Notebook_1_Python_and_NumPy.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# TensorFlow for ICT Applications\n",
    "\n",
    "## Lab session 1: introduction to Python and NumPy"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Some generic Python concepts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Importing a module and using it\n",
    "import math\n",
    "math.cos(3.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Lists are defined by squared brackets\n",
    "mylist = [2, -1, 3]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# You can index lists, but the indexing starts from zero!\n",
    "mylist[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Tuples are another fundamental data type\n",
    "mytyple = (3, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Finally we have dictionaries, where each element is indexed by a string\n",
    "mydict = {\n",
    "    'param1': 1,\n",
    "    'param2': 'a'\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Indexing of a dictionary is similar to a list, but requires the correct key\n",
    "mydict['param1']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Example of if in Python; the main block is indented!\n",
    "if len(mylist) > 2:\n",
    "    print('A very long list!')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Loops in Python work on everything which is iterable:\n",
    "for el in mylist:\n",
    "    print(el)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# List comprehensions are an easy way of manipulating lists\n",
    "mylist_2 = [2 * el for el in mylist]\n",
    "print(mylist_2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Manipulating matrices: NumPy arrays"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Several ways of generating matrices:\n",
    "A = np.array(mylist)\n",
    "B = np.random.rand(3, 2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Some functions requires a tuple for the shape!\n",
    "# Also note that (5,1) is a different shape from (5,)\n",
    "C = np.ones((5, 1)) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(B)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Each array is defined by its shape and the data type of its elements\n",
    "print(B.shape)\n",
    "print(B.dtype)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Basic indexing\n",
    "print(B[0, 0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Indexing with strides\n",
    "# Generic syntax: start:step:end\n",
    "print(B[0:3, 0:2:2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Backward indexing (starts from the end)\n",
    "print(B[-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Getting statistics from arrays\n",
    "print(np.sum(B, axis=0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Linear algebra is in the 'linalg' submodule\n",
    "np.linalg.pinv(B)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Some more matrices\n",
    "a = np.linspace(0, 1, 10)\n",
    "B = np.ones((2, 10))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Operators are element-wise in NumPy\n",
    "a * a"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Dot product\n",
    "a.dot(a)\n",
    "# Or a @ a"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Broadcasting\n",
    "a * B"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Reading material\n",
    "* A comprehensive set of lectures about scientific computing with Python: https://github.com/jrjohansson/scientific-python-lectures.\n",
    "* Very good introductory tutorial on NumPy: https://github.com/gertingold/euroscipy-numpy-tutorial."
   ]
  }
 ],
 "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": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}