Scribd
Upload
144 views

Machine Learning Algorithm Cheat Sheet - Laura Diane Hamilton

This document provides a cheat sheet for machine learning algorithms, summarizing their pros and cons and what tasks each is best suited for. Linear regression is good for numerical data with many features but cannot handle complex relationships. Decision trees are robust, fast, and good for classification tasks. Neural networks can model very complex relationships but require significant computing power. Support vector machines can model nonlinear relationships and are robust but require selecting a good kernel function. K-nearest neighbors is simple, powerful, and good for low-dimensional datasets.

Uploaded by

werwerwer
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
144 views

Machine Learning Algorithm Cheat Sheet - Laura Diane Hamilton

This document provides a cheat sheet for machine learning algorithms, summarizing their pros and cons and what tasks each is best suited for. Linear regression is good for numerical data with many features but cannot handle complex relationships. Decision trees are robust, fast, and good for classification tasks. Neural networks can model very complex relationships but require significant computing power. Support vector machines can model nonlinear relationships and are robust but require selecting a good kernel function. K-nearest neighbors is simple, powerful, and good for low-dimensional datasets.

Uploaded by

werwerwer
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

10/18/2016

MachineLearningAlgorithmCheatSheetLauraDianeHamilton

LauraDianeHamilton
TechnicalProductManageratGroupon
Resum
@lauradhamilton
linkedin
github
googleplus
email
rss

MachineLearningAlgorithmCheatSheet
September09,2014
Hereisacheatsheetthatshowswhichalgorithmsperformbestatwhichtasks.
Algorithm
Linear
regression

Decision
trees

Neural
networks

Support
Vector
Machines

KNearest
Neighbors

Pros
Veryfast(runsinconstant
time)
Easytounderstandthe
model
Lesspronetooverfitting
Fast
Robusttonoiseand
missingvalues
Accurate
Extremelypowerful
Canmodelevenvery
complexrelationships
Noneedtounderstandthe
underlyingdata
Almostworksbymagic
Canmodelcomplex,
nonlinearrelationships
Robusttonoise(because
theymaximizemargins)

Simple
Powerful
Notraininginvolved
(lazy)
Naturallyhandles

Cons
Unabletomodelcomplex
relationships
Unabletocapturenonlinear
relationshipswithoutfirsttransforming
theinputs

Goodat
Thefirstlookatadataset
Numericaldatawithlots
offeatures

Complextreesarehardtointerpret
Starclassification
Duplicationwithinthesamesubtree Medicaldiagnosis
ispossible
Creditriskanalysis
Pronetooverfitting
Longtrainingtime
Requiressignificantcomputing
powerforlargedatasets
Modelisessentiallyunreadable

Images
Video
Humanintelligence
typetaskslikedrivingor
flying
Robotics

Needtoselectagoodkernelfunction
Modelparametersaredifficultto
interpret
Sometimesnumericalstability
problems
Requiressignificantmemoryand
processingpower

Classifyingproteins
Textclassification
Imageclassification
Handwritingrecognition

Expensiveandslowtopredictnew
instances
Mustdefineameaningfuldistance
function
Performspoorlyonhigh
dimensionalitydatasets

http://www.lauradhamilton.com/machinelearningalgorithmcheatsheet

Lowdimensionaldatasets
Computersecurity:
intrusiondetection
Faultdetectionin
semiconducter
manufacturing
Videocontentretrieval
Geneexpression
1/2

10/18/2016

MachineLearningAlgorithmCheatSheetLauraDianeHamilton

multiclassclassificationand
regression

Proteinprotein
interaction

FollowmeonTwitterorsubscribetoRSS
GraphingwithRHowtoGettheIonicFrameworkRunningonUbuntu
Lauradhamilton.comisaparticipantintheAmazonServicesLLCAssociatesProgram,anaffiliateadvertising
programdesignedtoprovideameansforsitestoearnadvertisingfeesbyadvertisingandlinkingto
amazon.com.

Login
Email
Password
Login

http://www.lauradhamilton.com/machinelearningalgorithmcheatsheet

2/2

You might also like