Sitemap
Press enter or click to view image in full size

About Machine learning

4 min readJan 10, 2024
Press enter or click to view image in full size

Hey there! So, I wanted to share this cool story about machine learning, which is like teaching computers to be super smart. It’s part of this big, brainy thing called artificial intelligence. We use it to make computers learn from data and do smart stuff. There are different types, like the one where it learns from labeled data (we know the answers) or the one where it figures things out on its own. It’s like teaching a robot how to do tricks, and it’s super awesome! We also have some challenges in making sure everything is fair and nice. Anyway, I hope you enjoy my story from my final year machine learning module short note!

About Machine learning

The subfield of artificial intelligence application optimisation.

Support Other fields in CS

  • Vision
  • Data Mining
  • OCR
  • NLP

AI has two types of branches

  • algorithmic branch
  • game theory
  • optimization
  • data-driven(ML)
  • using real-word dataset
  • using a dataset and trying to learn certain patterns.

Why is ML now important??

  1. Available data.
  2. Hardware advancement (computational powers).
  3. Algorithms.
  4. Frameworks.
  5. Theoretical knowledge is developed by researchers.
  6. Cloud computing
Press enter or click to view image in full size

Terminology

  • Features.

that can be used to describe each item quantitatively.

  • Samples.

A sample is an item to process.

  • Feature vector.

numerical features that represent some objects.

  • Feature extraction.

preparation of the feature vector.

transforms the data in the high-dimensional space into a space of fewer dimensions.

  • Training/Evolution set

Set of data to discover potentially predictive relationships.

Press enter or click to view image in full size

ML Categories

  • Supervised Learning
  • Unsupervised Learning
  • Reinforcement Learning
  • Semi-Supervised Learning
  • Bayesian learning

Supervised Learning (Classification)

the correct classes (labels) of the training data are known

  • Used when the dataset has classes or labels
  • Includes a ‘training’ phase with the dataset and a ‘testing’ phase to validate the accuracy of the classifier
  • Algorithms –
  • Regression
  • Support Vector Machines
  • Neural Networks
  • Convolutional Neural Networks,
  • Decision Trees
  • Logistic Regression
  • Random Forest
  • Naïve Bayesian

Supervised learning examples

  • A bank may have borrower details (age, income, gender, etc.) from the past (features).
  • Also, it may have details of the borrowers who defaulted in the past (labels).
  • Based on the above, we can train a classifier to learn the patterns of borrowers who are likely to default on their payments.

Supervised learning types

  • Regression: predict continuous variables (salary, rent)
  • Binary classification (facial recognition, whether a tumor is benign or malignant)
  • Multi-class classification (the type of vehicle, the stage of progression of cancer — level 1, 2, and 3)

Linear regression

Press enter or click to view image in full size

Unsupervised learning

  • Used when the dataset does not have the labels (classes)
  • Used to group or cluster the data into clusters, which may then be used for decision-making, making recommendations, classification, etc.

Algorithms

  • K-means
  • Self-organizing maps
  • Deep belief Networks

Unsupervised learning examples

  • A Supermarket may store each buyer’s basket content details (features)
  • There is NO grouping (labels)
  • Need to group the buyers based on their buying patterns to best use the shelf space (recommendation)

Unsupervised learning types

  • K-means clustering
  • Self-organizing maps
  • Deep Belief Networks

Reinforcement Learning

Allows the machine or software agent to learn its behavior based on feedback or the environment.

Can be learned once and for all or keep on adapting as time goes by.

This can be used for:

  • No data is available.
  • When the environment is dynamically changing.
  • The reward function is used to measure the reward for a given action.
  • Based on the reward values.
  • Can be continued over time.
  • Can be deployed in both single-agent and multiple-agent systems.

What are using Algorithms??

  • Actor critic learning
  • Q learning
  • Montecarlo methods

Reinforcement learning examples

  • Can apply robots
  • Autonomous vehicles
  • Multi-agent system

Semi-Supervised Learning

It is a mix of supervised and unsupervised learning.

Labeled data is expensive and difficult to get.

Unlabeled data is cheaper and easier to get.

Algorithms

  • Self trying
  • Generative models
  • Semi-supervised support vector machines

Semi-supervised learning applications

Web page classification

Speech-to-text conversion

Video/image generation

Issues and Challenges in ML

  • Bias
  • Interpretability/Explainability
  • Generalisability
  • Ethical/Legal Issues
  • Social issues
  • Data availability
  • ‘External Factors’ that are not captured in historical data

Follow me on GitHub: MadhushaPrasad

--

--

MadhushaPrasad
MadhushaPrasad

Written by MadhushaPrasad

Open Source Enthusiast | Full Stack Developer👨🏻‍💻 | JavaScript & Type Script , Git & GitHub Lover | Undergraduate — Software Engineering‍💻 | SLIIT👨🏻‍🎓

No responses yet