tra-analysis/data analysis/cudaRegressTesting.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def factorial(n):\n",
    "    if n==0:\n",
    "        return 1\n",
    "    else:\n",
    "        return n*factorial(n-1)\n",
    "\n",
    "def take_all_pwrs(vec,pwr):\n",
    "    #todo: vectorize (kinda)\n",
    "    combins=torch.combinations(vec, r=pwr, with_replacement=True)\n",
    "    out=torch.ones(combins.size()[0])\n",
    "    for i in torch.t(combins):\n",
    "        out *= i\n",
    "    return out\n",
    "\n",
    "class LinearRegKernel():\n",
    "    parameters= []\n",
    "    weights=None\n",
    "    bias=None\n",
    "    def __init__(self, num_vars):\n",
    "        self.weights=torch.rand(num_vars, requires_grad=True)\n",
    "        self.bias=torch.rand(1, requires_grad=True)\n",
    "        self.parameters=[self.weights,self.bias]\n",
    "    def forward(self,mtx):\n",
    "        long_bias=self.bias.repeat([1,mtx.size()[1]])\n",
    "        return torch.matmul(self.weights,mtx)+long_bias\n",
    "    \n",
    "class SigmoidalRegKernel():\n",
    "    parameters= []\n",
    "    weights=None\n",
    "    bias=None\n",
    "    sigmoid=torch.nn.Sigmoid()\n",
    "    def __init__(self, num_vars):\n",
    "        self.weights=torch.rand(num_vars)\n",
    "        self.bias=torch.rand(1)\n",
    "        self.parameters=[self.weights,self.bias]\n",
    "    def forward(self,mtx):\n",
    "        long_bias=self.bias.repeat([1,mtx.size()[1]])\n",
    "        return self.sigmoid(torch.matmul(self.weights,mtx)+long_bias)\n",
    "\n",
    "class LogRegKernel():\n",
    "    parameters= []\n",
    "    weights=None\n",
    "    bias=None\n",
    "    def __init__(self, num_vars):\n",
    "        self.weights=torch.rand(num_vars)\n",
    "        self.bias=torch.rand(1)\n",
    "        self.parameters=[self.weights,self.bias]\n",
    "    def forward(self,mtx):\n",
    "        long_bias=self.bias.repeat([1,mtx.size()[1]])\n",
    "        return torch.log(torch.matmul(self.weights,mtx)+long_bias)\n",
    "\n",
    "class ExpRegKernel():\n",
    "    parameters= []\n",
    "    weights=None\n",
    "    bias=None\n",
    "    def __init__(self, num_vars):\n",
    "        self.weights=torch.rand(num_vars)\n",
    "        self.bias=torch.rand(1)\n",
    "        self.parameters=[self.weights,self.bias]\n",
    "    def forward(self,mtx):\n",
    "        long_bias=self.bias.repeat([1,mtx.size()[1]])\n",
    "        return torch.exp(torch.matmul(self.weights,mtx)+long_bias)\n",
    "\n",
    "class PolyRegKernel():\n",
    "    parameters= []\n",
    "    weights=None\n",
    "    bias=None\n",
    "    power=None\n",
    "    def __init__(self, num_vars, power):\n",
    "        self.power=power\n",
    "        num_terms=int(factorial(num_vars+power-1) / factorial(power) / factorial(num_vars-1))\n",
    "        self.weights=torch.rand(num_terms)\n",
    "        self.bias=torch.rand(1)\n",
    "        self.parameters=[self.weights,self.bias]\n",
    "    def forward(self,mtx):\n",
    "        #TODO: Vectorize the last part\n",
    "        cols=[]\n",
    "        for i in torch.t(mtx):\n",
    "            cols.append(take_all_pwrs(i,self.power))\n",
    "        new_mtx=torch.t(torch.stack(cols))\n",
    "        long_bias=self.bias.repeat([1,mtx.size()[1]])\n",
    "        return torch.matmul(self.weights,new_mtx)+long_bias\n",
    "\n",
    "def SGDTrain(kernel, data, ground, loss=torch.nn.MSELoss(), iterations=1000, learning_rate=.1, return_losses=False):\n",
    "    optim=torch.optim.SGD(kernel.parameters, lr=learning_rate)\n",
    "    if (return_losses):\n",
    "        losses=[]\n",
    "        for i in range(iterations):\n",
    "            with torch.set_grad_enabled(True):\n",
    "                optim.zero_grad()\n",
    "                pred=kernel.forward(data)\n",
    "                ls=loss(pred,ground)\n",
    "                losses.append(ls.item())\n",
    "                ls.backward()\n",
    "                optim.step()\n",
    "        return [kernel,losses]\n",
    "    else:\n",
    "        for i in range(iterations):\n",
    "            with torch.set_grad_enabled(True):\n",
    "                optim.zero_grad()\n",
    "                pred=kernel.forward(data)\n",
    "                ls=loss(pred,ground)\n",
    "                ls.backward()\n",
    "                optim.step() \n",
    "        return kernel\n",
    "\n",
    "def CustomTrain(kernel, optim, data, ground, loss=torch.nn.MSELoss(), iterations=1000, return_losses=False):\n",
    "    if (return_losses):\n",
    "        losses=[]\n",
    "        for i in range(iterations):\n",
    "            with torch.set_grad_enabled(True):\n",
    "                optim.zero_grad()\n",
    "                pred=kernel.forward(data)\n",
    "                ls=loss(pred,ground)\n",
    "                losses.append(ls.item())\n",
    "                ls.backward()\n",
    "                optim.step()\n",
    "        return [kernel,losses]\n",
    "    else:\n",
    "        for i in range(iterations):\n",
    "            with torch.set_grad_enabled(True):\n",
    "                optim.zero_grad()\n",
    "                pred=kernel.forward(data)\n",
    "                ls=loss(pred,ground)\n",
    "                ls.backward()\n",
    "                optim.step() \n",
    "        return kernel"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[1.0000, 2.0000]], grad_fn=<AddBackward0>)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model=SGDTrain(LinearRegKernel(3),torch.tensor([[1,2],[3,4],[5,6]]).to(torch.float),torch.tensor([[1,2]]).to(torch.float),iterations=10000, learning_rate=.01, return_losses=True)\n",
    "model[0].forward(torch.tensor([[1,2],[3,4],[5,6]]).to(torch.float))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}