Basics of Machine Learning
Jul 04, 2015
Basics of Machine Learning
Contents
Definition of Machine Learning
Unsupervised & Supervised Learning
Types of Unsupervised learning
Manifolds
LLE Algorithm
Definition of Machine learning
It is a branch of Artificial Intelligence, concerns the construction and study of systems that can learn from given data.
Dataset consists of data; data means it is a form of matrix.
In matrix rows are nothing but examples & columns are attributes of examples.
How pixels are stored as no’s in images ?
In images pixels will be used as no’s, if suppose an image is given of size 120*120, then the product will be 14,400 pixels.
Each pixel value will have 0 – 255 numbers.
If there are 25 images the matrix size is 25*14,400 pixels.
Pixels will be said based on intensity values 0 – Black1 – White
Gray scale
It is pronounced as ‘Grey Scale’.
These are also called ‘Monochromatic’
Grayscale is an image in which the value of each pixel is a single sample, that is it carries only intensity information.
Images of this sort, also known as black-and-white, are composed exclusively of shades of gray, varying from black at the Weakest intensity to white at strongest.
Supervised vs Unsupervised Learning
In theoretical point of view both differ only in the casual structure of the model.
Advantage of Unsupervised Learning
With unsupervised learning, it is possible to learn larger and more complex models than with supervised learning.
Unlabeled: This data might include photos, videos, audio recordings, etc. There is no explanation for each piece of unlabeled data – it just contains the data, and nothing else.
Labeled: This data typically takes a patch of unlabeled data & augments each piece of that unlabeled data with some sort of meaningful “tag”.
Two types of Unsupervised Learning
1. Dimensionality Reduction
2. Density Estimation
What is topology?
Topology is relationship between the points, “Location of point w.r.t another point around it.”
Topology means distances.
Example: Let us take points A,B,C
C ->>>>> 10 m ->>>>> A ->>>>> 5 m ->>>>> B (In High Dimension)
C ->>>>> 1 m ->>>>> A ->>>>> 0.5m ->>>>> B (In Low Dimension)
Dimensionality Reduction Types
1. Linear Method
(a) PCA – Principal Component Analysis
(b) MDS – Multi Dimensional Scaling
2. Non-Linear Method
(a) ISOMAP
(b) LLE – Locally Linear Embedding
Advantages of Dimensionality Reduction
Reduce Time complexity
Reduce Space complexity
More interpretable
Manifolds
“According to mathematics, it is a collection of points forming a certain kind of set, such as those of topologically closed surface.”
Example: Surface, Curve & point.
A Manifold has a dimension.
“A Manifold embedded in n-dimensional Euclidian space locally look like (n-1) dimensional vector space.”
LLE - Locally Linear Embedding
Main Aim of LLE is to convert high dimensional inputs to low dimensional outputs.
It is a Eigen vector method.
LLE is capable of generating highly non-linear embedding's.
In LLE, the transformation is non-linear.
In mathematics, linear in the sense no polynomials are involved in ‘X’.i.e. X^2, X^3 etc….
LLE Algorithm - Steps
Step – 1: Compute the neighbors of each data point, 𝑋𝑖
Step – 2: Compute the weights 𝑊𝑖𝑗
Step – 3: Compute the vectors 𝑌𝑖
Conversion of High Dimension to Low Dimension
Thank you
Presented by : Ch. Satya Pranav,
KL University