Introduction to Machine Learning

Grasp basics in machine learning

Transcript

2. About Me • Research and build Robots • Enjoy studying AI research papers • Got engaged to Machine Learning (I got a big ring) • Super hero ~ ( I gotchu you when you need me) • Work as a Site Reliability Engineer @AEDC @jennifazor Jenniferkwentoh.com
3. It’s a Field in IT as well •“Machine Learning is the field of study that gives computers the ability to learn without being explicitly programmed” ~ Arthur Samuel 1959 • Nothing to be scared of ~ it’s not just for topnotch developers • 2017, we have more and more female Machine Learning engineers • We would be doing a lot of snapping and Hi-fives (don’t worry we would code too)
4. Mom: c’mon baby you can do it, say daddy Baby: daa aa
5. More and more experience please! • We learn from both our success and failures and we use the experience to create strategies to make decisions for better performance. • With Machine Learning we are training our program to do the same. • Improving the way it performs a given task based on past experience
6. Whose hungry? • Toaster analogy I want bread Have to figure how she likes it
7. “Experience”== “Data”
8. Some family members Statistics and Data Mining Grand papa is Artificial Intelligence
9. First thing • Start with a question •WHAT EXACTLY AM I TRYING TO SOLVE?
10. Second Thing • Data •Data Data Data • Data Data and more Data
11. Third thing • Understand the classification Problem + Dataset = Right Classification/type to use
12. Our babies • Supervised Learning ~ has a given input (feature) and output (label) • examples: hand writing recognition, face recognition, movie recommendation • Unsupervised Learning ~ unlabeled data • examples: detecting money fraud in banks • Reinforcement Learning ~ unlabeled data
13. Relatives we can’t live without (uncle Joe’s kids) • Training models and algorithms • Classifier ~ takes data as input (feature) and assigns label to it as an output • Example: decision tree, SVM
14. Some Tools to Use • There are different brands of make-up brushes but they all get the job done Python, R, Java, C++, Javascript, Scala, Julia, C, matlab
15. The winner is? Open Source Libraries • Tensorflow • Scikit-learn • Pattern • Theano • The list goes on and on!
16. Weldone for listening so far, give someone a high-five Take a moment break you can go powder your nose
17. Let’s code a little •We would be using scikit-learn You can download and install easily with Anaconda
18. Steps Understand the problem Collect Data Train Classifier We can start making predictions (we would be using a decision tree classifier)
19. Training • Problem set : we want to identify fake and real BS-Mall makeup brushes • Training data • 10cm light brown fake (0) • 12cm light brown fake (0) • 13cm dark brown original (1) • 15cm dark brown original (1) features Labels
20. Decision Tree classifier Length >= 12cm? YES NO is color light brown ? FAKE ORIGINAL YES NO
21. On the features we used 0’s for light brown and 1’ s for dark brown On labels we used 0’s for light brown and 1’ s for dark brown
22. Resources • https://geekyrobotacademy.com • https://machinelearningmastery.com • Udacity Machine Learning https://udacity.com • Youtube • Google Developers (machine learning playlist) • Siraj Raval