truncated_gist.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import aesara\n",
    "import aesara.tensor as at\n",
    "from aesara.graph.basic import Apply\n",
    "from aesara.tensor.random.op import RandomVariable"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "outputs": [],
   "source": [
    "class TruncatedRV(RandomVariable):\n",
    "\n",
    "    def __init__(self, *args, base_op=None, **kwargs):\n",
    "        self.base_op = base_op\n",
    "        super().__init__(*args, **kwargs)\n",
    "\n",
    "    def make_node(self, *args, **kwargs):\n",
    "        *dist_params, lower, upper = args\n",
    "\n",
    "        lower = at.as_tensor_variable(lower)\n",
    "        upper = at.as_tensor_variable(upper)\n",
    "        base_apply = super().make_node(*dist_params, **kwargs)\n",
    "\n",
    "        return Apply(\n",
    "            self,\n",
    "            [*base_apply.inputs, lower, upper],\n",
    "            [out.type() for out in base_apply.outputs]\n",
    "        )\n",
    "\n",
    "    def rng_fn(self, *args):\n",
    "        rng, *dist_args, lower, upper, size = args\n",
    "        base_rng_fn = self.base_op.rng_fn\n",
    "\n",
    "        res = base_rng_fn(rng, *dist_args, size)\n",
    "        is_scalar = np.isscalar(res)\n",
    "\n",
    "        outside_bounds = (res < lower) | (res > upper)\n",
    "        while np.any(outside_bounds):\n",
    "            new_res = base_rng_fn(rng, *dist_args, size)\n",
    "            if is_scalar:\n",
    "                res = new_res\n",
    "            else:\n",
    "                res[outside_bounds] = new_res[outside_bounds]\n",
    "            outside_bounds = (res < lower) | (res > upper)\n",
    "\n",
    "        return res\n",
    "\n",
    "def truncate(rv, lower, upper):\n",
    "    base_op = rv.owner.op\n",
    "    truncated_rv = TruncatedRV(\n",
    "        f'truncated_{rv.name}',\n",
    "        base_op.ndim_supp,\n",
    "        list(base_op.ndims_params) + [base_op.ndim_supp]*2,\n",
    "        base_op.dtype,\n",
    "        inplace=False,\n",
    "        base_op=base_op,\n",
    "    )\n",
    "    rng, size, dtype, *dist_params = rv.owner.inputs\n",
    "    return truncated_rv(*dist_params, lower, upper, size=size)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "outputs": [],
   "source": [
    "x = at.random.uniform(0, 1, size=10)\n",
    "y = truncate(x, lower=0.4, upper=0.45)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "truncated_None_rv.1 [id A] ''   \n",
      " |RandomStateSharedVariable(<RandomState(MT19937) at 0x7FCE59180640>) [id B]\n",
      " |TensorConstant{(1,) of 10} [id C]\n",
      " |TensorConstant{11} [id D]\n",
      " |TensorConstant{0} [id E]\n",
      " |TensorConstant{1} [id F]\n",
      " |TensorConstant{0.4} [id G]\n",
      " |TensorConstant{0.45} [id H]\n"
     ]
    }
   ],
   "source": [
    "aesara.dprint(y)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "outputs": [
    {
     "data": {
      "text/plain": "array([0.42512075, 0.43904175, 0.44705009, 0.43412428, 0.42623413,\n       0.42201117, 0.42752124, 0.40940581, 0.43109232, 0.43450091])"
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y.eval()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}