Machine Learning: Basis and Wavelet - WordPress.com€¦ · · 2017-03-102017-03-10 · Humans can typically create one or two good models a week; machine learning can ... 2,4 (𝑥𝑥)
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Machine Learning: Basis and Wavelet
Haar DWT in 2 levels
𝒇𝒇
7 22 38 191
17 83 188 211
71 167 194 207
159 187 201 216
-20 -44
-31 -7
135
-40 -17
-46
13 -32
-17 1
-18 -42
-27 -4
32 157
146 204
김 화 평 (CSE ) Medical Image computing lab 서진근교수 연구실
+ +
- -
+ -
+ -
+ -
- +
+ +
+ +
Machine learning is the field of study that gives computers the ability to learn the feed-forward function without being explicitly programmed.
Supervised Learning Mission: Find a feed-forward function 𝒇𝒇 𝒙𝒙 = 𝒚𝒚 from labeled training data, {( 𝒙𝒙
(𝒊𝒊) ,𝒚𝒚
(𝒊𝒊)): 𝒊𝒊 = 𝟏𝟏, … ,𝒎𝒎}, such that 𝒇𝒇 𝒙𝒙 (𝒊𝒊) = 𝒚𝒚
(𝒊𝒊), 𝒊𝒊 = 𝟏𝟏, … ,𝒎𝒎.
Supervised learning is the machine learning technique of finding a feed-forward function 𝑓𝑓 𝒙𝒙 = 𝒚𝒚 iteratively from labeled training data, {( 𝒙𝒙
(𝑖𝑖) ,𝒚𝒚
(𝑖𝑖)): 𝑖𝑖 = 1, … ,𝑚𝑚}, such that 𝑓𝑓 𝒙𝒙 (𝑖𝑖) = 𝒚𝒚
(𝑖𝑖), 𝑖𝑖 = 1, … ,𝑚𝑚.
Machine learning: Why it is and why it matters. Humans can typically create one or two good models a week; machine learning can create thousands of models a week. Machine is faster! (Not smarter)
In big and complex data in the real world, it is very difficult to make explicit program to find all hidden insights. ML is possible to quickly and automatically produce models that can analyze bigger, more complex data and deliver faster, more accurate results – even on a very large scale.
Basis: Fourier Transform Basis
The Fourier transform of 𝑓𝑓 is defined by ℱ𝑓𝑓 𝜉𝜉 = ∫ 𝑓𝑓(𝑡𝑡)∞
−∞ 𝑒𝑒−2𝜋𝜋𝑖𝑖𝜋𝜋𝜋𝜋 𝑑𝑑𝑡𝑡. Each fourier transform acts as a basis to demonstrate the ability to distinguish different signals.
Every function can be expressed as a linear combination of basis functions 𝑓𝑓 = ∑𝑐𝑐𝑗𝑗𝜙𝜙𝑗𝑗,
where 𝜙𝜙0,𝜙𝜙1,⋯ is a set of orthonormal basis < 𝜙𝜙𝑛𝑛,𝜙𝜙𝑚𝑚 > = �1 𝑖𝑖𝑓𝑓 𝑛𝑛 = 𝑚𝑚,0 𝑖𝑖𝑓𝑓 𝑛𝑛 ≠ 𝑚𝑚.
Approximation by 4 principal components (basis) only
Wavelet basis functions: The family of functions {𝝍𝝍𝒋𝒋,𝒌𝒌: 𝒋𝒋,𝒌𝒌 ∈ ℤ}, dyadic translations and dilations of a mother wavelet function 𝝍𝝍, construct a complete orthonormal Hilbert basis.
Suppose we would like to apply lossy compression to a collection of m points 𝒙𝒙 𝟏𝟏 ,⋯ ,𝒙𝒙 𝒎𝒎 ⊂ 𝑹𝑹𝒏𝒏. Lossy compression means storing the points in a way that requires less memory but may lose some precision.
𝒙𝒙 𝟏𝟏 ,⋯ ,𝒙𝒙 𝟑𝟑𝟑𝟑 ⊂ 𝑹𝑹𝒏𝒏
𝐧𝐧 = 𝟑𝟑𝟑𝟑𝟏𝟏 × 𝟑𝟑𝟐𝟐𝟏𝟏
𝒙𝒙 𝟏𝟏
𝒙𝒙 𝟑𝟑𝟑𝟑 Slide Credit: Vaclav
Approximation by 4 principal components only
High-dimensional data 𝒙𝒙 𝒊𝒊 ’s often lies on or near a much lower dimensional, curved manifold. A good way to represent data points is by low-dimensional coordinates 𝑹𝑹𝒅𝒅 . The low-dimensional representation of the data should capture information about high-dimensional pairwise distance.
Let f:𝒙𝒙 ∈ Rn → 𝒄𝒄 ∈ 𝑅𝑅𝑙𝑙 l ≤ n be an encoding function which represents each data point x by a point c = f(x) in the low-dimensional space Rl. PCA is defined by our choice of the decoding function g: 𝒄𝒄 ∈ Rl → 𝒙𝒙 ∈ 𝑅𝑅𝑛𝑛 such that g f 𝒙𝒙 ≈ 𝒙𝒙. Let g(c) = Dc where D ∈ Rn×l defines the decoding. PCA constraints the columns of D to be orthonormal vectors in Rn.
This optimization problem may be solved using eigenvalue decomposition. Specifically, 𝒅𝒅∗ is given by the eigenvector of 𝑿𝑿𝑻𝑻𝑿𝑿 corresponding to the largest eigenvalue.
32nd row
1st row
= 𝒅𝒅∗= 𝒅𝒅𝒓𝒓𝒊𝒊 𝒎𝒎𝒊𝒊𝒏𝒏𝒅𝒅
𝑿𝑿 − 𝑿𝑿𝒅𝒅𝒅𝒅𝑻𝑻 𝑭𝑭𝟑𝟑
The first principle component
Slide Credit: Vaclav
More detailed explanation in computing the first principle component d∗