 201\n", "Trajectory 32 number of steps: 201\n", "Trajectory 33 number of steps: 201\n", "Trajectory 34 number of steps: 7\n", "Adding molecule from trajectory 34 at time 3.00 fs\n", "Trajectory 35 number of steps: 11\n", "Adding molecule from trajectory 35 at time 5.00 fs\n", "Trajectory 36 number of steps: 201\n", "Trajectory 37 number of steps: 5\n", "Adding molecule from trajectory 37 at time 2.00 fs\n", "Trajectory 38 number of steps: 201\n", "Trajectory 39 number of steps: 201\n", "Trajectory 40 number of steps: 201\n", "Trajectory 41 number of steps: 201\n", "Trajectory 42 number of steps: 201\n", "Trajectory 43 number of steps: 11\n", "Adding molecule from trajectory 43 at time 5.00 fs\n", "Trajectory 44 number of steps: 201\n", "Trajectory 45 number of steps: 201\n", "Trajectory 46 number of steps: 201\n", "Trajectory 47 number of steps: 201\n", "Trajectory 48 number of steps: 201\n", "Trajectory 49 number of steps: 201\n", "Trajectory 50 number of steps: 201\n", "Trajectory 51 number of steps: 12\n", "Adding molecule from trajectory 51 at time 5.50 fs\n", "Trajectory 52 number of steps: 201\n", "Trajectory 53 number of steps: 6\n", "Adding molecule from trajectory 53 at time 2.50 fs\n", "Trajectory 54 number of steps: 201\n", "Trajectory 55 number of steps: 201\n", "Trajectory 56 number of steps: 201\n", "Trajectory 57 number of steps: 10\n", "Adding molecule from trajectory 57 at time 4.50 fs\n", "Trajectory 58 number of steps: 201\n", "Trajectory 59 number of steps: 201\n", "Trajectory 60 number of steps: 201\n", "Trajectory 61 number of steps: 201\n", "Trajectory 62 number of steps: 201\n", "Trajectory 63 number of steps: 201\n", "Trajectory 64 number of steps: 201\n", "Trajectory 65 number of steps: 201\n", "Trajectory 66 number of steps: 201\n", "Trajectory 67 number of steps: 201\n", "Trajectory 68 number of steps: 201\n", "Trajectory 69 number of steps: 201\n", "Trajectory 70 number of steps: 201\n", "Trajectory 71 number of steps: 201\n", "Trajectory 72 number of steps: 201\n", "Trajectory 73 number of steps: 201\n", "Trajectory 74 number of steps: 201\n", "Trajectory 75 number of steps: 201\n", "Trajectory 76 number of steps: 201\n", "Trajectory 77 number of steps: 201\n", "Trajectory 78 number of steps: 201\n", "Trajectory 79 number of steps: 201\n", "Trajectory 80 number of steps: 201\n", "Trajectory 81 number of steps: 201\n", "Trajectory 82 number of steps: 5\n", "Adding molecule from trajectory 82 at time 2.00 fs\n", "Trajectory 83 number of steps: 201\n", "Trajectory 84 number of steps: 201\n", "Trajectory 85 number of steps: 201\n", "Trajectory 86 number of steps: 13\n", "Adding molecule from trajectory 86 at time 6.00 fs\n", "Trajectory 87 number of steps: 201\n", "Trajectory 88 number of steps: 201\n", "Trajectory 89 number of steps: 201\n", "Trajectory 90 number of steps: 201\n", "Trajectory 91 number of steps: 201\n", "Trajectory 92 number of steps: 201\n", "Trajectory 93 number of steps: 201\n", "Trajectory 94 number of steps: 201\n", "Trajectory 95 number of steps: 201\n", "Trajectory 96 number of steps: 201\n", "Trajectory 97 number of steps: 201\n", "Trajectory 98 number of steps: 201\n", "Trajectory 99 number of steps: 201\n", "Trajectory 100 number of steps: 201\n", "Iteration 0 takes 72.60129306931049 s\n", " Labeling time: 38.02987134503201 s\n", " Training time: 15.152436579111964 s\n", " Sampling time: 19.418985145166516 s\n", "\n", "\n", "\n", "\n", "Active learning iteration 1\n", "Points to label: 15\n", "New labeled points: 15\n", "0 points are abandoned due to failed calculation\n", "Number of points in the labeled data set: 315\n" ] } ], "source": [ "# start active learning\n", "ml.al(molecule=optmol,\n", " ml_model='KREG',\n", " sampler = 'md',\n", " sampler_kwargs=sampler_kwargs,\n", " min_new_points=5,\n", " reference_method=dft)" ] }, { "cell_type": "code", "execution_count": null, "id": "a8e05325-1dd2-4b6f-9284-149bb5664ef5", "metadata": {}, "outputs": [], "source": [ "# check the final labeled database\n", "db = ml.molecular_database.load(filename='labeled_db.json', format='json')\n", "len(db)" ] }, { "cell_type": "code", "execution_count": null, "id": "27f81e60-e537-462f-9a64-ba27b5751cc9", "metadata": {}, "outputs": [], "source": [ "for imol in range(len(db)):\n", " print(db[imol].internuclear_distance(0, 1), db[imol].energy)" ] }, { "cell_type": "code", "execution_count": null, "id": "ea683fe5-af9f-4597-9153-b138f950468c", "metadata": {}, "outputs": [], "source": [ "db.view()" ] }, { "cell_type": "code", "execution_count": null, "id": "d7fb4d47-20df-4e78-9de4-121ec068ea06", "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": "3.9.13" } }, "nbformat": 4, "nbformat_minor": 5 }