Describes about Machine Learning

1. WHAT IS MACHINE
LEARNING
Bhaskara Reddy Sannapureddy, Senior Project Manager @Infosys, +91-7702577769

2. WHAT IS MACHINE
LEARNING?
Automating automation
Getting computers to program themselves
Writing software is the bottleneck
Let the data do the work instead!

3. TRADITIONAL PROGRAMMING VS
Computer
Data
Program
Output
Computer
Data
Output
Program
MACHINE LEARNING
Traditional Programming
Machine Learning

4. MAGIC?
No, more like gardening
Seeds = Algorithms
Nutrients = Data
Gardener = You
Plants = Programs

5. SAMPLE APPLICATIONS
Web search
Computational biology
Finance
E-commerce
Space exploration
Robotics
Information extraction
Social networks
Debugging
[Your favorite area]

6. ML IN A NUTSHELL
 Tens of thousands of machine learning algorithms
 Hundreds new every year
 Every machine learning algorithm has three components:
• Representation
• Evaluation
• Optimization

7. REPRESENTATION
 Decision trees
 Sets of rules / Logic programs
 Instances
 Graphical models (Bayes/Markov nets)
 Neural networks
 Support vector machines
 Model ensembles
 Etc.

8. EVALUATION
Accuracy
Precision and recall
Squared error
Likelihood
Posterior probability
Cost / Utility
Margin
Entropy
K-L divergence
Etc.

9. OPTIMIZATION
 Combinatorial optimization
• E.g.: Greedy search
 Convex optimization
• E.g.: Gradient descent
 Constrained optimization
• E.g.: Linear programming

10. TYPES OF LEARNING
 Supervised (inductive) learning
• Training data includes desired outputs
 Unsupervised learning
• Training data does not include desired outputs
 Semi-supervised learning
• Training data includes a few desired outputs
 Reinforcement learning
• Rewards from sequence of actions

11. INDUCTIVE LEARNING
 Given examples of a function (X, F(X))
 Predict function F(X) for new examples X
• Discrete F(X): Classification
• Continuous F(X): Regression
• F(X) = Probability(X): Probability estimation

12. SUPERVISED AND
UNSUPERVISED LEARNING
 Supervised learning
• Decision tree induction
• Rule induction
• Instance-based learning
• Bayesian learning
• Neural networks
• Support vector machines
• Model ensembles
• Learning theory
 Unsupervised learning
• Clustering
• Dimensionality reduction

13. MACHINE LEARNING
PROBLEMS

14. ML IN PRACTICE
 Understanding domain, prior knowledge, and goals
 Data integration, selection, cleaning,
pre-processing, etc.
 Learning models
 Interpreting results
 Consolidating and deploying discovered knowledge
 Loop

15. CLUSTERING STRATEGIES
 K-means
• Iteratively re-assign points to the nearest cluster center
 Agglomerative clustering
• Start with each point as its own cluster and iteratively merge the closest clusters
 Mean-shift clustering
• Estimate modes of pdf
 Spectral clustering
• Split the nodes in a graph based on assigned links with similarity weights
As we go down this chart, the clustering strategies have
more tendency to transitively group points even if they are
not nearby in feature space

16. THE MACHINE LEARNING
FRAMEWORK
 Apply a prediction function to a feature representation of the
image to get the desired output:
Slide credit: L. Lazebnik

17. THE MACHINE LEARNING
FRAMEWORK
y = f(x)
output prediction
function
Image
feature
 Training: given a training set of labeled examples {(x1,y1), …, (xN,yN)}, estimate the
prediction function f by minimizing the prediction error on the training set
 Testing: apply f to a never before seen test example x and output the predicted value y = f(x)
Slide credit: L. Lazebnik

18. THANK YOU