Scribd
Upload
174 views19 pages

ML Unit 1 Notes

Uploaded by

Riyaz Abdul
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
174 views19 pages

ML Unit 1 Notes

Uploaded by

Riyaz Abdul
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 19
Unik Tnbredulin be Mochtue Leaatig op You Thule about eoupulers, Tay computaliion wy watdiel alodshen: Cora pubes cuudertaut precern — everyTitug tntarwa of Goud 1S. olka hou tats are gist af Hearing and -secgrition of thts. Thabo oseally teen trom history sn dc poritvts oF Supurvi Aer at a be the vufurod With teary ore of emmpett (Comptortes) vo fy heaven Nery aeita ete do te ana of geet of wo at) \SMo Nt uell oud In fla abet tes (larry & seco fearfeg ‘ wot ane Traditional Programming Data Data ‘Output Output Machine Learning Figure 1.3: Machine learning versus traditional programming In traditional programming, the computer follows a set of predefined rules to process the input data and produce the outeome, In machine learning, the computer tries to mimic human thinking. It interacts with the input data, expected outcome, and environment, and it derives patterns that are represented by one or more mathematical models. The models are then used to interact with future input data and to generate outcomes. Unlike in automation, the computer in a machine learning setting doesn't receive explicit and instructive coding. ately ofkwe Ah Rime tvetver Trocetng ta befd Mic Hey owapple , ond Hy 1h alfo oH apple , a4 ey 74 Cupeeke and 80 o- Batally yor bau call fe ou oe ayperviton. tidy [ola of learning * happening by Chew! obj wt [EA ategery oY Jabel Cox; apple, xp cate ae pele, ose ti edie 4 bicaly ld oh Aapervized Herat, vA Super sty oy yee Kearny heppes Me earning 13 dome Some tyaivs Couns. Supper dr B oyyied = Wavnsr- pt ou axel fw : ° In A body Wor ge wats doe eee ungein Seep Unsupervised Learning Ze) (28s oN 7 “8 fn rigs 4 el Ye wall woe Ni Yr | bts! 1 ae Sar er Gwscince Sivmillaw vals tthe aber — orclody 14 Supervisirg. you ofod aly cates [tdegorish Jobely) Lek’ dpa ahah feud oF travis Jappess In absente of- Bupeviaer Fo aww ter) feo ab the oboe picture: In qrare ave Bo So wor obpds. 7 don't \cuora whet 15 te clan Sabo for given fap Ruages ti, vi apertitny cou't Fel the ohh abel. Bub 1 Gy Afvide Tau info Shullbr objet togetlen thr objuts one gre lar? L 3 he 4g pron 4 ww Reinforcement Learning environment Ninety eos ray Called agtach é: eagle ob tee Jeary 5 ih Piggies ate aay For bkawnp * 1 a Wow you! are af ee r udev position: From Ta gation oe vei have to teke ent. 04— auy ation dh bated on Aies vault f ple actions. 4 ae ee fa yo Ore fn a ord Sitede- Why machine learning Course introduction 2 Hon: abuts iL aaipitng haf Trvoge Loge RetegshO: Iebut Hyg td OTE Goda Se Goel ae Cs Gpaleols 4 Dine es dorvrat lenod all The Automated Longuoge Tronations s we ei But with ME Ea ec ean travshate om be amolley Nanguege i PS eh soos puhy S/he ee wf Ronn Wier sil ee} ee ed deoypwontte youd piss et Frathic preotens age on inital Calrun4 oa ‘hy ton wontor tala tyro i cross-validation —_—s ee Asa gentle reminder, you will see cross-validat don't panie if you ever find this section difficult to understan very soon. ple ti in this book. So jon in action multiple times later in this ; ae das you will become an expert of it When the training size is very large, it's often sufficient to split it into training. ¥ i= 0 testing (three subsets) and conduct a performance check onthe latter two. Cross-validation is less preferable in this ease since i's computationally costly to train a model for each single round. But if you can afford it, there's no reason not to use cross-validation. When the size isn't so large, cross-validation is definitely a good choice. There are mainly two cross-validation schemes in use: exhaustive and non-exhaustive. In the exhaustive scheme, we leave out a fixed number of observations in each round as testing (or validation) samples and use the remaining observations as training samples. This process is repeated until all possible different subsets of samples are used for testing onee. For instance, we can apply Leave-One-Out-Cross-Validation (LOOCY), which lets each sample be in the testing set once. For a dataset of the size n, LOOCV requires n rounds of cross-validation. This ‘can be slow when n gets large. This following diagram presents the workflow of LOOCV: N samples . Testing . set A non-exhaustive scheme, on partitions, The most widely randomly split the original d becomes the testing set, and the We repeat this process k times, average the k sets of test results, 10. The following table illustrates the setup for Round Fold 1 1 Testing 2 Training 3 Training 4 ‘Training 5 Training Parametric Machine Learning Algorithms Assumptions can greatly simplify the learning process, but can also limit what can be leamed. Algorithms that simplify the function to a known form are called parametric machine learning algorithms. Parametric machine learning algorithms make assumptions about the mapping function and have a fixed number of parameters. No matter how much data is used to lear the model, this will not change how many parameters the algorithm has. With a parametric algorithm, we are selecting the Fold 2 Training Training Training with each fold being the desi for the purpose of evaluation. five-fold: Fold Fold Training Training Training Training Testing ‘Training Training Testing Training ‘Training ignated testing ‘Common values for & are Fold 5 ‘Training Training Training Training Testing form of the function and then learning its coefficients using the The algorithms involve two steps: 1. Selecta form for the function, 2. Learn the coefficients for t ir set once. Finally, we BO+B/ *X1+B2*X2=0 i initial function, the This assumption greatly simplifies the learning process: afer eee ae Seen remaining problem is simply to estimate the coefficients BO, BI, and of input variables X1 and X2. 1g algorithms include: Some more examples of parametric machine learniny + Logistic Regression Linear Discriminant Analysis + Pereeptron + Naive Bayes Simple Neural Networks arametric Machine Learning Algorithms Non: ratamerh eMathine Learning Algneiims Non-parametric algorithms do not make assumptions regarding the form of the mapping function between input data and output, Consequently, they are five to learn any functional form from the training data. A simple example is the K-nearest neighbors (KNN) algorithm, KNN does not make any: assumptions about the functional form, but instead uses the pattern that points have similar output when they are close, Some more examples of popular nonparametric machine learning algorithms are: + k-Nearest Neighbors + Decision Trees like CART and C45 + Support Vector Machines g® Sree edits pea sive Reguemor ty eos commelly abt pe uli te oullsfes ela! au Lables + ie jo ped techies oe aceds Var iors Sderbry - YE Cont BY weresues 3) Rus u) thedcowe s) Exgineer fg SS Bdaebtor. 2) Spots & Eek erletwmnk - pel Berea de cyrtied pot y al welal to preded® thu ap] cbt dotane wh baths sh | orto oF a Loe os Yo preked: ee een © Bag wey Me, Tey Ce ate Nata get! Decwlabion which © Cu De Bice aes weed wrt hye Bowe Wels ot / gel gre output Rig ante ate Sela eyes poise fo Sy 7 calush of ey es Wolk caene @ Sales prektion ® Fee oe pretation. ev. oun eX beau, steve diy fae Thy Leo probing tt. weeh. Veter qleagieg por Ff Susper-vited Jearnteg. rrenein “a here bpewtely yee ane gol iC Nee Sobel. Regre Btn Wako dur elidion tye gk US ie he ial tet wad CL Sraply fae 5 there 3 a warkek fg Compt aud they Drove Colbabed Sonne ef. the dato — catoiws age bathethosct amount spit om workehfag and cokes o PORTA tee Copy gel - to Nave Rome they dee Gwe aq fox oe dave Some better Porsgs from we Gu Come AD daue wore ) No we Want woe Concrefe Con We get which 20eA OE tt. debe) 10a oukpuk Neco thas 1 wha We Wout A ans we Troe UPS Pal avesmived pS Were er shuple Newear SO SL pret ded Goat 10 ra ae (0) Si NEAT aM LAO REL BY 2a aaa S aria (case oars uve SE Tas yj CoG val have plotfed all pointA . Koro Thy iA Let's agumt we we unug to pret Y ae Fleugs whet aly: Sealed om ml ws will be Calling 4 cepted variable oF targa vwyiable, a4 At, Vartoble, (martertg predict calls 1A Called of de budget) usivg which &e Jud variable o- preddor. Ix Sinple LR we f vorrabl.c waleat sd tadopendtrt yarlable , led bo pred dopisatt Peat Nek ea ty hy ee vrelaffoud.ge ble oS al . ‘ Pec aor ory aud Salen ott dito Wor Hue 5 Str eels db ptiws We Gu uuderd rad AG ale Prevensly So Let'4 ag em a rant

You might also like