async?

Async subroutines.ipynb

@@ -0,0 +1,234 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "fa8b70aa",
+   "metadata": {},
+   "source": [
+    "# To sync or not to sync\n",
+    "\n",
+    "`Estimator` and `Sampler` primitives currently only allow asynchronous calls:\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "b6d26f95",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "est = Estimator()\n",
+    "job = est.run([circuit], [operator], [values])\n",
+    "result = job.result()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "c88fc5eb",
+   "metadata": {},
+   "source": [
+    "In context of the algorithms, we discussed whether we want to follow the same interface for advanced subroutines like gradients or fidelities, as"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "2d32cec8",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "grad = ParameterShiftEstimatorGradient()\n",
+    "job = grad.run([circuit], [operator], [values])\n",
+    "result = job.result()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "40dfd676",
+   "metadata": {},
+   "source": [
+    " Alternatively we would only allow synchronous calls as"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "e9495b9a",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "grad = ParameterShiftEstimatorGradient()\n",
+    "result = grad.evaluate([circuit], [operator], [values])"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "4fef8a91",
+   "metadata": {},
+   "source": [
+    "### Pro sync\n",
+    "\n",
+    "* easier syntax\n",
+    "* \"where do we draw the line?\" -- if subroutines have `.run` do algorithms also need it?"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "0a0a12e3",
+   "metadata": {},
+   "source": [
+    "### Pro async\n",
+    "\n",
+    "* consistency with other estimator and sampler\n",
+    "* subroutines and primitives are used in the same piece of code -- different signatures look weird"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "9c6cfc4e",
+   "metadata": {},
+   "source": [
+    "### Examples"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "385ec5be",
+   "metadata": {},
+   "source": [
+    "#### VQE \"sync\""
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "963dab2e",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class VQE(MinimumEigensolver):\n",
+    "    ...\n",
+    "    \n",
+    "    def compute_minimum_eigenvalue(self, operator):\n",
+    "        \n",
+    "        def energy(theta):\n",
+    "            job = self.estimator.run([self.ansatz], [operator], [theta])\n",
+    "            return job.result()\n",
+    "            \n",
+    "        def gradient(theta):\n",
+    "            return self.gradient.evaluate([self.ansatz], [operator], [theta])\n",
+    "        \n",
+    "        result = self.optimizer.minimize(energy, initial_guess, jac=gradient)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "25f430b5",
+   "metadata": {},
+   "source": [
+    "### VQE \"all async\""
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "c10ca846",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class VQE(MinimumEigensolver):\n",
+    "    ...\n",
+    "    \n",
+    "    def compute_minimum_eigenvalue(self, operator):\n",
+    "        \n",
+    "        def energy(theta):\n",
+    "            job = self.run([self.ansatz], [operator], [theta])\n",
+    "            return job.result()\n",
+    "            \n",
+    "        def gradient(theta):\n",
+    "            job = self.gradient.run([self.ansatz], [operator], [theta])\n",
+    "            return job.result()\n",
+    "        \n",
+    "        result = self.optimizer.minimize(energy, initial_guess, jac=gradient)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "80102684",
+   "metadata": {},
+   "source": [
+    "\n",
+    "\n",
+    "#### VQD \"sync\""
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "0d2b6b00",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class VQD(Eigensolver):\n",
+    "    ...\n",
+    "    \n",
+    "    def energy_evaluation(operator, theta, penalty):\n",
+    "        job = self.estimator.run([self.ansatz], [operator], [theta])\n",
+    "        energy = job.result()\n",
+    "        \n",
+    "        overlap = self.fidelity.evaluate([self.ansatz], [self.ansatz], [self.optimal_point], [theta])\n",
+    "        \n",
+    "        return energy + penalty * overlap"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "4d592842",
+   "metadata": {},
+   "source": [
+    "#### VQD \"all async\""
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d2de0598",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class VQD(Eigensolver):\n",
+    "    ...\n",
+    "    \n",
+    "    def energy_evaluation(operator, theta, penalty):\n",
+    "        jobs = [\n",
+    "            self.estimator.run([self.ansatz], [operator], [theta]),\n",
+    "            self.fidelity.run([self.ansatz], [self.ansatz], [self.optimal_point], [theta])\n",
+    "        ]\n",
+    "        energy, overlap = [job.result() for job in jobs]\n",
+    "                \n",
+    "        return energy + penalty * overlap"
+   ]
+  }
+ ],
+ "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": "3.10.4"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}