forked from nyu-dl/Intro_to_ML_Lecture_Note
-
Notifications
You must be signed in to change notification settings - Fork 0
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
1 parent
25cdcf5
commit 7414f9b
Showing
1 changed file
with
268 additions
and
0 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,268 @@ | ||
\documentclass{article} | ||
\usepackage{authblk} | ||
\usepackage{mathptmx} | ||
\usepackage{url,latexsym,amsmath,amsthm,xspace,rotating,multirow,multicol,xspace,amssymb,paralist} | ||
\usepackage{euscript} | ||
\usepackage{fancybox,xcolor} | ||
\usepackage{longtable} | ||
\usepackage{paralist} | ||
\usepackage[normalem]{ulem} | ||
\usepackage[pdftex]{hyperref} | ||
\usepackage{algorithmicx} | ||
\usepackage{algpseudocode} | ||
\usepackage{algorithm} | ||
\usepackage{cancel} | ||
\usepackage{mathtools} | ||
|
||
\usepackage{url} | ||
\usepackage{latexsym} | ||
|
||
\usepackage{times} | ||
\usepackage{amsmath} | ||
\usepackage{amsthm} | ||
\usepackage{amssymb} | ||
\usepackage{graphicx} | ||
\usepackage{xspace} | ||
\usepackage{tabularx} | ||
\usepackage{multicol} | ||
\usepackage{multirow} | ||
%\usepackage{hyperref} | ||
\usepackage{url} | ||
%\usepackage{natbib} | ||
\usepackage{wrapfig} | ||
\usepackage{comment} | ||
\usepackage{listings} | ||
\usepackage{color} | ||
\usepackage[utf8]{inputenc} | ||
\usepackage{fancyvrb} | ||
\usepackage{booktabs} | ||
\usepackage{color} | ||
\usepackage[normalem]{ulem} | ||
|
||
\newcommand{\obs}{\text{obs}} | ||
\newcommand{\mis}{\text{mis}} | ||
|
||
\newcommand{\qt}[1]{\left<#1\right>} | ||
\newcommand{\ql}[1]{\left[#1\right]} | ||
\newcommand{\hess}{\mathbf{H}} | ||
\newcommand{\jacob}{\mathbf{J}} | ||
\newcommand{\hl}{HL} | ||
\newcommand{\cost}{\mathcal{L}} | ||
\newcommand{\lout}{\mathbf{r}} | ||
\newcommand{\louti}{r} | ||
\newcommand{\outi}{y} | ||
\newcommand{\out}{\mathbf{y}} | ||
\newcommand{\gauss}{\mathbf{G_N}} | ||
\newcommand{\eye}{\mathbf{I}} | ||
\newcommand{\softmax}{\text{softmax}} | ||
\newcommand{\targ}{\mathbf{t}} | ||
\newcommand{\metric}{\mathbf{G}} | ||
\newcommand{\sample}{\mathbf{z}} | ||
\newcommand{\f}{\text{f}} | ||
%\newcommand{\log}{\text{log}} | ||
|
||
\newcommand{\bmx}[0]{\begin{bmatrix}} | ||
\newcommand{\emx}[0]{\end{bmatrix}} | ||
\newcommand{\qexp}[1]{\left<#1\right>} | ||
\newcommand{\vect}[1]{\mathbf{#1}} | ||
\newcommand{\vects}[1]{\boldsymbol{#1}} | ||
\newcommand{\matr}[1]{\mathbf{#1}} | ||
\newcommand{\var}[0]{\operatorname{Var}} | ||
\newcommand{\std}[0]{\operatorname{std}} | ||
\newcommand{\cov}[0]{\operatorname{Cov}} | ||
\newcommand{\diag}[0]{\operatorname{diag}} | ||
\newcommand{\matrs}[1]{\boldsymbol{#1}} | ||
\newcommand{\va}[0]{\vect{a}} | ||
\newcommand{\vb}[0]{\vect{b}} | ||
\newcommand{\vc}[0]{\vect{c}} | ||
\newcommand{\ve}[0]{\vect{e}} | ||
|
||
\newcommand{\vh}[0]{\vect{h}} | ||
\newcommand{\vv}[0]{\vect{v}} | ||
\newcommand{\vx}[0]{\vect{x}} | ||
\newcommand{\vp}[0]{\vect{p}} | ||
\newcommand{\vz}[0]{\vect{z}} | ||
\newcommand{\vw}[0]{\vect{w}} | ||
\newcommand{\vs}[0]{\vect{s}} | ||
\newcommand{\vf}[0]{\vect{f}} | ||
\newcommand{\vi}[0]{\vect{i}} | ||
\newcommand{\vo}[0]{\vect{o}} | ||
\newcommand{\vd}[0]{\vect{d}} | ||
\newcommand{\vy}[0]{\vect{y}} | ||
\newcommand{\vg}[0]{\vect{g}} | ||
\newcommand{\vm}[0]{\vect{m}} | ||
\newcommand{\vu}[0]{\vect{u}} | ||
\newcommand{\vL}[0]{\vect{L}} | ||
\newcommand{\vr}[0]{\vect{r}} | ||
\newcommand{\vone}[0]{\vect{1}} | ||
|
||
\newcommand{\mW}[0]{\matr{W}} | ||
\newcommand{\mE}[0]{\matr{E}} | ||
\newcommand{\mG}[0]{\matr{G}} | ||
\newcommand{\mX}[0]{\matr{X}} | ||
\newcommand{\mY}[0]{\matr{Y}} | ||
\newcommand{\mQ}[0]{\matr{Q}} | ||
\newcommand{\mU}[0]{\matr{U}} | ||
\newcommand{\mF}[0]{\matr{F}} | ||
\newcommand{\mV}[0]{\matr{V}} | ||
\newcommand{\mA}{\matr{A}} | ||
\newcommand{\mC}{\matr{C}} | ||
\newcommand{\mD}{\matr{D}} | ||
\newcommand{\mL}[0]{\matr{L}} | ||
\newcommand{\mR}[0]{\matr{R}} | ||
\newcommand{\mS}{\matr{S}} | ||
\newcommand{\mI}{\matr{I}} | ||
\newcommand{\td}[0]{\text{d}} | ||
\newcommand{\TT}[0]{\vects{\theta}} | ||
\newcommand{\vsig}[0]{\vects{\sigma}} | ||
\newcommand{\valpha}[0]{\vects{\alpha}} | ||
\newcommand{\vmu}[0]{\vects{\mu}} | ||
\newcommand{\vzero}[0]{\vect{0}} | ||
\newcommand{\tf}[0]{\text{m}} | ||
\newcommand{\tdf}[0]{\text{dm}} | ||
\newcommand{\grad}[0]{\nabla} | ||
\newcommand{\alert}[1]{\textcolor{red}{#1}} | ||
\newcommand{\N}[0]{\mathcal{N}} | ||
\newcommand{\YY}[0]{\mathcal{Y}} | ||
\newcommand{\BB}[0]{\mathcal{B}} | ||
\newcommand{\LL}[0]{\mathcal{L}} | ||
\newcommand{\HH}[0]{\mathcal{H}} | ||
\newcommand{\RR}[0]{\mathbb{R}} | ||
\newcommand{\MM}[0]{\mathcal{M}} | ||
\newcommand{\OO}[0]{\mathbb{O}} | ||
\newcommand{\II}[0]{\mathbb{I}} | ||
\newcommand{\Scal}[0]{\mathcal{S}} | ||
\newcommand{\sigmoid}{\sigma} | ||
\newcommand{\sign}{\text{sign}} | ||
\newcommand{\E}[0]{\mathbb{E}} | ||
\newcommand{\enabla}[0]{\ensuremath{% | ||
\overset{\raisebox{-0.3ex}[0.5ex][0ex]{% | ||
\ensuremath{\scriptscriptstyle e}}}{\nabla}}} | ||
\newcommand{\enhnabla}[0]{\nabla_{\hspace{-0.5mm}e}\,} | ||
\newcommand{\eos}[0]{\ensuremath{\left< \text{eos}\right>}} | ||
|
||
|
||
\newcommand{\todo}[1]{{\Large\textcolor{red}{#1}}} | ||
\newcommand{\done}[1]{{\Large\textcolor{green}{#1}}} | ||
\newcommand{\dd}[1]{\ensuremath{\mbox{d}#1}} | ||
|
||
\DeclareMathOperator*{\argmax}{\arg \max} | ||
\DeclareMathOperator*{\argmin}{\arg \min} | ||
\newcommand{\newln}{\\&\quad\quad{}} | ||
|
||
\newcommand{\BP}{\text{BP}} | ||
\newcommand{\PPL}{\text{PPL}} | ||
\newcommand{\PL}{\text{PL}} | ||
\newcommand{\MatSum}{\text{MatSum}} | ||
\newcommand{\MatMul}{\text{MatMul}} | ||
\newcommand{\KL}{\text{KL}} | ||
\newcommand{\data}{\text{data}} | ||
\newcommand{\rect}{\text{rect}} | ||
\newcommand{\maxout}{\text{maxout}} | ||
\newcommand{\train}{\text{train}} | ||
\newcommand{\hinge}{\text{hinge}} | ||
\newcommand{\val}{\text{val}} | ||
\newcommand{\init}{\text{init}} | ||
\newcommand{\fenc}{\text{fenc}} | ||
\newcommand{\renc}{\text{renc}} | ||
\newcommand{\enc}{\text{enc}} | ||
\newcommand{\dec}{\text{dec}} | ||
\newcommand{\test}{\text{test}} | ||
\newcommand{\tra}{\text{tra}} | ||
\newcommand{\Ax}{\mathcal{A}_x} | ||
\newcommand{\Ay}{\mathcal{A}_y} | ||
\newcommand{\ola}{\overleftarrow} | ||
\newcommand{\ora}{\overrightarrow} | ||
\newcommand{\ov}{\overline} | ||
\newcommand{\ts}{\rule{0pt}{2.6ex}} % Top strut | ||
\newcommand{\ms}{\rule{0pt}{0ex}} % Middle strut | ||
\newcommand{\bs}{\rule[-1.2ex]{0pt}{0pt}} % Bottom strut | ||
\newcommand{\specialcell}[2][c]{% | ||
\begin{tabular}[#1]{@{}c@{}}#2\end{tabular}} | ||
|
||
|
||
%\usepackage{bibentry} | ||
%\nobibliography* | ||
|
||
\begin{document} | ||
|
||
\title{Homework Assignment 6} | ||
\author{Lecturer: Kyunghyun Cho} | ||
%\affil{ | ||
% Courant Institute of Mathematical Sciences and \\ | ||
% Center for Data Science,\\ | ||
% New York University | ||
%} | ||
|
||
\maketitle | ||
\pagenumbering{arabic} | ||
|
||
\paragraph{1.} | ||
|
||
(a) Why is it more efficient to process data points if they are lower-dimensional | ||
vectors? State one reason. | ||
|
||
\vspace{5mm} | ||
|
||
(b) What is a potential trouble of reducing the dimensionality of input vectors | ||
before training a classifier? State one reason. | ||
|
||
\vspace{5mm} | ||
|
||
\paragraph{2.} | ||
|
||
(a) Given a training set $D=\left\{ \vx_1, \ldots, \vx_N \right\}$, show that the | ||
reconstruction error of principal component analysis (PCA) could be written down | ||
as | ||
\begin{align*} | ||
\frac{1}{N} \sum_{n=1}^N \| \vx_n - \hat{\vx_n} \|^2_2 = | ||
\sum_{j=q+1}^d \vw_i^\top \mC \vw_i, | ||
\end{align*} | ||
where $\vw_i$ is the $i$-th principal component or the eigenvector of the input | ||
covariance matrix $\mC$. | ||
|
||
|
||
\vspace{15mm} | ||
|
||
(b) Show that | ||
\begin{align*} | ||
&\Sigma = \mW^\top \mC \mW \\ | ||
\iff& | ||
\sigma_j^2 = \vw_j^\top \mC \vw_j,\text{ for all }j=1,\ldots,d, | ||
\end{align*} | ||
where $\mW$ is the weight matrix of PCA, $\mC$ is the input covariance | ||
matrix, and | ||
\begin{align*} | ||
\Sigma = \diag(\sigma_1^2, \ldots, \sigma_q^2) = | ||
\left[ | ||
\begin{array}{c c c c} | ||
\sigma_1^2 & 0 & \cdots & 0 \\ | ||
0 & \sigma2^2 & \cdots & 0 \\ | ||
\vdots & 0 & \cdots & \vdots \\ | ||
\vdots & \vdots & \cdots & \vdots \\ | ||
0 & 0 & \cdots & \sigma_q^2 | ||
\end{array} | ||
\right] | ||
\end{align*} | ||
is the covariance matrix of the code vectors. | ||
|
||
\vspace{30mm} | ||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
|
||
%\bibliographystyle{abbrv} | ||
%\bibliography{../lecture_note} | ||
|
||
|
||
\end{document} | ||
|
||
|
||
|
||
|
||
|
||
|