{ "cells": [ { "cell_type": "markdown", "id": "5d904dee", "metadata": {}, "source": [ "# Example 10: Relativitistic Velocity Addition" ] }, { "cell_type": "markdown", "id": "6465ec94", "metadata": {}, "source": [ "In this example, we will symbolically regress $f(u,v)=\\frac{u+v}{1+uv}$. In relavitity, we know the rapidity trick $f(u,v)={\\rm tanh}({\\rm arctanh}\\ u+{\\rm arctanh}\\ v)$. Can we rediscover rapidity trick with KAN?" ] }, { "cell_type": "markdown", "id": "94056ef6", "metadata": {}, "source": [ "Intialize model and create dataset" ] }, { "cell_type": "code", "execution_count": 1, "id": "0a59179d", "metadata": {}, "outputs": [], "source": [ "from kan import *\n", "\n", "# initialize KAN with G=3\n", "model = KAN(width=[2,1,1], grid=10, k=3)\n", "\n", "# create dataset\n", "f = lambda x: (x[:,[0]]+x[:,[1]])/(1+x[:,[0]]*x[:,[1]])\n", "dataset = create_dataset(f, n_var=2, ranges=[-0.9,0.9])" ] }, { "cell_type": "markdown", "id": "cb1f817e", "metadata": {}, "source": [ "Train KAN and plot" ] }, { "cell_type": "code", "execution_count": 2, "id": "a87b97b0", "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "train loss: 6.27e-04 | test loss: 5.79e-04 | reg: 5.51e+00 : 100%|██| 20/20 [00:04<00:00, 4.76it/s]\n" ] } ], "source": [ "model.fit(dataset, opt=\"LBFGS\", steps=20);" ] }, { "cell_type": "code", "execution_count": 3, "id": "3f1cfc9d", "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "model.plot(beta=10)" ] }, { "cell_type": "markdown", "id": "5ca6421a", "metadata": {}, "source": [ "Retrain the model, the loss remains similar, meaning that the locking does not degrade model behavior, justifying our hypothesis that these two activation functions are the same. Let's now determine what this function is using $\\texttt{suggest_symbolic}$" ] }, { "cell_type": "code", "execution_count": 4, "id": "2ccb7048", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ " function fitting r2 r2 loss complexity complexity loss total loss\n", "0 arctanh 0.999999 -16.470396 4 4 -16.470396\n", "1 tan 0.999842 -12.540685 3 3 -12.540685\n", "2 arcsin 0.998866 -9.771875 4 4 -9.771875\n", "3 arccos 0.998866 -9.771725 4 4 -9.771725\n", "4 x^0.5 0.982258 -5.815842 2 2 -5.815842\n" ] }, { "data": { "text/plain": [ "('arctanh',\n", " ((x)>,\n", " (x)>,\n", " 4,\n", " (x, y_th)>),\n", " 0.9999989867210388,\n", " 4)" ] }, "execution_count": 4, "metadata": {}, "output_type": "execute_result" } ], "source": [ "model.suggest_symbolic(0,1,0,weight_simple=0.0)" ] }, { "cell_type": "markdown", "id": "0092be41", "metadata": {}, "source": [ "We can see that ${\\rm arctanh}$ is at the top of the suggestion list! So we can set both to arctanh, retrain the model, and plot it." ] }, { "cell_type": "code", "execution_count": 5, "id": "1bb96fe1", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "r2 is 0.9999993443489075\n", "r2 is 0.9999989867210388\n" ] }, { "data": { "text/plain": [ "tensor(1.0000)" ] }, "execution_count": 5, "metadata": {}, "output_type": "execute_result" } ], "source": [ "model.fix_symbolic(0,0,0,'arctanh')\n", "model.fix_symbolic(0,1,0,'arctanh')" ] }, { "cell_type": "code", "execution_count": 6, "id": "83b852a3", "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "train loss: 3.91e-04 | test loss: 4.70e-04 | reg: 5.54e+00 : 100%|██| 20/20 [00:02<00:00, 7.30it/s]\n" ] } ], "source": [ "model.fit(dataset, opt=\"LBFGS\", steps=20, update_grid=False);" ] }, { "cell_type": "code", "execution_count": 7, "id": "9ccd0923", "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "model.plot(beta=10)" ] }, { "cell_type": "markdown", "id": "4b98a727", "metadata": {}, "source": [ "We will see that ${\\rm tanh}$ is at the top of the suggestion list! So we can set it to ${\\rm tanh}$, retrain the model to machine precision, plot it and finally get the symbolic formula." ] }, { "cell_type": "code", "execution_count": 8, "id": "99ad38b9", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ " function fitting r2 r2 loss complexity complexity loss total loss\n", "0 tanh 0.999974 -14.743149 3 3 -0.548630\n", "1 x 0.945782 -4.204828 1 1 -0.040966\n", "2 0 0.000000 0.000014 0 0 0.000003\n", "3 cos 0.995867 -7.915010 2 2 0.016998\n", "4 sin 0.995867 -7.915010 2 2 0.016998\n" ] }, { "data": { "text/plain": [ "('tanh',\n", " ((x)>,\n", " (x)>,\n", " 3,\n", " (x, y_th)>),\n", " 0.9999735355377197,\n", " 3)" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "model.suggest_symbolic(1,0,0)" ] }, { "cell_type": "code", "execution_count": 9, "id": "af24c80d", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "r2 is 0.9999735355377197\n" ] }, { "data": { "text/plain": [ "tensor(1.0000)" ] }, "execution_count": 9, "metadata": {}, "output_type": "execute_result" } ], "source": [ "model.fix_symbolic(1,0,0,'tanh')" ] }, { "cell_type": "code", "execution_count": 10, "id": "01936f17", "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "train loss: 6.55e-08 | test loss: 6.85e-08 | reg: 5.57e+00 : 100%|██| 20/20 [00:01<00:00, 17.11it/s]\n" ] } ], "source": [ "model.fit(dataset, opt=\"LBFGS\", steps=20);" ] }, { "cell_type": "code", "execution_count": 11, "id": "76bcc188", "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "
" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "model.plot()" ] }, { "cell_type": "code", "execution_count": 12, "id": "b62b0246", "metadata": {}, "outputs": [ { "data": { "text/latex": [ "$\\displaystyle \\tanh{\\left(\\operatorname{atanh}{\\left(x_{1} \\right)} + \\operatorname{atanh}{\\left(x_{2} \\right)} \\right)}$" ], "text/plain": [ "tanh(atanh(x_1) + atanh(x_2))" ] }, "execution_count": 12, "metadata": {}, "output_type": "execute_result" } ], "source": [ "formula = model.symbolic_formula()[0][0]\n", "nsimplify(ex_round(formula, 4))" ] } ], "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": "3.9.7" } }, "nbformat": 4, "nbformat_minor": 5 }