Download presentation
Presentation is loading. Please wait.
Published byJudith Booth Modified over 9 years ago
1
Machine Learning Lecture for Methodological Foundations of Biomedical Informatics Fall 2015 (BMSC-GA 4449) Sisi Ma NYU Langone Medical Center CHIBI
2
What type of problems can machine learning solve? Re Real Estate Artificial Intelligence Retail Sales Conservation Climate Current Active Projects on Kaggle as of Oct, 26 th,2015Kaggle
3
What type of problems can machine learning solve? Predominantly: Classification
4
How to classify? Main Ways to Classify: -Unsupervised -Supervised
5
Unsupervised Learning Group similar items together Comics credit: http://nlp.cs.berkeley.edu/comics.shtml
6
Unsupervised Learning Since the definition of similarity is arbitrary, one can get different labeling solutions.
7
Unsupervised Learning The solution depend on both: (1) what variables were used to construct the similarity metric (2) how the similarity metric were constructed.
8
Unsupervised Learning The solution depend on both: (1) what variables were used to construct the similarity metric (2) how the similarity metric were constructed.
9
Unsupervised Learning The solution depend on both: (1) what variables were used to construct the similarity metric (2) how the similarity metric were constructed. Lowe, 2012
10
Unsupervised Learning The solution depend on both: (1) what variables were used to construct the similarity metric (2) how the similarity metric were constructed. Image Credit: https://en.wikipedia.org/wiki/Metric_(mathematics)
11
Unsupervised Learning How do we know the solution is good? It corresponds to something we care about.
12
Unsupervised Learning
13
Supervised Learning
14
Overfitting Duda, 2ed
15
Supervised Learning Overfitting Image Credit: https://commons.wikimedia.org/wiki/File:Overfitting.svg
16
Supervised Learning How do I know if I am overfitting? Validation Data
17
Supervised Learning How do I know if I am overfitting? Duda, 2ed
18
Supervised Learning Support Vector Machine 18 Key Characteristics of SVM Maximum gap to prevent overfitting QP problems can be solved with standard methods. Soft margins to tolerate noise Kernel trick for linearly non-separable data Statnikov et al., 2011 Most modern algorithms have built in mechanism to minimize overfitting.
19
Predictive Modeling: A Simplified General Framework 19 Validation Data
20
Predictive Modeling: Cross Validation for error estimation and model selection 20 Ma et al., 2015 (in preparation)
21
Machine Learning vs Statistics Robert Tibshiriani
22
Machine Learning vs Statistics Robert Tibshiriani
23
Machine Learning vs Statistics Machine LearningStatistics One major difference between machine learning and statistics : How is the model evaluated?
24
Machine Learning vs Statistics What is a good model? According to most statistician, in practice especially Most commonly evaluated by R-squared Breiman, 2001
25
Machine Learning vs Statistics Validation Data What is a good model? According to machine learning researcher.
26
The Future
28
What’s the job?
33
Homework Research bias-variance decomposition and answer the following question from ” An Introduction to Statistical Learning”.
34
Resources
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.