SLIDE 1
AI and Predictive Analytics in Data-Center Environments
Supervised Learning Methods
Josep Ll. Berral @BSC
Intel Academic Education Mindshare Initiative for AI
AI and Predictive Analytics in Data-Center Environments Supervised - - PowerPoint PPT Presentation
AI and Predictive Analytics in Data-Center Environments Supervised - - PowerPoint PPT Presentation
AI and Predictive Analytics in Data-Center Environments Supervised Learning Methods Josep Ll. Berral @BSC Intel Academic Education Mindshare Initiative for AI Introduction If we have data and it is labeled, we can learn their relation and
SLIDE 2
SLIDE 3
Supervised Learning
- Supervised Learning
- Training data is already labeled
- Want to predict new unlabeled data
SLIDE 4
Supervised Learning
Set of Examples <features> + <label> Observe Label Model Observe New Set of Examples <features> Automatic Labeling <labels> Later...
SLIDE 5
Supervised Learning
- Labeling data:
- By hand
- By known methods
- By posterior metrics
- From known data
in
- ut
“cat” “cat” “dog” “cat” “dog” “cat” “cat” “cat” “dog” “dog”
Observe Label
dataset
SLIDE 6
Typical Flow
Model “dog” What is this? Right! [Reinforce]
SLIDE 7
Typical Flow
Model “dog” What is this? No! It is a “cat” [Adapt] Model “dog” What is this? Right! [Reinforce]
SLIDE 8
Typical Flow
Model “dog” What is this? No! It is a “cat” [Adapt] Model “dog” What is this? Right! [Reinforce] Model “cat” What is this? Right! [Reinforce]
SLIDE 9
Typical Flow
Model “dog” What is this? No! It is a “cat” [Adapt] Model “dog” What is this? Right! [Reinforce] Model “cat” What is this? Right! [Reinforce] Iterate until the model is accurate
SLIDE 10
Good Procedures
- Keep some data “unseen”, to avoid overfitting / memorizing
Training Set Model Algorithm
SLIDE 11
Good Procedures
- Keep some data “unseen”, to avoid overfitting / memorizing
Training Set Validation Set Model Algorithm Predict Model Evaluate Is it good? No Tune the Algorithm
SLIDE 12
Good Procedures
- Keep some data “unseen”, to avoid overfitting / memorizing
Training Set Validation Set Model Algorithm Predict Model Evaluate Is it good? No Tune the Algorithm Test Set Model Predict Final Evaluation Yes
SLIDE 13
Supervised Learning
Algorithms & Methods!
SLIDE 14
Algorithms & Methods
- Classification
- The outputs are “classes”
E.g.:
- Regression
- The outputs are “quantities”
E.g.:
“cat” “dog” Max Speed Car Properties
SLIDE 15
Some Methods
- Regression algorithms
- Linear & Polynomial Regression, Gaussian Processes, ...
“Attempt to find a function/set-of-functions that match with the example points”
- The learning process minimizes the regression error
SLIDE 16
Some Methods
- Trees and Forests
- Decision Trees, Regression Trees, Random Forests
“Attempt to find a set of recursive partition that minimize the classification or regression error”
“A” “B” “C”
“A” “B” “C” “A” “C” “B” “B” “A” “C”
SLIDE 17
Some Methods
- k – Nearest Neighbors
“Compare new samples with some memorized ones, and classify/predict as the ‘k’ nearest ones”
?
SLIDE 18
Some Methods
- Bayesian Methods
- Naïve Bayes, Bayesian Networks, ...
“Compute probabilities of classes, events and relations, then apply Bayes theorem” P(A|B) = P(A) · P(B|A) / P(B) P(Class|Example) = P(Class) · P(Example|Class) / P(Example) = P(Class & Example) / P(Example)
SLIDE 19
Some Methods
- Support Vector Machines
“Find the function that best divides classes, with a minimal tolerance for errors”
SLIDE 20
Summary
- Supervised Learning:
- Models learn from labeled data, and “human-driven” tuning
- Methods for Regression & Classification
- Lots of Algorithms to be applied
- Each with its characteristics
- Strong and weak points
- Different consumption of resources