Consider the machine learning problem in the figure below (only four examples of each class are shown for simplicity, imagine that we have 1,000 examples.

Slides:



Advertisements
Similar presentations
5.4 Correlation and Best-Fitting Lines
Advertisements

On Comparing Classifiers: Pitfalls to Avoid and a Recommended Approach Author: Steven L. Salzberg Presented by: Zheng Liu.
DECISION TREES. Decision trees  One possible representation for hypotheses.
Naïve-Bayes Classifiers Business Intelligence for Managers.
Machine Learning Instance Based Learning & Case Based Reasoning Exercise Solutions.
NEURAL NETWORKS Perceptron
My name is Dustin Boswell and I will be presenting: Ensemble Methods in Machine Learning by Thomas G. Dietterich Oregon State University, Corvallis, Oregon.
K-NEAREST NEIGHBORS AND DECISION TREE Nonparametric Supervised Learning.
Counting Outcomes and Tree Diagrams
Data Mining Classification: Alternative Techniques
LIAL HORNSBY SCHNEIDER
Algebra I 6.0 Students graph a linear equation and compute the x- and y-intercepts. They are also able to sketch the region defined by linear inequality.
Classification Dr Eamonn Keogh Computer Science & Engineering Department University of California - Riverside Riverside,CA Who.
Copyright © Cengage Learning. All rights reserved. 10 Introduction to the Derivative.
Induction of Decision Trees
Classification Continued
Three kinds of learning
Support Vector Machines
ML ALGORITHMS. Algorithm Types Classification (supervised) Given -> A set of classified examples “instances” Produce -> A way of classifying new examples.
SIEVE—Search Images Effectively through Visual Elimination Ying Liu, Dengsheng Zhang and Guojun Lu Gippsland School of Info Tech,
Graphing Linear Inequalities
Machine Learning CS 165B Spring 2012
Learning what questions to ask. 8/29/03Decision Trees2  Job is to build a tree that represents a series of questions that the classifier will ask of.
Outline Classification Linear classifiers Perceptron Multi-class classification Generative approach Naïve Bayes classifier 2.
1 1 Slide Evaluation. 2 2 n Interactive decision tree construction Load segmentchallenge.arff; look at dataset Load segmentchallenge.arff; look at dataset.
1 What you will learn today 1. New vocabulary 2. How to determine if data points are related 3. How to develop a linear regression equation 4. How to graph.
Copyright © Cengage Learning. All rights reserved. 3 Introduction to the Derivative.
Learning Theory Reza Shadmehr Linear and quadratic decision boundaries Kernel estimates of density Missing data.
Ensemble Learning Spring 2009 Ben-Gurion University of the Negev.
Data Reduction via Instance Selection Chapter 1. Background KDD  Nontrivial process of identifying valid, novel, potentially useful, and ultimately understandable.
Working collaboratively towards creating positive possible selves Sarah Fletcher
A Content-Based Approach to Collaborative Filtering Brandon Douthit-Wood CS 470 – Final Presentation.
Gang WangDerek HoiemDavid Forsyth. INTRODUCTION APROACH (implement detail) EXPERIMENTS CONCLUSION.
CS378 Final Project The Netflix Data Set Class Project Ideas and Guidelines.
ISQS 7342 Dr. zhangxi Lin By: Tej Pulapa. DT in Forecasting Targeted Marketing - Know before hand what an online customer loves to see or hear about.
ORTHOGRAPHIC PROJECTION. There is a pile of both. Which one do you choose and why?
Photography Composition Basics. The Rule of Thirds.
Histograms, Frequency Polygons, and Ogives. What is a histogram?  A graphic representation of the frequency distribution of a continuous variable. Rectangles.
Combining multiple learners Usman Roshan. Decision tree From Alpaydin, 2010.
Lecture 5: Statistical Methods for Classification CAP 5415: Computer Vision Fall 2006.
CONFIDENTIAL 1 Good Afternoon! Today we are going to learn about ROUNDING whole numbers. 1.) Find the next number in the pattern. 22, 23, 25, 28, ____.
Morphing Project II. What is morphing?  Changing one object to another through seamless transition.
1 Copyright © Cengage Learning. All rights reserved. 2 Polynomial and Rational Functions.
Graphing Linear Inequations y > y is greater than  all points above the line are a solution y < y is less than  all points below the line are a solution.
Instance-Based Learning Evgueni Smirnov. Overview Instance-Based Learning Comparison of Eager and Instance-Based Learning Instance Distances for Instance-Based.
Copyright © Cengage Learning. All rights reserved.
k-Nearest neighbors and decision tree
Graphing Linear Inequalities
MIRA, SVM, k-NN Lirong Xia. MIRA, SVM, k-NN Lirong Xia.
Improving the Performance of Fingerprint Classification
*Currently, self driving cars do a bit of both.
Classification Nearest Neighbor
Classification with Perceptrons Reading:
Features & Decision regions
Machine Learning Week 1.
Students: Meiling He Advisor: Prof. Brain Armstrong
We understand classification algorithms in terms of the expressiveness or representational power of their decision boundaries. However, just because your.
PROBLEM 1 Training Examples: Class 1 Training Examples: Class 2
Data Mining extracting knowledge from a large amount of data
Experiments in Machine Learning
2.7 Two-variable inequalities (linear) 3.3 Systems of Inequalities
Day 54 – Standard Deviation
Lecture 7: Simple Classifier (KNN)
Midterm Exam Closed book, notes, computer Format:
Advanced Artificial Intelligence Classification
Ensemble learning.
CS412 – Machine Learning Sentiment Analysis - Turkish Tweets
Machine Learning with Clinical Data
MAS 622J Course Project Classification of Affective States - GP Semi-Supervised Learning, SVM and kNN Hyungil Ahn
MIRA, SVM, k-NN Lirong Xia. MIRA, SVM, k-NN Lirong Xia.
Presentation transcript:

Consider the machine learning problem in the figure below (only four examples of each class are shown for simplicity, imagine that we have 1,000 examples to train our machine learning algorithms). Which of the following algorithms could potentially achieve close to 100% classification accuracy on this problem? Hint, plot out a bunch of examples in a 2D plot A) Nearest Neighbor and Decision tree B) Simple Linear Classifier C) Decision tree onlyD) No algorithm can do well on this problem. A learning problem. (The rule used to determine class is this, if the sum of the two heights is less than or equal to ten, it is class A, otherwise it is class B).

Consider the machine learning problem below. Suppose I learn a Nearest Neighbor for this problem. The resulting classifier would have an accuracy rate of about… A) ) 99 to 100%B) 20 to 25% C) 70 to 90%.D) Nearest Neighbor is not defined for this problem Now suppose I learn a simple linear classifier for this problem. The resulting classifier would have an accuracy rate of about… A) ) 99 to 100%B) 45 to 55% C) 70 to 90%.D) 0%

A) Do nothingB) Make the decision tree more accurate. C) Make both algorithms more accurateD) Make nearest neighbor more accurate. A) Abandoning the projectB) Does not effect my recommendation C) Decision treeD) Nearest neighbor A) Does not effect my recommendationB) Decision tree C) Abandoning the projectD) Nearest neighbor I am working for the government on a secret project. The send me a dataset with ninety-two million instances and 17 continuous features. I have no idea what the features are. I quickly test both the nearest neighbor and decision tree classifiers, they both have about 87% accuracy on this problem, so I have no reason to choose one over the other… Given the above, suppose that the government tells me that it now knows that features 2 and 3 are useless for the problem at hand. Using this information I can Given the above, suppose that the government tells me that it the problem is to classify whether a image on a radar screen is an civilian airplane or incoming missile. In the latter case the government has 0.03 seconds to try to shoot the missile down. This information would make me lean towards recommending Given the above, suppose that the government tells me that one of the goals of the project is to gain some understanding of the data, not just to build a accurate classifier. This information would make me lean towards recommending

Look at the classification problem to the left. How well to you think a simple linear classifier would work? Draw the decision boundary for it (just a rough sketch) How well to you think a nearest neighbor would work? Draw the decision boundary for it (just a rough sketch) How well to you think a decision tree would work? Draw the decision tree, and draw the decision boundary (just a rough sketch)

Look at the classification problem to the left. How well to you think a simple linear classifier would work? Draw the decision boundary for it (just a rough sketch) How well to you think a nearest neighbor would work? Draw the decision boundary for it (just a rough sketch) How well to you think a decision tree would work? Draw the decision tree, and draw the decision boundary (just a rough sketch)

Look at the classification problem to the left. How well to you think a simple linear classifier would work? Draw the decision boundary for it (just a rough sketch) How well to you think a nearest neighbor would work? Draw the decision boundary for it (just a rough sketch) How well to you think a decision tree would work? Draw the decision tree, and draw the decision boundary (just a rough sketch)

Look at the classification problem to the left. How well to you think a simple linear classifier would work? Draw the decision boundary for it (just a rough sketch) How well to you think a nearest neighbor would work? Draw the decision boundary for it (just a rough sketch) How well to you think a decision tree would work? Draw the decision tree, and draw the decision boundary (just a rough sketch) Note: This is a tricky one. There is a good decision tree for this problem, but it is not clear that the algorithm we discussed would find it. For now, just draw the best decision tree for this problem

Look at the classification problem to the left. In this example we know both the x ans y value for all our examples: However, suppose that we only had the X value How well to you think a simple linear classifier would work? Draw the decision boundary for it (just a rough sketch) How well to you think a decision tree would work? Draw the decision tree, and draw the decision boundary (just a rough sketch) Now suppose that we only had the Y value How well to you think a simple linear classifier would work? Draw the decision boundary for it (just a rough sketch) How well to you think a decision tree would work? Draw the decision tree, and draw the decision boundary (just a rough sketch) X Y

Look at the classification problem to the left. How well to you think a simple linear classifier would work? Draw the decision boundary for it (just a rough sketch) How well to you think a nearest neighbor would work? Draw the decision boundary for it (just a rough sketch) How well to you think a decision tree would work? Draw the decision tree, and draw the decision boundary (just a rough sketch) Note: This is a tricky one. There is a good decision tree for this problem, but it is not clear that the algorithm we discussed would find it. For now, just draw the best decision tree for this problem

(Western Pipistrelle (Parastrellus hesperus) Photo by Michael Durham

Characteristic frequency Call duration Western pipistrelle calls We can easily measure two features of bat calls. Their characteristic frequency and their call duration