Passionate about Machine learning? Same for us here @Dreamix. Machine learning is so vast today that you probably use it dozens of times a day without knowing it. In the past years, machine learning has given us effective web search,self-driving cars, practical speech recognition. Now is the time to learn more about it. Enjoy!

1. Machine Learning
Boris Velichkov
www.dreamix.eu

2. About Boris
● Java EE
● Oracle ADF
● Consultancy
● Artificial Intelligence
● Data Mining
● Object-oriented programming and
modelling of software
● Algorithms design/analysis and data
structures
● Agile methodologies
Java EE/Oracle Developer @Dreamix

3. Machine
Learning?
www.dreamix.eu

4. How do we learn?
www.dreamix.eu

5. How do we learn?

6. Machine Learning
Supervised Learning Unsupervised Learning
Classification Regression Clustering
www.dreamix.eu

7. Classification
Attribute 1 Attribute 2 … … Attribute
N
Class
„Yes“
„No“
„Yes“
„Yes“
X1
.
.
.
Xm
Dataset
Xt = (a1, a2, … , aN)
Class(Xt) = ?
www.dreamix.eu

8. Classification
Humidity Wind Temperature Class (Play Tennis)
high mid 35° No
mid strong 5° No
mid weak 20° Yes
high weak 25° Yes
low strong 18° No
X1
X2
X3
X4
X5
Dataset
Xt = (low, weak, 21°)
Class(Xt) = ?
www.dreamix.eu

9. Regression
Attribute 1 Attribute 2 … … Attribute
N
Class
1.5
0.7
1.8
1.2
X1
.
.
.
Xm
Dataset
Xt = (a1, a2, … , aN)
Class(Xt) = ?
www.dreamix.eu

10. Regression
Humidity Wind Outlook Class (T°)
high mid rainy 35
mid strong overcast 5
mid weak sunny 20
high weak overcast 25
low strong overcast 18
X1
X2
X3
X4
X5
Dataset
Xt = (low, weak, sunny)
Class(Xt) = ?
www.dreamix.eu

11. Classification Algorithms
www.dreamix.eu

12. Classification Algorithms
• Linear classifiers
– Fisher's linear discriminant
– Logistic regression
– Naive Bayes classifier
– Perceptron
www.dreamix.eu

13. Classification Algorithms
• Linear classifiers
– Fisher's linear discriminant
– Logistic regression
– Naive Bayes classifier
– Perceptron
www.dreamix.eu

14. Classification Algorithms
• Linear classifiers
– Fisher's linear discriminant
– Logistic regression
– Naive Bayes classifier
– Perceptron
www.dreamix.eu

15. Classification Algorithms
• Support vector machines
– Least squares support vector machines
www.dreamix.eu

16. Classification Algorithms
• Quadratic classifiers
www.dreamix.eu

17. Classification Algorithms
• Kernel estimation
– k-nearest neighbor
K = 3K = 4
www.dreamix.eu

18. Classification Algorithms
• Boosting (meta-algorithm)
– AdaBoost
www.dreamix.eu

19. Classification Algorithms
• Decision trees
– Random forests
www.dreamix.eu

20. Classification Algorithms
• Neural networks
www.dreamix.eu

21. Classification Algorithms
• Learning vector quantization
I
N
P
U
T
O
U
T
P
U
T
„winner-take-
www.dreamix.eu

22. Cross-validation
www.dreamix.eu

23. Clustering
www.dreamix.eu

24. Clustering Example
www.dreamix.eu

25. Hierarchical Clustering
A
C
B
D E
F G
A B C D E F G
Dendrogram
S
i
m
i
l
a
r
i
t
y
www.dreamix.eu

26. Aglommerative vs. Divisive
Clustering
✓ Aglommerative ( bottom-up )
✓ Divisive ( partitional, top-down )
www.dreamix.eu

27. Agglomerative Clustering
A B C D E F G H I
www.dreamix.eu

28. Cluster Similarity
–Single Link: Similarity of the two most
similar members.
–Complete Link: Similarity of the two
least similar members.
–Group Average: Average similarity
between members.
www.dreamix.eu

29. Direct Clustering Method
www.dreamix.eu

30. K-Means Clustering
www.dreamix.eu

31. K-Means Clustering
www.dreamix.eu

32. K-Means Clustering
www.dreamix.eu

33. K-Means Clustering
www.dreamix.eu

34. K-Means Clustering
www.dreamix.eu

35. K-Medians Clustering
www.dreamix.eu

36. Expectation-Maximization
Clustering
www.dreamix.eu

37. Classification
Clustering
www.dreamix.eu

38. Classification
Class 1
Class 2
www.dreamix.eu

39. Classification
Class 1
Class 2

40. Classification
Class 1
Class 2

41. Classification
Class 1
Class 2

42. Clustering

43. Clustering

44. Clustering

45. Clustering

46. Clustering

47. AI & Money
IBM
& Watson
Google
& DeepMind
Ontotext
& BBC
. . .
www.dreamix.eu

48. DEMOwww.dreamix.eu

49. mix.eu

51. Thank
You ! www.dreamix.eu