ee046746_tut_08_deep_uncertainty.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# <img src=\"https://img.icons8.com/bubbles/100/000000/3d-glasses.png\" style=\"height:50px;display:inline\"> EE 046746 - Technion - Computer Vision\n",
    "\n",
    "#### Elias Nehme\n",
    "\n",
    "## Tutorial 08 - Uncertainty Quantification in Deep Learning\n",
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## <img src=\"https://img.icons8.com/bubbles/50/000000/checklist.png\" style=\"height:50px;display:inline\"> Agenda\n",
    "---\n",
    "* [Why We Need Uncertainty?](#-Why-We-Need-Uncertainty?)\n",
    "* [Types of Uncertainty](#-Types-of-Uncertainty)\n",
    "    * [Epistemic/Model Uncertainty](#-Epistemic/Model-Uncertainty)\n",
    "    * [Aleatoric/Data Uncertainty](#-Aleatoric/Data-Uncertainty)\n",
    "* [Computing Uncertainty](#-Computing-Uncertainty)\n",
    "    * [Likelihood Modelling](#-Likelihood-Modelling)\n",
    "    * [Bayesian Neural Networks](#-Bayesian-Neural-Networks)\n",
    "    * [Evidental Deep Learning](#-Evidental-Deep-Learning)\n",
    "* [Computer Vision Applications](#-Computer-Vision-Applications)\n",
    "* [Existing Libraries](#-Blitz---Bayesian-Neural-Networks-in-PyTorch)\n",
    "* [Recommended Videos](#-Recommended-Videos)\n",
    "* [Credits](#-Credits)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/doodle/96/000000/average-value.png\" style=\"height:50px;display:inline\"> Why We Need Uncertainty?\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/dice.gif\" width=\"400\">\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### With great power..\n",
    "---\n",
    "\n",
    "<img src=\"./assets/uncertainty_why4.jpg\" width=\"400\">\n",
    "\n",
    "* Control handed-over to automated systems; with many dangerous scenarios\n",
    "    * Medical diagnosis\n",
    "    * Automotive and Self-driving Cars\n",
    "    * High frequency Trading (affecting economic markets on global scale)\n",
    "    * etc.\n",
    "\n",
    "* <a href=\"http://www.cs.ox.ac.uk/people/yarin.gal/website/bdl101/MLSS_2019_BDL_1.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Knowing what you don't know\n",
    "---\n",
    "* A deep learning model should be able to say: “sorry, I don’t know”\n",
    "* If it can tell us how certain it is we can plan accordingly\n",
    "\n",
    "<img src=\"./assets/uncertainty_why1.jpg\" width=\"1200\">\n",
    "\n",
    "* <a href=\"https://www.slideshare.net/perone/uncertainty-estimation-in-deep-learning/5\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Diagnosis in medicine\n",
    "---\n",
    "* Suppose you are trying to classify lesions in MRI\n",
    "\n",
    "<img src=\"./assets/uncertainty_why2.jpg\" width=\"1200\">\n",
    "\n",
    "* Later on: change in acquisition setup leads to differently looking samples\n",
    "* Problem: we will still get predictions with **high probability** for an Out-of-distrbution (OOD) sample.\n",
    "* <a href=\"https://www.slideshare.net/perone/uncertainty-estimation-in-deep-learning/7\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Out-of-distribution detection\n",
    "---\n",
    "* Systems applied to toy data deployed in real-life settings\n",
    "\n",
    "<img src=\"./assets/uncertainty_why9.jpg\" width=\"1200\">\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Out-of-distribution detection\n",
    "---\n",
    "* Likelihood vs confidence\n",
    "\n",
    "<img src=\"./assets/uncertainty_why10.jpg\" width=\"1200\">\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Out-of-distribution detection\n",
    "---\n",
    "* Likelihood vs confidence\n",
    "\n",
    "<img src=\"./assets/uncertainty_why11.jpg\" width=\"1200\">\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Out-of-distribution detection\n",
    "---\n",
    "* Likelihood vs confidence\n",
    "\n",
    "<img src=\"./assets/uncertainty_why12.jpg\" width=\"1200\">\n",
    "\n",
    "* Classifier forced to choose between 2 possible outcomes even if it have no clue\n",
    "* July 2015, classification system erroneously identified two African American humans as gorillas, raising concerns of racial discrimination.\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Autonomous Driving\n",
    "---\n",
    "* Semantic Segmentation used as a strong cue for motion planning\n",
    "\n",
    "<img src=\"./assets/uncertainty_why5.jpg\" width=\"1200\">\n",
    "\n",
    "* <a href=\"http://www.cs.ox.ac.uk/people/yarin.gal/website/bdl101/MLSS_2019_BDL_1.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Autonomous Driving\n",
    "---\n",
    "* Ground Truth Segmentation Labels\n",
    "\n",
    "<img src=\"./assets/uncertainty_why6.jpg\" width=\"1200\">\n",
    "\n",
    "* <a href=\"http://www.cs.ox.ac.uk/people/yarin.gal/website/bdl101/MLSS_2019_BDL_1.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Autonomous Driving\n",
    "---\n",
    "* Predicted Segmentation Labels\n",
    "\n",
    "<img src=\"./assets/uncertainty_why7.jpg\" width=\"1200\">\n",
    "\n",
    "* <a href=\"http://www.cs.ox.ac.uk/people/yarin.gal/website/bdl101/MLSS_2019_BDL_1.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Autonomous Driving\n",
    "---\n",
    "* Predicted Uncertainty per-pixel\n",
    "\n",
    "<img src=\"./assets/uncertainty_why8.jpg\" width=\"1200\">\n",
    "\n",
    "* May 2016, tragically experienced the first fatality from an assisted driving system. White side of tractor trailer confused against a brightly lit sky.\n",
    "* <a href=\"http://www.cs.ox.ac.uk/people/yarin.gal/website/bdl101/MLSS_2019_BDL_1.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Active learning\n",
    "---\n",
    "* Agent chooses which unlabelled data is most informative\n",
    "* Asks external \\oracle\" (eg human annotator) for a label only for that\n",
    "* Acquisition function: ranks points based on their potential informativeness, e.g. model uncertainty\n",
    "\n",
    "<img src=\"./assets/uncertainty_why3.jpg\" width=\"1200\">\n",
    "\n",
    "* <a href=\"http://www.cs.ox.ac.uk/people/yarin.gal/website/bdl101/MLSS_2019_BDL_2.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "* Many more applications including:\n",
    "    * Explore/exploit dilemma in Reinforcement Learning\n",
    "    * Model understanding/Dataset bias understanding\n",
    "    * Sample-adaptive loss-function tuning\n",
    "    * etc"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/officexs/60/000000/area-chart.png\" style=\"height:50px;display:inline\"> Types of Uncertainty\n",
    "--- \n",
    "* Epistemic/Model Uncertainty\n",
    "* Aleatoric/Data Uncertainty"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/dusk/64/000000/artificial-intelligence.png\" style=\"height:50px;display:inline\"> Epistemic/Model Uncertainty\n",
    "--- \n",
    "* Episteme in Latin is \"Knowledge/understanding\"\n",
    "* Also called model uncertaintly or reducible uncertainty\n",
    "* Describes the uncertainty in the model itself (network weights)\n",
    "* High when we do not have good coverage of the sample space\n",
    "* Can be reduced with more training data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "<img src=\"./assets/uncertainty_types3.png\" width=\"800\">\n",
    "\n",
    "* Cool online demo for epistemic uncertainty: https://mlg.eng.cam.ac.uk/yarin/blog_3d801aa532c1ce.html"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/ios/60/000000/area-chart.png\" style=\"height:50px;display:inline\"> Aleatoric/Data Uncertainty\n",
    "--- \n",
    "* Aleator in Latin means \"dice player\"\n",
    "\n",
    "<img src=\"./assets/uncertainty_types2.png\" width=\"400\">\n",
    "\n",
    "* Rolling a pair of dice again and again will not reduce the uncertainty\n",
    "* <a href=\"http://www.cs.ox.ac.uk/people/yarin.gal/website/bdl101/MLSS_2019_BDL_2.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/ios/60/000000/area-chart.png\" style=\"height:50px;display:inline\"> Aleatoric/Data Uncertainty\n",
    "--- \n",
    "* Also called data uncertainty or irreducible uncertainty\n",
    "* Describes the confidence in the input data\n",
    "* High when the input data is noisy\n",
    "* Can not be reduced by adding more training data\n",
    "* To decrease it, we need to acquire data with a better sensor (less noise)\n",
    "* Further classified into: \n",
    "    * homoscedastic - fixed for all samples\n",
    "    * heteroscedastic - sample-dependent noise"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Visual Example 1\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_types1.png\" width=\"1200\">\n",
    "\n",
    "* Data generation model is given by $y_i = \\omega_0 + \\omega_1 x_i + \\epsilon$\n",
    "* Here we have heteroscedastic data uncertainty since $\\epsilon = \\epsilon(x_i)$\n",
    "\n",
    "* <a href=\"https://towardsdatascience.com/my-deep-learning-model-says-sorry-i-dont-know-the-answer-that-s-absolutely-ok-50ffa562cb0b#:~:text=There%20are%20two%20major%20different,epistemic%20uncertainty%20and%20aleatoric%20uncertainty.&text=Epistemic%20uncertainty%20describes%20what%20the,to%20limited%20data%20and%20knowledge.\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Visual Example 2\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_types4.png\" width=\"1200\">\n",
    "\n",
    "* Extrapolation is an obvious case of high epistemic uncertainty\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/office/80/000000/abacus.png\" style=\"height:50px;display:inline\"> Computing Uncertainty\n",
    "---\n",
    "* Likelihood Modelling\n",
    "* Bayesian Neural Networks\n",
    "* Dropout at Test time\n",
    "* Evidental Deep Learning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/pastel-glyph/64/000000/increase-profits.png\" style=\"height:50px;display:inline\"> Likelihood Modelling\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_mle1.png\" width=\"1200\">\n",
    "\n",
    "* Main idea is simple: swap point estimate with parameterized likelihood\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/pastel-glyph/64/000000/increase-profits.png\" style=\"height:50px;display:inline\"> Likelihood Modelling: Classification\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_mle2.png\" width=\"1200\">\n",
    "\n",
    "* Classification: Output is a probability distribution with Softmax and Cross-entropy\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/pastel-glyph/64/000000/increase-profits.png\" style=\"height:50px;display:inline\"> Likelihood Modelling: Classification\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_mle3.png\" width=\"1200\">\n",
    "\n",
    "* Underlying assumption is that the output is Categorially distributed\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/pastel-glyph/64/000000/increase-profits.png\" style=\"height:50px;display:inline\"> Likelihood Modelling: Regression\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_mle4.png\" width=\"1200\">\n",
    "\n",
    "* Same logic only assume the output is Gaussian\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/pastel-glyph/64/000000/increase-profits.png\" style=\"height:50px;display:inline\"> Likelihood Modelling: Regression\n",
    "---\n",
    "\n",
    "* Where are the uncertainty \"labels\" coming from?\n",
    "    * We do not need labels for uncertainty!\n",
    "* The negative log-likelihood for a Gaussian distribution is composed of 2 terms ($\\hat{y}\\equiv \\mu$):\n",
    "$$-\\log\\left(\\mathcal{N}\\left(y\\lvert \\mu, \\sigma^2\\right)\\right) = \\frac{1}{2} \\frac{\\left(y - \\hat{y}\\left(x\\right)\\right)^2}{\\hat{\\sigma}^2\\left(x\\right)} + \\frac{1}{2} \\log\\left(\\hat{\\sigma}^2\\left(x\\right)\\right) + C$$\n",
    "* The first term favors very large uncertainty\n",
    "* The second term favors very small uncertainty\n",
    "* The compromise is reached such that for small errors the uncertainty is small\n",
    "* Note that this requires another head for the net (to output sigma)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/pastel-glyph/64/000000/increase-profits.png\" style=\"height:50px;display:inline\"> Likelihood Modelling: Summary\n",
    "---\n",
    "\n",
    "<img src=\"./assets/uncertainty_mle5.png\" width=\"1200\">\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/officel/40/000000/normal-distribution-histogram.png\" style=\"height:50px;display:inline\"> Bayesian Neural Networks\n",
    "---\n",
    "\n",
    "<img src=\"./assets/uncertainty_bnn5.png\" width=\"1200\">\n",
    "\n",
    "* Epistemic uncertainty is significantly harder to estimate\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/officel/40/000000/normal-distribution-histogram.png\" style=\"height:50px;display:inline\"> Bayesian Neural Networks: key idea\n",
    "---\n",
    "\n",
    "<img src=\"./assets/uncertainty_bnn4.png\" width=\"1200\">\n",
    "\n",
    "* Main idea: put probability distrbutions over weights\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/officel/40/000000/normal-distribution-histogram.png\" style=\"height:50px;display:inline\"> Bayesian Neural Networks: intractable directly\n",
    "---\n",
    "\n",
    "<img src=\"./assets/uncertainty_bnn6.png\" width=\"1200\">\n",
    "\n",
    "* Straighforward Bayesian inference is intractable\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/officel/40/000000/normal-distribution-histogram.png\" style=\"height:50px;display:inline\"> Bayesian Neural Networks: stochastic variational inference\n",
    "---\n",
    "\n",
    "* The underlying approximations rely on \"Stochastic Variational Inference\".\n",
    "* The posterior is approximated using a parametric distrbution (eg Gaussian)\n",
    "* Optimization is performed using the \"evidence lower bound\" (ELBO)\n",
    "* \"Monte Carlo\" (MC) Integration is involved to approximate the gradients.\n",
    "* Still becomes extremely heavy for very large models."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/officel/40/000000/normal-distribution-histogram.png\" style=\"height:50px;display:inline\"> Bayesian Neural Networks: Efficient approximations\n",
    "---\n",
    "\n",
    "<img src=\"./assets/uncertainty_bnn7.png\" width=\"1200\">\n",
    "\n",
    "* What? Dropout at test time? \n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"assets/dropout_icon.png\" style=\"height:50px;display:inline\"> Dropout at Inference \n",
    "--- \n",
    "\n",
    "* It can be proven that Dropout with L2 weight decay = approximate Bayesian inference of a Gaussian process.\n",
    "* Main idea: Dropout **both** at **training** and at **test** time.\n",
    "* Intuition: Training $2^m$ different models simultaneously, where $m$ is the number of nodes in the network. \n",
    "* For each batch, a randomly sampled set of these models is trained.\n",
    "\n",
    "<img src=\"assets/dropout_mc.gif\" style=\"height:500px\">\n",
    "\n",
    "* <a href=\"https://www.yuritan.nl/posts/prediction_uncertainty/\">Image source</a>\n",
    "* <a href=\"https://jmlr.org/papers/v15/srivastava14a.html\">Icon source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/cotton/64/000000/combo-chart--v1.png\" style=\"height:50px;display:inline\"> Evidental Deep Learning\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_el1.png\" width=\"1200\">\n",
    "\n",
    "* There is a need for a single-pass model that is able to provide both uncertainties \n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/cotton/64/000000/combo-chart--v1.png\" style=\"height:50px;display:inline\"> Evidental Deep Learning\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_el2.png\" width=\"1200\">\n",
    "\n",
    "* Each model from the ensemble leads to a single point $\\left(\\mu, \\sigma^2\\right)$\n",
    "* The model uncertainty is then given by the variance of $\\mu$ - $Var\\left[\\mu\\right]$\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/cotton/64/000000/combo-chart--v1.png\" style=\"height:50px;display:inline\"> Evidental Deep Learning\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_el3.png\" width=\"1200\">\n",
    "\n",
    "* Ultimately we are interested in the marginal distributions over $\\mu$ and $\\sigma$ \n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/cotton/64/000000/combo-chart--v1.png\" style=\"height:50px;display:inline\"> Evidental Deep Learning\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_el4.png\" width=\"1200\">\n",
    "\n",
    "* Key idea: Instead of training ensemble of models to output scalars $\\mu$ and $\\sigma$, output their distributions directly\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/cotton/64/000000/combo-chart--v1.png\" style=\"height:50px;display:inline\"> Evidental Deep Learning\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_el5.png\" width=\"1200\">\n",
    "\n",
    "* For Classification: The prior will be on the individual probabilities\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/cotton/64/000000/combo-chart--v1.png\" style=\"height:50px;display:inline\"> Evidental Deep Learning\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_el6.png\" width=\"1200\">\n",
    "\n",
    "* Key idea: Instead of training ensemble of models to output scalars, output nested distributions directly\n",
    "* The previous choices are not random, stem from \"conjugate priors\" in Bayesian analysis:\n",
    "    * If posterior and prior are in the same \"probability distribution family\" (eg exponential) then the prior is said to be conjugate to the likelihood.\n",
    "    * Examples include: Poisson Likelihood-Gamma Prior, Categorial Likelihood-Dirichlet Prior,...\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/cotton/64/000000/combo-chart--v1.png\" style=\"height:50px;display:inline\"> Evidental Deep Learning\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_el7.png\" width=\"1200\">\n",
    "\n",
    "* Pytorch implementation: https://github.com/dougbrion/pytorch-classification-uncertainty\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/ios/50/000000/summary-list.png\" style=\"height:50px;display:inline\"> Techniques - Summary\n",
    "---\n",
    "\n",
    "<img src=\"./assets/uncertainty_comparison.png\" width=\"1200\">\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/bubbles/100/000000/3d-glasses.png\" style=\"height:50px;display:inline\"> Computer Vision Applications\n",
    "--- \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/bubbles/100/000000/3d-glasses.png\" style=\"height:50px;display:inline\"> Application 1: Out-of-distribution detection\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_app1.png\" width=\"1200\">\n",
    "\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/bubbles/100/000000/3d-glasses.png\" style=\"height:50px;display:inline\"> Application 1: Out-of-distribution detection\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_app4.png\" width=\"1200\">\n",
    "\n",
    "* Standard Softmax on the left, Evidental deep learning on the right\n",
    "* Rotation angle effectively induces out-of-distribution samples\n",
    "* Erroneously high confidence with standard method\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/bubbles/100/000000/3d-glasses.png\" style=\"height:50px;display:inline\"> Application 1: Out-of-distribution detection\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_app6.png\" width=\"800\">\n",
    "\n",
    "* The model can easily tell that it is just flipping a coin\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/bubbles/100/000000/3d-glasses.png\" style=\"height:50px;display:inline\"> Application 2: Autonomous Driving\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_app3.png\" width=\"1200\">\n",
    "\n",
    "* Aleatoric uncertainty captures object boundaries where labels are noisy\n",
    "* Bottom row - failure case shows increase in epistemic uncertainty\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/bubbles/100/000000/3d-glasses.png\" style=\"height:50px;display:inline\"> Application 2: Autonomous Driving\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_app5.png\" width=\"1200\">\n",
    "\n",
    "* Multi-task Learning: Improves over learning each task individually\n",
    "* Improve generalization by sharing info between tasks: \"what is learned from one task can help learn another task\"\n",
    "* Optimal weighting between the importance of the different tasks given by the homoscedastic aleatoric uncertainty\n",
    "$$ \\mathcal{L}_{total} = \\sum\\limits_i \\alpha_i \\mathcal{L}_i \\approx \\frac{1}{2\\sigma_1^2} \\mathcal{L}_1 + \\frac{1}{2\\sigma_2^2} \\mathcal{L}_2 + \\log\\left(\\sigma_1\\right) + \\log\\left(\\sigma_2\\right)$$\n",
    "* <a href=\"https://openaccess.thecvf.com/content_cvpr_2018/html/Kendall_Multi-Task_Learning_Using_CVPR_2018_paper.html\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### <img src=\"https://img.icons8.com/bubbles/100/000000/3d-glasses.png\" style=\"height:50px;display:inline\"> Application 2: Autonomous Driving\n",
    "--- \n",
    "\n",
    "<img src=\"./assets/uncertainty_app2.png\" width=\"1200\">\n",
    "\n",
    "* Monocular depth estimation\n",
    "* 3D Object recognition in point clouds\n",
    "* Pose Estimation in SLAM\n",
    "* etc.\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Image source</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### <img src=\"assets/blitz_bnn3.png\" style=\"height:50px;display:inline\"> Blitz - Bayesian Neural Networks in PyTorch\n",
    "---\n",
    "<img src=\"assets/blitz_bnn.png\" style=\"height:400px\">\n",
    "\n",
    "* <a href=\"https://github.com/piEsposito/blitz-bayesian-deep-learning\">Blitz is a bayesian neural netowrk library for PyTorch</a>\n",
    "* Acompanies the ~<a href=\"http://proceedings.mlr.press/v37/blundell15.html\">first paper</a> that revived this field from ICML 2015. \n",
    "* Relatively easy to use, however, quite heavy for large models."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### <img src=\"assets/pyro_logo_wide.png\" style=\"height:50px;display:inline\"> Pyro - Deep Universal Probabilistic Programming\n",
    "---\n",
    "<img src=\"assets/pyro_gp.png\" style=\"height:400px\">\n",
    "\n",
    "* <a href=\"https://github.com/piEsposito/blitz-bayesian-deep-learning\">Pyro</a> is a deep probabilistic programming library for PyTorch\n",
    "* It has extensive tools, and can deal with a variety of applications including generative models etc."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "* Other available resources online:\n",
    "    * https://github.com/Harry24k/bayesian-neural-network-pytorch\n",
    "    * https://github.com/dougbrion/pytorch-classification-uncertainty\n",
    "    * https://github.com/cpark321/uncertainty-deep-learning\n",
    "    * https://github.com/ivannz/mlss2019-bayesian-deep-learning\n",
    "    * etc."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/color/96/000000/code.png\" style=\"height:50px;display:inline\"> Simple code example: BNN with dropout\n",
    "* Code based on example by <a href=\"https://github.com/cpark321\">Chanwoo Park</a>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "# imports\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import math\n",
    "import torch.optim as optim\n",
    "from torch.autograd import Variable\n",
    "import matplotlib.pyplot as plt\n",
    "from torchvision import datasets, transforms\n",
    "%matplotlib inline\n",
    "\n",
    "# GPU ID\n",
    "device = torch.device(\"cuda:0\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "# simple network model for 1D regression\n",
    "class SimpleModel(torch.nn.Module):\n",
    "    def __init__(self, dropout_rate, decay):\n",
    "        super(SimpleModel, self).__init__()\n",
    "        self.dropout_rate = dropout_rate\n",
    "        self.decay = decay\n",
    "        self.f = torch.nn.Sequential(\n",
    "            torch.nn.Linear(1,20),\n",
    "            torch.nn.ReLU(),\n",
    "            torch.nn.Dropout(p=self.dropout_rate),\n",
    "            torch.nn.Linear(20, 20),\n",
    "            torch.nn.ReLU(),\n",
    "            torch.nn.Dropout(p=self.dropout_rate),\n",
    "            torch.nn.Linear(20,1)\n",
    "        )\n",
    "    def forward(self, X):        \n",
    "        return self.f(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "# estimate uncertainty by MC dropout sampling\n",
    "def uncertainity_estimate(x, model, num_samples, l2):\n",
    "    outputs = np.hstack([model(x).cpu().detach().numpy() for i in range(num_samples)]) # n번 inference, output.shape = [20, N]\n",
    "    y_mean = outputs.mean(axis=1)\n",
    "    y_variance = outputs.var(axis=1)\n",
    "    tau = l2 * (1. - model.dropout_rate) / (2. * N * model.decay)\n",
    "    y_variance += (1. / tau)\n",
    "    y_std = np.sqrt(y_variance)\n",
    "    return y_mean, y_std"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "# generate data\n",
    "N = 200 ##  number of points\n",
    "min_value = -10\n",
    "max_value = 10\n",
    "\n",
    "\n",
    "x_obs = np.linspace(min_value, max_value, N)\n",
    "noise = np.random.normal(loc = 10, scale = 80, size = N)\n",
    "y_obs =  x_obs**3 + noise\n",
    "\n",
    "x_test = np.linspace(min_value - 10, max_value + 10, N)\n",
    "y_test = x_test**3 + noise\n",
    "\n",
    "# Normalise data:\n",
    "x_mean, x_std = x_obs.mean(), x_obs.std()\n",
    "y_mean, y_std = y_obs.mean(), y_obs.std()\n",
    "x_obs = (x_obs - x_mean) / x_std\n",
    "y_obs = (y_obs - y_mean) / y_std\n",
    "x_test = (x_test - x_mean) / x_std\n",
    "y_test = (y_test - y_mean) / y_std"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(16,8))\n",
    "plt.plot(x_obs, y_obs)\n",
    "plt.grid()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iter: 0, Loss: 0.9956\n",
      "Iter: 2000, Loss: 0.3633\n",
      "Iter: 4000, Loss: 0.2682\n",
      "Iter: 6000, Loss: 0.2930\n",
      "Iter: 8000, Loss: 0.2467\n",
      "Iter: 10000, Loss: 0.2529\n",
      "Iter: 12000, Loss: 0.2155\n",
      "Iter: 14000, Loss: 0.2215\n",
      "Iter: 16000, Loss: 0.2392\n",
      "Iter: 18000, Loss: 0.1714\n"
     ]
    }
   ],
   "source": [
    "# define setting and optimizer\n",
    "model = SimpleModel(dropout_rate=0.5, decay=1e-6).to(device)\n",
    "criterion  = torch.nn.MSELoss()\n",
    "optimizer = torch.optim.SGD(model.parameters(), lr=0.01, weight_decay=model.decay)\n",
    "\n",
    "# train the model\n",
    "for iter in range(20000):\n",
    "    y_pred = model(torch.Tensor(x_obs).view(-1,1).to(device))\n",
    "    y_true = Variable(torch.Tensor(y_obs).view(-1,1).to(device))\n",
    "    optimizer.zero_grad()\n",
    "    loss = criterion(y_pred, y_true)\n",
    "    loss.backward()\n",
    "    optimizer.step()\n",
    "    \n",
    "    if iter % 2000 == 0:\n",
    "        print(\"Iter: {}, Loss: {:.4f}\".format(iter, loss.item()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot the fit to the training data\n",
    "plt.figure(figsize=(16,8))\n",
    "y_pred = model(torch.Tensor(x_obs).view(-1,1).to(device))\n",
    "plt.plot(x_obs, y_obs, ls=\"none\", marker=\"o\", color=\"0.1\", alpha=0.8, label=\"observed\")\n",
    "plt.plot(x_obs, y_pred.cpu().detach().numpy(), ls=\"-\", color=\"b\", label=\"mean\")\n",
    "plt.grid()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# estimate uncertainty\n",
    "iters_uncertainty = 200\n",
    "lengthscale = 0.01\n",
    "n_std = 3 # number of standard deviations to plot\n",
    "y_mean, y_std = uncertainity_estimate(torch.Tensor(x_test).view(-1,1).to(device), model, iters_uncertainty, lengthscale)\n",
    "\n",
    "# plot the confidence intervals\n",
    "plt.figure(figsize=(16,8))\n",
    "plt.plot(x_obs, y_obs, ls=\"none\", marker=\"o\", color=\"0.1\", alpha=0.8, label=\"observed\")\n",
    "plt.plot(x_test, y_mean, ls=\"-\", color=\"b\", label=\"mean\")\n",
    "plt.plot(x_test, y_test, ls='--', color='r', label='true')\n",
    "for i in range(n_std):\n",
    "    plt.fill_between( x_test,\n",
    "        y_mean - y_std * ((i+1.)),\n",
    "        y_mean + y_std * ((i+1.)),\n",
    "        color=\"b\",\n",
    "        alpha=0.1)\n",
    "plt.legend(fontsize=20)\n",
    "plt.grid()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### <img src=\"https://img.icons8.com/bubbles/50/000000/video-playlist.png\" style=\"height:50px;display:inline\"> Recommended Videos\n",
    "---\n",
    "#### <img src=\"https://img.icons8.com/cute-clipart/64/000000/warning-shield.png\" style=\"height:30px;display:inline\"> Warning!\n",
    "* These videos do not replace the lectures and tutorials.\n",
    "* Please use these to get a better understanding of the material, and not as an alternative to the written material.\n",
    "\n",
    "#### Video By Subject\n",
    "* Bayesian Neural Networks (Theory)\n",
    "    * Variational Inference and ELBO - <a href=\"https://www.youtube.com/watch?v=G6tUZRHnJYc\">Yarin Gal - Part I</a>\n",
    "    * Dropout as efficient Bayesian Inference - <a href=\"https://www.youtube.com/watch?v=7p56lnNi74s&t=5s&ab\">Yarin Gal - Part II</a>\n",
    "* Bayesian Neural Networks vs Dropout (Practice)\n",
    "    * MNIST demonstration - <a href=\"https://www.youtube.com/watch?v=Puc_ujh5QZs&t=739s&ab\"> Elise Jennings, Argonne National Laboratory</a>\n",
    "* Evidental Deep Learning - <a href=\"https://www.youtube.com/watch?v=toTcf7tZK8c&ab\"> MIT 6.S191 - Lecture 7</a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "## <img src=\"https://img.icons8.com/dusk/64/000000/prize.png\" style=\"height:50px;display:inline\"> Credits\n",
    "----\n",
    "* <a href=\"http://mlg.eng.cam.ac.uk/yarin/blog_3d801aa532c1ce.html\">What My Deep Model Doesn't Know...</a> - Yarin Gal\n",
    "* <a href=\"http://mlg.eng.cam.ac.uk/yarin/blog_3d801aa532c1ce.html\">Deep Learning Is Not Good Enough,\n",
    "We Need Bayesian Deep Learning for Safe AI</a> - Alex Kendall\n",
    "* <a href=\"http://www.cs.ox.ac.uk/people/yarin.gal/website/bdl101/\">Bayesian Deep Learning 101 @ MLSS19</a> - Yarin Gal\n",
    "* <a href=\"http://introtodeeplearning.com/slides/6S191_MIT_DeepLearning_L7.pdf\">Slides Lecture 7 MIT 6.S191</a> (Introduction to Deep Learning) -  Alexander Amini\n",
    "* <a href=\"https://towardsdatascience.com/my-deep-learning-model-says-sorry-i-dont-know-the-answer-that-s-absolutely-ok-50ffa562cb0b#:~:text=There%20are%20two%20major%20different,epistemic%20uncertainty%20and%20aleatoric%20uncertainty.&text=Epistemic%20uncertainty%20describes%20what%20the,to%20limited%20data%20and%20knowledge.\"> Uncertainty in Deep Learning. How to Measure?</a> - Michel Kana\n",
    "* <a href=\"#\">Uncertainty Estimation in Deep Learning</a> (Slideshare) - Christian Perone\n",
    "* Research papers:\n",
    "    * <a href=\"http://proceedings.mlr.press/v37/blundell15.html\">Weight Uncertainty in Neural Networks</a>\n",
    "    * <a href=\"http://proceedings.mlr.press/v48/gal16.html\">Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning</a>\n",
    "    * <a href=\"https://papers.nips.cc/paper/2017/hash/2650d6089a6d640c5e85b2b88265dc2b-Abstract.html\">What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?</a>\n",
    "    * <a href=\"https://openaccess.thecvf.com/content_cvpr_2018/html/Kendall_Multi-Task_Learning_Using_CVPR_2018_paper.html\">Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics</a>\n",
    "    * <a href=\"https://papers.nips.cc/paper/2018/hash/a981f2b708044d6fb4a71a1463242520-Abstract.html\">Evidential Deep Learning to Quantify Classification Uncertainty</a>\n",
    "    * <a href=\"https://openaccess.thecvf.com/content_CVPR_2019/html/Barron_A_General_and_Adaptive_Robust_Loss_Function_CVPR_2019_paper.html\">A General and Adaptive Robust Loss Function</a>\n",
    "    * <a href=\"https://www.nips.cc/virtual/2020/public/poster_aab085461de182608ee9f607f3f7d18f.html\">Deep Evidential Regression</a>\n",
    "\n",
    "* Icons from <a href=\"https://icons8.com/\">Icon8.com</a> - https://icons8.com\n"
   ]
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "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": "3.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}