Transcription
Dr. Janell StraachFaculty, OMCSRice Universityjanell.straach@rice.eduGrace Series Talk to UT DallasSeptember 2019
We start learning when we are born SoundsTouchSmellReachEatMotor skillsVerbal sounds
Machine Learning Definition Alan Turing posed it this way "Can machines think?"We now prefer to pose it as "Can machines do what we (asthinking entities) can do?“The name machine learning was coined in 1959 by Arthur SamuelTom Mitchell’s definition is the one we will use"A computer program is said to learn from experience E withrespect to some class of tasks T and performance measure P if itsperformance at tasks in T, as measured by P, improves withexperience E.”-- Tom Mitchell
We start learning when we are born SoundsTouchSmellReachEatMotor skillsVerbal sounds "A computer program is said tolearn from experience E withrespect to some class oftasks T and performancemeasure P if its performance attasks in T, as measured by P,improves with experience E.”-- Tom Mitchell
My journeyWorkSchool0 to 66 to 1515 to 17Part TimeKid stuffElem & Jr HighHigh School17 to 23Part Time23 to 2828-3636-4950 Full TimePart TimeMaster’s & PhDFull TimeMBABest of bothBachelor’s &Master’s
Categories of Learning Supervised LearningInput data includes label (desiredsolution) which is used to “train” thesystemUnsupervised LearningInput data is not labeledSemisupervised LearningSome input data is labeled and some isnotReinforcement LearningSystem interacts and receives rewards orpenalties based on decisions todetermine a strategy (called a policy)
Deep Learning
Deep Learning comparison
My journeyWork0 to 6SchoolKid stuff6 to 15Elem & Jr High15 to 1717 to 23Part TimePart Time23 to 2828-36Full TimePart Time36-4950 High SchoolBachelor’s &Master’sMaster’s & PhDFull TimeMBABest of forcementDeep LearningDeep Learning?
Perceptrons--Background
Perceptron (Artificial Neuron)
Common Activation Functions
Classification
Classification – My strengthsStrengthsWeaknesses
Things I learned: Most people willadd an “e” to theend of my name Talking is mystrength?
Conceptually: Neural NetworksForward Activity - Backward Error
JS Neural Network Connections FriendsFamilyWorkMental HealthPhysical HealthPersonal GrowthCommunitySecurityetc
Clustering -- k-means algorithmRandomly pick k centroids from the sample points as initial cluster centersRepeatAssign each sample to the nearest centroidMove the centroids to the center of the samples that were assigned to itUntil cluster assignments do not change or tolerance or maximum iterationsreached
JS ClustersPassionImportantEnjoy
Some JS ationLearningFairnessFriendsDiversityLaughter
Sometimes we have to “un”learn
Bias in Machine LearningX Boys are not allowedon the top rowX Only women areallowed to beteachersX Girls must weardresses
Never stop learning! Learn your strengthsand leverage them! Adjust prioritiesconstantly Sort through the noisein life! Be willing to unlearnthings! Embrace pivots!
janell.straach@rice.edu
Rice University janell.straach@rice.edu Grace Series Talk to UT Dallas September 2019. . 28-36 Part Time Master's & PhD 36-49 Full Time MBA 50 Best of both. Categories of Learning Supervised Learning Input data includes label (desired solution) which is used to "train" the