Introduction to Artificial Intelligence
| Natural Intelligence | Artificial Intelligence |
|---|---|
| the task is done by human using Intelligence | if we add this human natural intelligence to a machine |
Artificial Intelligence is
The Science and Engineering of making Intelligent Machines ~ John McCarthy
Road Map to Artificial Intelligence
1. Programming Language
- Python
- R
- Lisp
- Prolog
- Java
2. Mathematical Knowledge
- Linear Algebra
- Probability and Distribution
- Statistics
- Vector Calculus
- Matrix Decomposition
3. Machine Learning Algorithm
An algorithm is a step by step method of solving a problem
| Classification | Algorithm | |
|---|---|---|
| Supervised Learning | Decision Trees Naive Bayes Classification Ordinary Least Squares Regression Logistic Regression Support Vector Machines Ensemble Methods | |
| Unsupervised Learning | Clustering Algorithms Principal Component Analysis Singular Value Decomposition Dijkstra’s Algorithm |
4. Machine Learning Tool
- Google’s TensorFlow
- Torch
- Microsoft Cognitive Toolkit
- IBM Watson
- Amazon Web Services
- Accord.NET Framework
- Caffe
- Eclipse Deeplearning4j
- Apache Mahout
- Theano
- H20
- PredictionlO
- ai-one
- Protégé
- DiffBlue
- Nervana Neon
- OpenNN
- Veles
- Scikit-learn
5. AI Case Study
- Real Estate Price Prediction
- Image Recognition System
- Weather Prediction
- Self-Driving Car
- Automatic Music Composition
- Diagnosing of Medical Diseases
- Identifying the Fake News
- Books Recommendation
Machine Learning
| Self Learning Algorithm | Machine Learning |
|---|---|
| Self Learning Algorithm is the algorithm which will learn by itself and takes own decision/predictions | Machine learning is a subset of Artificial Intelligence. Which works based on self learning algorithm using past experience or dayaset without being explicitly programmed. |
Machine learning is the science of getting computers to act without being explicitly programmed ~ Stanford
Machine learning algorithms can figure out how to perform important tasks by generalizing from examples. ~ University of Washington
Machine learning is based on algorithms that can learn from data without relying on rules-based programming. ~ McKinsey & Co.
| Supervised Learning | Unsupervised Learning | Reinforcement Learning |
|---|---|---|
| It infer a function from labeled data and use this function on new examples/never seen data/test data. | It is the training of an artificial intelligence algorithm using information that is neither classified nor labeled and allowing the algorithm to act on that information without guidance | It enables an agent to learn in an interactive environment by trial and error using feedback from its own actions and experiences |
Supervised Learning
Supervised Learning Output
| Classification | Regression | |
|---|---|---|
| Definition | Classification algorithms are used when the desired output is a Discrete label It is called as Binary Classification (Only two possible outcomes) | Regression algorithms are used when the desired output is a is a Real or Continuous value (Quantity) |
| Example | 1. when filtering emails “spam” or “not spam” 2. when looking at transaction data, “fraudulent”, or “authorized” | 1. Predicting house price based on area 2. Predict the number of copies a music album will be sold next month |
| Model | - Logistic Regression - Decision Tree - Gradient-Boosted Tree - Multilayer Perceptron - One-vs-rest - Naive Bayes - Kernel Approximation - K-Nearest Neighbors - Support Vector Machine - Random Forest | - Ordinary Least Squares Regression (OLSR) - Linear Regression - Stepwise Regression - Multivariate Adaptive Regression Splines (MARS) - Locally Estimated Scatterplot Smoothing (LOESS) - Random Forest |
Unsupervised Learning
Unsupervised Learning Output
| Clustering | Dimensionality Reduction | |
|---|---|---|
| Example | - Recommender Systems - Targeted Marketing - Customer Segmentation - Recommend products to customers based on Past purchase | - Big Data Visualization - Meaningful Compression - Structure Discovery - Features Elicitation |
List of UnSupervised Algorithms
- K- Means , K- Medoids Fuzzy C-Means
- Hierarchical
- Gaussian Mixture
- Neural Networks
- Hidden Markov Model
Reinforcement Learning
Types of Task
| Episodic Task | Continuous Task | |
|---|---|---|
| Definition | It is having starting point and an ending point (a terminal state). This creates an episode - a list of States, Actions, Rewards, and New States. | This Task will continue forever (No terminal state). The agent has to learn how to choose the best actions and simultaneously interacts with the environment. |
| Example | Super Mario Begin : Episode begin at the launch Ending : when you’re killed or you’re reach the end of the level | Automated stock trading |
Types of Learning
| Monte Carlo Approach | Temporal Difference Learning |
|---|---|
| Collecting the rewards at the end of the episode and then calculating the maximum expected future reward | It will update its value estimation for the non-terminal states occurring at that experience. |
| We start a new game with the added knowledge. The agent makes better decisions with each iteration | It is also called TD(O) or one step TD |
List of Reinforcement Learning Algorithms
- Monte Carlo
- Q-Learning
- Q-Learning with Normalized Advantage Functions NAF
- State-Action-Reward-State-Action (SARSA)
- Deep Q Network (DQN)
- Deep Deterministic Policy Gradient (DDPG)