Fajl koji je korišćen može se naći na

Save this PDF as:
 WORD  PNG  TXT  JPG

Size: px
Start display at page:

Download "Fajl koji je korišćen može se naći na"

Transcription

1 Machine learning Tumačenje matrice konfuzije i podataka Fajl koji je korišćen može se naći na Fajl se odnosi na pečurke (Edible mushrooms). Svaka instanca je definisana sa 23 atributa (raznih: oblik šešira, boja, boje spora...). Poslednji atribut je klasa i može uzimati vrednosti Y ili N (Y - jestiva, N - poisonous). Podaci se propuštaju kroz klasifikator da bi se omogućilo da se na osnovu njih, neka nova pečurka, sa određenim atributima, svrsta u jednu od ovih klasa. Rezultat klasifikacije nam govori koliko je model dobar. Šta se dešava: Klasifikator napravi model koji treba da na osnovu prethodno poznatih podataka, klasifikuje novu instancu pečurke u jednu od dveju klasa. Zatim se taj model proverava nekim podacima za koje se znaju klase, da bi se procenila uspešnost modela. Pošto se model pravi nad velikom količinom podataka, naravno da neće biti savršen (što je i dobro - ali to je neka druga priča) već će i praviti greške. Rezultat koji dobijemo pokazuje uspešnost modela. U našem slučaju, model je napravljen na celokupnom setu podataka. Zatim je isti taj set (ceo) korišćen za proveru. Rezultat je dobijen korišćenjem filtera Naive Bayes, 10-folds iteration, u programu Weka:

2 Ovo je deo izlaza koji je nama potreban. === Stratified cross-validation === === Summary === Correctly Classified Instances % Incorrectly Classified Instances % Kappa statistic Mean absolute error Root mean squared error Relative absolute error % Root relative squared error % Total Number of Instances 8124 === Detailed Accuracy By Class === TP Rate FP Rate Precision Recall F-Measure ROC Area Class Y N Weighted Avg === Confusion Matrix === a b <-- classified as a = Y b = N

3 Matrica konfuzije Matrica konfuzije se pravi za jednu klasu. Matrica je data u obliku: TP - True positives FP - False positives FN - False negatives TN - True negatives Pogledajmo šta to znači na gornjem primeru: a b <-- classified as a = Y b = N U ovom slučaju, matrica je data za klasu a (koja predstavlja Y - jestive pečurke). Značenje je sledeće: 4176 je true positives - to znači da je za 4176 instanci klasifikator predvideo da su jestive, i da je bio u pravu, da one zaista jesu jestive 32 je false negatives - model je za 32 instance predvideo da su otrovne, a one su zapravo jestive 307 je false positives - model je za 307 instanci predvideo da su jestive, a one su zapravo otrovne 3609 je true negatives - model je za 3609 instanci smatrao da su otrovne i bio je u pravu, one jesu otrovne. Odavde možemo videti koliko je model dobar - poželjno je da na glavnoj dijagonali matrice bude što veći broj elemenata - time je ispravnost modela veća. Čitanje matrice: po vertikali su brojevi elemenata koje je model klasifikovao u neku od klasa. Po horizontali su elementi koji zapravo pripadaju nekoj od klasa (prema podacima iz fajla). Matrica konfuzije za klasu b (N - otrovne pečurke) bi bila: a b <-- classified as a = N b = Y

4 Računanje pokazatelja ispravnosti Dobijeni rezultat: TP Rate FP Rate Precision Recall F-Measure ROC Area Class Y N Weighted Avg Matrice konfuzije: Klasa Y Klasa N a b <-- classified as a = Y b = N a b <-- classified as a = N b = Y Podaci TP Rate - pokazuje osetljivost modela. Ovaj broj pokazuje udeo dobro predviđenih instanci neke klase, u ukupnom broju instanci koje zapravo pripadaju toj klasi. Npr: Za klasu Y, model je od 4208 instanci, uspešno klasifikovao FP Rate - pokazuje specifičnost modela, tačnije njegovu sposobnost da uspešno klasifikuje negativne instance (tačnije one koje ne pripadaju klasi za koju se pravi matrica konfuzije). Npr: za klasu Y, od 3916 instanci koje su u drugoj klasi, pogrešio je u klasifikaciji 307 instanci.

5 Precision - pokazuje preciznost modela. Ovaj broj pokazuje udeo dobro predviđenih instanci neke klase, u ukupnom broju instanci koje je model svrstao u datu klasu. Drugim rečima, pokazuje koliki deo rezultata u jednoj klasi je uspešno klasifikovan. Npr: Za klasu Y, model je ukupno klasifikovao 4383 instance da će pripadati toj klasi, od toga, 4176 zaista pripada toj klasi. Recall je jednak TP Rate. F- measure je harmonijska sredina Recall i Precission. Služi da označi tačnost. ROC Area - to je površina ispod ROC krive. Ova kriva se dobija iscrtavanjem vrednosti koje predstavljaju odnos TP rate i FP rate. Ovo je kumulativna funkcija. Accuracy - nema ga u izveštaju a odnosi se na procenat uspešno klasifikovanih instanci u odnosu na ukupan broj instanci. Računa se kao: U našem slučaju to je:

6 Još malo o Recall i Precision. Ovo su dve najznačajnije mere ispravnosti modela. Recall možemo posmatrati kao meru kompletnosti (kvantiteta) a Precission kao meru tačnosti (kvaliteta) Recall pokazuje koliko je relevantnih rezultata algoritam vratio, a Precision pokazuje koliki je udeo relevantnih u nerelevantnim rezultatima koje je algoritam vratio. by Hijavata

KLASIFIKACIJA NAIVNI BAJES. NIKOLA MILIKIĆ URL:

KLASIFIKACIJA NAIVNI BAJES. NIKOLA MILIKIĆ   URL: KLASIFIKACIJA NAIVNI BAJES NIKOLA MILIKIĆ EMAIL: nikola.milikic@fon.bg.ac.rs URL: http://nikola.milikic.info ŠTA JE KLASIFIKACIJA? Zadatak određivanja klase kojoj neka instanca pripada instanca je opisana

More information

Projektovanje paralelnih algoritama II

Projektovanje paralelnih algoritama II Projektovanje paralelnih algoritama II Primeri paralelnih algoritama, I deo Paralelni algoritmi za množenje matrica 1 Algoritmi za množenje matrica Ovde su data tri paralelna algoritma: Direktan algoritam

More information

Mathcad sa algoritmima

Mathcad sa algoritmima P R I M J E R I P R I M J E R I Mathcad sa algoritmima NAREDBE - elementarne obrade - sekvence Primjer 1 Napraviti algoritam za sabiranje dva broja. NAREDBE - elementarne obrade - sekvence Primjer 1 POČETAK

More information

Red veze za benzen. Slika 1.

Red veze za benzen. Slika 1. Red veze za benzen Benzen C 6 H 6 je aromatično ciklično jedinjenje. Njegove dve rezonantne forme (ili Kekuléove structure), prema teoriji valentne veze (VB) prikazuju se uobičajeno kao na slici 1 a),

More information

TEORIJA SKUPOVA Zadaci

TEORIJA SKUPOVA Zadaci TEORIJA SKUPOVA Zadai LOGIKA 1 I. godina 1. Zapišite simbolima: ( x nije element skupa S (b) d je član skupa S () F je podskup slupa S (d) Skup S sadrži skup R 2. Neka je S { x;2x 6} = = i neka je b =

More information

CptS 570 Machine Learning School of EECS Washington State University. CptS Machine Learning 1

CptS 570 Machine Learning School of EECS Washington State University. CptS Machine Learning 1 CptS 570 Machine Learning School of EECS Washington State University CptS 570 - Machine Learning 1 IEEE Expert, October 1996 CptS 570 - Machine Learning 2 Given sample S from all possible examples D Learner

More information

LINEARNI MODELI STATISTIČKI PRAKTIKUM 2 2. VJEŽBE

LINEARNI MODELI STATISTIČKI PRAKTIKUM 2 2. VJEŽBE LINEARNI MODELI STATISTIČKI PRAKTIKUM 2 2. VJEŽBE Linearni model Promatramo jednodimenzionalni linearni model. Y = β 0 + p β k x k + ε k=1 x 1, x 2,..., x p - varijable poticaja (kontrolirane) ε - sl.

More information

Preliminarno ispitivanje sadrž aja slike pomoć u histograma slike koris ć enjem SVM algoritma i neuronske mrež e

Preliminarno ispitivanje sadrž aja slike pomoć u histograma slike koris ć enjem SVM algoritma i neuronske mrež e Preliminarno ispitivanje sadrž aja slike pomoć u histograma slike koris ć enjem SVM algoritma i neuronske mrež e Student Igor Valjević Mentor prof. dr. Vladimir Filipović Matematički fakultet Univerziteta

More information

BROJEVNE KONGRUENCIJE

BROJEVNE KONGRUENCIJE UNIVERZITET U NOVOM SADU PRIRODNO-MATEMATIČKI FAKULTET DEPARTMAN ZA MATEMATIKU I INFORMATIKU Vojko Nestorović BROJEVNE KONGRUENCIJE - MASTER RAD - Mentor, dr Siniša Crvenković Novi Sad, 2011. Sadržaj Predgovor...............................

More information

Smart Home Health Analytics Information Systems University of Maryland Baltimore County

Smart Home Health Analytics Information Systems University of Maryland Baltimore County Smart Home Health Analytics Information Systems University of Maryland Baltimore County 1 IEEE Expert, October 1996 2 Given sample S from all possible examples D Learner L learns hypothesis h based on

More information

Evaluation & Credibility Issues

Evaluation & Credibility Issues Evaluation & Credibility Issues What measure should we use? accuracy might not be enough. How reliable are the predicted results? How much should we believe in what was learned? Error on the training data

More information

CLASSIFICATION NAIVE BAYES. NIKOLA MILIKIĆ UROŠ KRČADINAC

CLASSIFICATION NAIVE BAYES. NIKOLA MILIKIĆ UROŠ KRČADINAC CLASSIFICATION NAIVE BAYES NIKOLA MILIKIĆ nikola.milikic@fon.bg.ac.rs UROŠ KRČADINAC uros@krcadinac.com WHAT IS CLASSIFICATION? A supervised learning task of determining the class of an instance; it is

More information

Performance Evaluation

Performance Evaluation Performance Evaluation Confusion Matrix: Detected Positive Negative Actual Positive A: True Positive B: False Negative Negative C: False Positive D: True Negative Recall or Sensitivity or True Positive

More information

Data Mining: Concepts and Techniques. (3 rd ed.) Chapter 8. Chapter 8. Classification: Basic Concepts

Data Mining: Concepts and Techniques. (3 rd ed.) Chapter 8. Chapter 8. Classification: Basic Concepts Data Mining: Concepts and Techniques (3 rd ed.) Chapter 8 1 Chapter 8. Classification: Basic Concepts Classification: Basic Concepts Decision Tree Induction Bayes Classification Methods Rule-Based Classification

More information

Matthew Piccoli University of Pennsylvania

Matthew Piccoli University of Pennsylvania Matthew Piccoli University of Pennsylvania Base Controller Lots of deprecated code Separate odometry from controller Use common basekinematics class Continuously updated throughout the summer Choices Safe

More information

SUPERVISED LEARNING: INTRODUCTION TO CLASSIFICATION

SUPERVISED LEARNING: INTRODUCTION TO CLASSIFICATION SUPERVISED LEARNING: INTRODUCTION TO CLASSIFICATION 1 Outline Basic terminology Features Training and validation Model selection Error and loss measures Statistical comparison Evaluation measures 2 Terminology

More information

Introduction to Supervised Learning. Performance Evaluation

Introduction to Supervised Learning. Performance Evaluation Introduction to Supervised Learning Performance Evaluation Marcelo S. Lauretto Escola de Artes, Ciências e Humanidades, Universidade de São Paulo marcelolauretto@usp.br Lima - Peru Performance Evaluation

More information

Zadatci sa ciklusima. Zadatak1: Sastaviti progra koji određuje z ir prvih prirod ih rojeva.

Zadatci sa ciklusima. Zadatak1: Sastaviti progra koji određuje z ir prvih prirod ih rojeva. Zadatci sa ciklusima Zadatak1: Sastaviti progra koji određuje z ir prvih prirod ih rojeva. StrToIntDef(tekst,broj) - funkcija kojom se tekst pretvara u ceo broj s tim da je uvedena automatska kontrola

More information

MAGNETIC FIELD OF ELECTRICAL RADIANT HEATING SYSTEM

MAGNETIC FIELD OF ELECTRICAL RADIANT HEATING SYSTEM UDK 537.612:697.27 DOI: 10.7562/SE2017.7.02.03 Original article www.safety.ni.ac.rs MIODRAG MILUTINOV 1 ANAMARIJA JUHAS 2 NEDA PEKARIĆ-NAĐ 3 1,2,3 University of Novi Sad, Faculty of Technical Sciences,

More information

Performance Evaluation and Comparison

Performance Evaluation and Comparison Outline Hong Chang Institute of Computing Technology, Chinese Academy of Sciences Machine Learning Methods (Fall 2012) Outline Outline I 1 Introduction 2 Cross Validation and Resampling 3 Interval Estimation

More information

Classifier Evaluation. Learning Curve cleval testc. The Apparent Classification Error. Error Estimation by Test Set. Classifier

Classifier Evaluation. Learning Curve cleval testc. The Apparent Classification Error. Error Estimation by Test Set. Classifier Classifier Learning Curve How to estimate classifier performance. Learning curves Feature curves Rejects and ROC curves True classification error ε Bayes error ε* Sub-optimal classifier Bayes consistent

More information

Šta je to mašinsko učenje?

Šta je to mašinsko učenje? MAŠINSKO UČENJE Šta je to mašinsko učenje? Disciplina koja omogućava računarima da uče bez eksplicitnog programiranja (Arthur Samuel 1959). 1. Generalizacija znanja na osnovu prethodnog iskustva (podataka

More information

Slika 1. Slika 2. Da ne bismo stalno izbacivali elemente iz skupa, mi ćemo napraviti još jedan niz markirano, gde će

Slika 1. Slika 2. Da ne bismo stalno izbacivali elemente iz skupa, mi ćemo napraviti još jedan niz markirano, gde će Permutacije Zadatak. U vreći se nalazi n loptica različitih boja. Iz vreće izvlačimo redom jednu po jednu lopticu i stavljamo jednu pored druge. Koliko različitih redosleda boja možemo da dobijemo? Primer

More information

Model Accuracy Measures

Model Accuracy Measures Model Accuracy Measures Master in Bioinformatics UPF 2017-2018 Eduardo Eyras Computational Genomics Pompeu Fabra University - ICREA Barcelona, Spain Variables What we can measure (attributes) Hypotheses

More information

ABOUT SOME VARIOUS INTERPRETATIONS OF THE FATIGUE CRITERION AT LOW NUMBER OF STRAIN CYCLES UDC Miodrag Janković

ABOUT SOME VARIOUS INTERPRETATIONS OF THE FATIGUE CRITERION AT LOW NUMBER OF STRAIN CYCLES UDC Miodrag Janković The Scientific Journal FACTA UNIVERSITATIS Series: Mechanical Engineering Vol.1, N o 8, 2001, pp. 955-964 ABOUT SOME VARIOUS INTERPRETATIONS OF THE FATIGUE CRITERION AT LOW NUMBER OF STRAIN CYCLES UDC

More information

FTN Novi Sad Katedra za motore i vozila. Drumska vozila Uputstvo za izradu vučnog proračuna motornog vozila. 1. Ulazni podaci IZVOR:

FTN Novi Sad Katedra za motore i vozila. Drumska vozila Uputstvo za izradu vučnog proračuna motornog vozila. 1. Ulazni podaci IZVOR: 1. Ulazni podaci IZVOR: WWW.CARTODAY.COM 1. Ulazni podaci Masa / težina vozila Osovinske reakcije Raspodela težine napred / nazad Dimenzije pneumatika Čeona površina Koeficijent otpora vazduha Brzinska

More information

Bayesian Decision Theory

Bayesian Decision Theory Introduction to Pattern Recognition [ Part 4 ] Mahdi Vasighi Remarks It is quite common to assume that the data in each class are adequately described by a Gaussian distribution. Bayesian classifier is

More information

Algoritam za množenje ulančanih matrica. Alen Kosanović Prirodoslovno-matematički fakultet Matematički odsjek

Algoritam za množenje ulančanih matrica. Alen Kosanović Prirodoslovno-matematički fakultet Matematički odsjek Algoritam za množenje ulančanih matrica Alen Kosanović Prirodoslovno-matematički fakultet Matematički odsjek O problemu (1) Neka je A 1, A 2,, A n niz ulančanih matrica duljine n N, gdje su dimenzije matrice

More information

Data Mining and Knowledge Discovery: Practice Notes

Data Mining and Knowledge Discovery: Practice Notes Data Mining and Knowledge Discovery: Practice Notes dr. Petra Kralj Novak Petra.Kralj.Novak@ijs.si 7.11.2017 1 Course Prof. Bojan Cestnik Data preparation Prof. Nada Lavrač: Data mining overview Advanced

More information

AIR CURTAINS VAZDU[NE ZAVESE V H

AIR CURTAINS VAZDU[NE ZAVESE V H AIR CURTAINS V 15.000 H 21.000 KLIMA Co. 2 KLIMA Co. Flow and system stress should be known factors in air flow. The flow is gas quantity flowing through the system during given time unit and is measured

More information

INVESTIGATION OF UPSETTING OF CYLINDER BY CONICAL DIES

INVESTIGATION OF UPSETTING OF CYLINDER BY CONICAL DIES INVESTIGATION OF UPSETTING OF CYLINDER BY CONICAL DIES D. Vilotic 1, M. Plancak M 1, A. Bramley 2 and F. Osman 2 1 University of Novi Sad, Yugoslavia; 2 University of Bath, England ABSTRACT Process of

More information

PRIPADNOST RJEŠENJA KVADRATNE JEDNAČINE DANOM INTERVALU

PRIPADNOST RJEŠENJA KVADRATNE JEDNAČINE DANOM INTERVALU MAT KOL Banja Luka) ISSN 0354 6969 p) ISSN 1986 58 o) Vol. XXI )015) 105 115 http://www.imvibl.org/dmbl/dmbl.htm PRIPADNOST RJEŠENJA KVADRATNE JEDNAČINE DANOM INTERVALU Bernadin Ibrahimpašić 1 Senka Ibrahimpašić

More information

Stephen Scott.

Stephen Scott. 1 / 35 (Adapted from Ethem Alpaydin and Tom Mitchell) sscott@cse.unl.edu In Homework 1, you are (supposedly) 1 Choosing a data set 2 Extracting a test set of size > 30 3 Building a tree on the training

More information

FIZIKALNA KOZMOLOGIJA VII. VRLO RANI SVEMIR & INFLACIJA

FIZIKALNA KOZMOLOGIJA VII. VRLO RANI SVEMIR & INFLACIJA FIZIKALNA KOZMOLOGIJA VII. VRLO RANI SVEMIR & INFLACIJA KOZMIČKI SAT ranog svemira Ekstra zračenje u mjerenju CMB Usporedba s rezultatima LEP-a Usporedba CMB i neutrina Vj.: Pozadinsko zračenje neutrina

More information

Uvod u relacione baze podataka

Uvod u relacione baze podataka Uvod u relacione baze podataka Ana Spasić 2. čas 1 Mala studentska baza dosije (indeks, ime, prezime, datum rodjenja, mesto rodjenja, datum upisa) predmet (id predmeta, sifra, naziv, bodovi) ispitni rok

More information

A SPECTRAL ATLAS OF λ BOOTIS STARS

A SPECTRAL ATLAS OF λ BOOTIS STARS Serb. Astron. J. 188 (2014), 75-84 UDC 524.3 355.3 DOI: 10.2298/SAJ1488075P Professional paper A SPECTRAL ATLAS OF λ BOOTIS STARS E. Paunzen 1 and U. Heiter 2 1 Department of Theoretical Physics and Astrophysics,

More information

DYNAMIC HEAT TRANSFER IN WALLS: LIMITATIONS OF HEAT FLUX METERS

DYNAMIC HEAT TRANSFER IN WALLS: LIMITATIONS OF HEAT FLUX METERS DYNAMI EAT TRANFER IN WALL: LIMITATION OF EAT FLUX METER DINAMIČKI PRENO TOPLOTE U ZIDOVIMA: OGRANIČENJA MERAČA TOPLOTNOG PROTOKA (TOPLOTNOG FLUKA) 1 I. Naveros a, b,. Ghiaus a a ETIL UMR58, INA-Lyon,

More information

ADAPTIVE NEURO-FUZZY MODELING OF THERMAL VOLTAGE PARAMETERS FOR TOOL LIFE ASSESSMENT IN FACE MILLING

ADAPTIVE NEURO-FUZZY MODELING OF THERMAL VOLTAGE PARAMETERS FOR TOOL LIFE ASSESSMENT IN FACE MILLING http://doi.org/10.24867/jpe-2017-01-016 JPE (2017) Vol.20 (1) Original Scientific Paper Kovač, P., Rodić, D., Gostimirović, M., Savković, B., Ješić. D. ADAPTIVE NEURO-FUZZY MODELING OF THERMAL VOLTAGE

More information

MATHEMATICAL ANALYSIS OF PERFORMANCE OF A VIBRATORY BOWL FEEDER FOR FEEDING BOTTLE CAPS

MATHEMATICAL ANALYSIS OF PERFORMANCE OF A VIBRATORY BOWL FEEDER FOR FEEDING BOTTLE CAPS http://doi.org/10.24867/jpe-2018-02-055 JPE (2018) Vol.21 (2) Choudhary, M., Narang, R., Khanna, P. Original Scientific Paper MATHEMATICAL ANALYSIS OF PERFORMANCE OF A VIBRATORY BOWL FEEDER FOR FEEDING

More information

An Algorithm for Computation of Bond Contributions of the Wiener Index

An Algorithm for Computation of Bond Contributions of the Wiener Index CROATICA CHEMICA ACTA CCACAA68 (1) 99-103 (1995) ISSN 0011-1643 CCA-2215 Original Scientific Paper An Algorithm for Computation of Bond Contributions of the Wiener Index Istvan Lukouits Central Research

More information

Linear Classifiers as Pattern Detectors

Linear Classifiers as Pattern Detectors Intelligent Systems: Reasoning and Recognition James L. Crowley ENSIMAG 2 / MoSIG M1 Second Semester 2014/2015 Lesson 16 8 April 2015 Contents Linear Classifiers as Pattern Detectors Notation...2 Linear

More information

Class 4: Classification. Quaid Morris February 11 th, 2011 ML4Bio

Class 4: Classification. Quaid Morris February 11 th, 2011 ML4Bio Class 4: Classification Quaid Morris February 11 th, 211 ML4Bio Overview Basic concepts in classification: overfitting, cross-validation, evaluation. Linear Discriminant Analysis and Quadratic Discriminant

More information

Mehurasto sortiranje Brzo sortiranje Sortiranje učešljavanjem Sortiranje umetanjem. Overviev Problemi pretraživanja Heš tabele.

Mehurasto sortiranje Brzo sortiranje Sortiranje učešljavanjem Sortiranje umetanjem. Overviev Problemi pretraživanja Heš tabele. Bubble sort Razmotrimo još jedan vrlo popularan algoritam sortiranja podataka, vrlo sličan prethodnom algoritmu. Algoritam je poznat pod nazivom Bubble sort algoritam (algoritam mehurastog sortiranja),

More information

The Solution to Assignment 6

The Solution to Assignment 6 The Solution to Assignment 6 Problem 1: Use the 2-fold cross-validation to evaluate the Decision Tree Model for trees up to 2 levels deep (that is, the maximum path length from the root to the leaves is

More information

NIPP. Implementing rules for metadata. Ivica Skender NSDI Working group for technical standards.

NIPP. Implementing rules for metadata. Ivica Skender NSDI Working group for technical standards. Implementing rules for metadata Ivica Skender NSDI Working group for technical standards ivica.skender@gisdata.com Content Working group for technical standards INSPIRE Metadata implementing rule Review

More information

Performance Evaluation

Performance Evaluation Performance Evaluation David S. Rosenberg Bloomberg ML EDU October 26, 2017 David S. Rosenberg (Bloomberg ML EDU) October 26, 2017 1 / 36 Baseline Models David S. Rosenberg (Bloomberg ML EDU) October 26,

More information

Strojno učenje 3 (I dio) Evaluacija modela. Tomislav Šmuc

Strojno učenje 3 (I dio) Evaluacija modela. Tomislav Šmuc Strojno učenje 3 (I dio) Evaluacija modela Tomislav Šmuc Pregled i. Greške (stvarna; T - na osnovu uzorka primjera) ii. Resampling metode procjene greške iii. Usporedba modela ili algoritama (na istim

More information

Regularization. CSCE 970 Lecture 3: Regularization. Stephen Scott and Vinod Variyam. Introduction. Outline

Regularization. CSCE 970 Lecture 3: Regularization. Stephen Scott and Vinod Variyam. Introduction. Outline Other Measures 1 / 52 sscott@cse.unl.edu learning can generally be distilled to an optimization problem Choose a classifier (function, hypothesis) from a set of functions that minimizes an objective function

More information

Diagnostics. Gad Kimmel

Diagnostics. Gad Kimmel Diagnostics Gad Kimmel Outline Introduction. Bootstrap method. Cross validation. ROC plot. Introduction Motivation Estimating properties of an estimator. Given data samples say the average. x 1, x 2,...,

More information

CONSTRUCTION OF GENERATOR CAPABILITY CURVES USING THE NEW METHOD FOR DETERMINATION OF POTIER REACTANCE

CONSTRUCTION OF GENERATOR CAPABILITY CURVES USING THE NEW METHOD FOR DETERMINATION OF POTIER REACTANCE CONSTRUCTION OF GENERATOR CAPABILITY CURVES USING THE NEW METHOD FOR DETERMINATION OF POTIER REACTANCE M.M. Kostić *, M. Ivanović *, B. Kostić *, S. Ilić** and D. Ćirić** Electrical Engineering Institute

More information

Philippe Jodin. Original scientific paper UDC: :519.6 Paper received:

Philippe Jodin. Original scientific paper UDC: :519.6 Paper received: The paper was presented at the Tenth Meeting New Trends in Fatigue and Fracture (NTF0) Metz, France, 30 August September, 00 Philippe Jodin APPLICATION OF NUMERICAL METHODS TO MIXED MODES FRACTURE MECHANICS

More information

.. Cal Poly CSC 466: Knowledge Discovery from Data Alexander Dekhtyar.. for each element of the dataset we are given its class label.

.. Cal Poly CSC 466: Knowledge Discovery from Data Alexander Dekhtyar.. for each element of the dataset we are given its class label. .. Cal Poly CSC 466: Knowledge Discovery from Data Alexander Dekhtyar.. Data Mining: Classification/Supervised Learning Definitions Data. Consider a set A = {A 1,...,A n } of attributes, and an additional

More information

Machine Learning in Action

Machine Learning in Action Machine Learning in Action Tatyana Goldberg (goldberg@rostlab.org) August 16, 2016 @ Machine Learning in Biology Beijing Genomics Institute in Shenzhen, China June 2014 GenBank 1 173,353,076 DNA sequences

More information

EXPERIMENTAL ANALYSIS OF THE STRENGTH OF A POLYMER PRODUCED FROM RECYCLED MATERIAL

EXPERIMENTAL ANALYSIS OF THE STRENGTH OF A POLYMER PRODUCED FROM RECYCLED MATERIAL A. Jurić et al. EXPERIMENTAL ANALYSIS OF THE STRENGTH OF A POLYMER PRODUCED FROM RECYCLED MATERIAL Aleksandar Jurić, Tihomir Štefić, Zlatko Arbanas ISSN 10-651 UDC/UDK 60.17.1/.:678.74..017 Preliminary

More information

ANALYSIS OF INFLUENCE OF PARAMETERS ON TRANSFER FUNCTIONS OF APERIODIC MECHANISMS UDC Života Živković, Miloš Milošević, Ivan Ivanov

ANALYSIS OF INFLUENCE OF PARAMETERS ON TRANSFER FUNCTIONS OF APERIODIC MECHANISMS UDC Života Živković, Miloš Milošević, Ivan Ivanov UNIVERSITY OF NIŠ The scientific journal FACTA UNIVERSITATIS Series: Mechanical Engineering Vol.1, N o 6, 1999 pp. 675-681 Editor of series: Nenad Radojković, e-mail: radojkovic@ni.ac.yu Address: Univerzitetski

More information

Iskazna logika 1. Matematička logika u računarstvu. oktobar 2012

Iskazna logika 1. Matematička logika u računarstvu. oktobar 2012 Matematička logika u računarstvu Department of Mathematics and Informatics, Faculty of Science,, Serbia oktobar 2012 Iskazi, istinitost, veznici Intuitivno, iskaz je rečenica koja je ima tačno jednu jednu

More information

CHAPTER 4: PREDICTION AND ESTIMATION OF RAINFALL DURING NORTHEAST MONSOON

CHAPTER 4: PREDICTION AND ESTIMATION OF RAINFALL DURING NORTHEAST MONSOON 43 CHAPTER 4: PREDICTION AND ESTIMATION OF RAINFALL DURING NORTHEAST MONSOON After analyzing the weather patterns and forecast of seasonal rainfall for Cauvery delta region, a method to predict the dry

More information

VELOCITY PROFILES AT THE OUTLET OF THE DIFFERENT DESIGNED DIES FOR ALUMINIUM EXTRUSION

VELOCITY PROFILES AT THE OUTLET OF THE DIFFERENT DESIGNED DIES FOR ALUMINIUM EXTRUSION VELOCITY PROFILES AT THE OUTLET OF THE DIFFERENT DESIGNED DIES FOR ALUMINIUM EXTRUSION J.Caloska, J. Lazarev, Faculty of Mechanical Engineering, University Cyril and Methodius, Skopje, Republic of Macedonia

More information

EXPERIMENTAL INVESTIGATION OF EXTRUSION SPEED AND TEMPERATURE EFFECTS ON ARITHMETIC MEAN SURFACE ROUGHNESS IN FDM- BUILT SPECIMENS

EXPERIMENTAL INVESTIGATION OF EXTRUSION SPEED AND TEMPERATURE EFFECTS ON ARITHMETIC MEAN SURFACE ROUGHNESS IN FDM- BUILT SPECIMENS EXPERIMENTAL INVESTIGATION OF EXTRUSION SPEED AND TEMPERATURE EFFECTS ON ARITHMETIC MEAN SURFACE ROUGHNESS IN FDM- BUILT SPECIMENS Ognjan Lužanin *, Dejan Movrin, Miroslav Plančak University of Novi Sad,

More information

Sveučilište J. J. Strossmayera u Osijeku Odjel za matematiku Sveučilišni nastavnički studij matematike i informatike. Sortiranje u linearnom vremenu

Sveučilište J. J. Strossmayera u Osijeku Odjel za matematiku Sveučilišni nastavnički studij matematike i informatike. Sortiranje u linearnom vremenu Sveučilište J. J. Strossmayera u Osijeku Odjel za matematiku Sveučilišni nastavnički studij matematike i informatike Tibor Pejić Sortiranje u linearnom vremenu Diplomski rad Osijek, 2011. Sveučilište J.

More information

Evaluation. Andrea Passerini Machine Learning. Evaluation

Evaluation. Andrea Passerini Machine Learning. Evaluation Andrea Passerini passerini@disi.unitn.it Machine Learning Basic concepts requires to define performance measures to be optimized Performance of learning algorithms cannot be evaluated on entire domain

More information

Analysis of Data Mining Techniques for Weather Prediction

Analysis of Data Mining Techniques for Weather Prediction ISSN (Print) : 0974-6846 ISSN (Online) : 0974-5645 Indian Journal of Science and Technology, Vol 9(38), DOI: 10.17485/ijst/2016/v9i38/101962, October 2016 Analysis of Data Mining Techniques for Weather

More information

Hypothesis Evaluation

Hypothesis Evaluation Hypothesis Evaluation Machine Learning Hamid Beigy Sharif University of Technology Fall 1395 Hamid Beigy (Sharif University of Technology) Hypothesis Evaluation Fall 1395 1 / 31 Table of contents 1 Introduction

More information

Lecture 4 Discriminant Analysis, k-nearest Neighbors

Lecture 4 Discriminant Analysis, k-nearest Neighbors Lecture 4 Discriminant Analysis, k-nearest Neighbors Fredrik Lindsten Division of Systems and Control Department of Information Technology Uppsala University. Email: fredrik.lindsten@it.uu.se fredrik.lindsten@it.uu.se

More information

PRECIPITATION FORECAST USING STATISTICAL APPROACHES UDC 55:311.3

PRECIPITATION FORECAST USING STATISTICAL APPROACHES UDC 55:311.3 FACTA UNIVERSITATIS Series: Working and Living Environmental Protection Vol. 10, N o 1, 2013, pp. 79-91 PRECIPITATION FORECAST USING STATISTICAL APPROACHES UDC 55:311.3 Mladjen Ćurić 1, Stanimir Ţivanović

More information

DEVELOPMENT OF A MATHEMATICAL MODEL TO PREDICT THE PERFORMANCE OF A VIBRATORY BOWL FEEDER FOR HEADED COMPONENTS

DEVELOPMENT OF A MATHEMATICAL MODEL TO PREDICT THE PERFORMANCE OF A VIBRATORY BOWL FEEDER FOR HEADED COMPONENTS http://doi.org/10.24867/jpe-2018-02-060 JPE (2018) Vol.21 (2) Tiwari, I., Laksha, Khanna, P. Original Scientific Paper DEVELOPMENT OF A MATHEMATICAL MODEL TO PREDICT THE PERFORMANCE OF A VIBRATORY BOWL

More information

Programiranje u realnom vremenu Bojan Furlan

Programiranje u realnom vremenu Bojan Furlan Programiranje u realnom vremenu Bojan Furlan Tri procesa sa D = T imaju sledeće karakteristike: Proces T C a 3 1 b 6 2 c 18 5 (a) Pokazati kako se može konstruisati ciklično izvršavanje ovih procesa. (b)

More information

ZANIMLJIV NAČIN IZRAČUNAVANJA NEKIH GRANIČNIH VRIJEDNOSTI FUNKCIJA. Šefket Arslanagić, Sarajevo, BiH

ZANIMLJIV NAČIN IZRAČUNAVANJA NEKIH GRANIČNIH VRIJEDNOSTI FUNKCIJA. Šefket Arslanagić, Sarajevo, BiH MAT-KOL (Banja Luka) XXIII ()(7), -7 http://wwwimviblorg/dmbl/dmblhtm DOI: 75/МК7A ISSN 5-6969 (o) ISSN 986-588 (o) ZANIMLJIV NAČIN IZRAČUNAVANJA NEKIH GRANIČNIH VRIJEDNOSTI FUNKCIJA Šefket Arslanagić,

More information

Data Mining and Analysis: Fundamental Concepts and Algorithms

Data Mining and Analysis: Fundamental Concepts and Algorithms Data Mining and Analysis: Fundamental Concepts and Algorithms dataminingbook.info Mohammed J. Zaki 1 Wagner Meira Jr. 2 1 Department of Computer Science Rensselaer Polytechnic Institute, Troy, NY, USA

More information

O homomorfizam-homogenim geometrijama ranga 2

O homomorfizam-homogenim geometrijama ranga 2 UNIVERZITET U NOVOM SADU PRIRODN0-MATEMATIČKI FAKULTET DEPARTMAN ZA MATEMATIKU I INFORMATIKU Eva Jungael O homomorfzam-homogenm geometrjama ranga 2 -završn rad- Nov Sad, oktoar 2009 Predgovor Za strukturu

More information

Evaluation requires to define performance measures to be optimized

Evaluation requires to define performance measures to be optimized Evaluation Basic concepts Evaluation requires to define performance measures to be optimized Performance of learning algorithms cannot be evaluated on entire domain (generalization error) approximation

More information

U X. 1. Multivarijantna statistička analiza 1

U X. 1. Multivarijantna statistička analiza 1 . Multivarijantna statistička analiza Standardizovana (normalizovana) vrednost obeležja Normalizovano odstupanje je mera varijacije koja pokazuje algebarsko odstupanje jedne vrednosti obeležja od aritmetičke

More information

Performance. Learning Classifiers. Instances x i in dataset D mapped to feature space:

Performance. Learning Classifiers. Instances x i in dataset D mapped to feature space: Learning Classifiers Performance Instances x i in dataset D mapped to feature space: Initial Model (Assumptions) Training Process Trained Model Testing Process Tested Model Performance Performance Training

More information

Performance evaluation of binary classifiers

Performance evaluation of binary classifiers Performance evaluation of binary classifiers Kevin P. Murphy Last updated October 10, 2007 1 ROC curves We frequently design systems to detect events of interest, such as diseases in patients, faces in

More information

DEVELOPMENT OF MATHEMATICAL MODELS TO PREDICT THE EFFECT OF INPUT PARAMETERS ON FEED RATE OF A RECIPROCATORY TUBE FUNNEL FEEDER

DEVELOPMENT OF MATHEMATICAL MODELS TO PREDICT THE EFFECT OF INPUT PARAMETERS ON FEED RATE OF A RECIPROCATORY TUBE FUNNEL FEEDER http://doi.org/10.24867/jpe-2018-01-067 JPE (2018) Vol.21 (1) Jain, A., Bansal, P., Khanna, P. Preliminary Note DEVELOPMENT OF MATHEMATICAL MODELS TO PREDICT THE EFFECT OF INPUT PARAMETERS ON FEED RATE

More information

Stats notes Chapter 5 of Data Mining From Witten and Frank

Stats notes Chapter 5 of Data Mining From Witten and Frank Stats notes Chapter 5 of Data Mining From Witten and Frank 5 Credibility: Evaluating what s been learned Issues: training, testing, tuning Predicting performance: confidence limits Holdout, cross-validation,

More information

THE USE OF SCRIPT IN THE SOFTWARE GEMCOM ***

THE USE OF SCRIPT IN THE SOFTWARE GEMCOM *** MINING AND METALLURGY INSTITUTE BOR UDK: 622 ISSN: 2334-8836 (Štampano izdanje) ISSN: 2406-1395 (Online) UDK: 681.51:551:517.1(045)=111 doi:10.5937/mmeb1504053v Abstract Zoran Vaduvesković *, Daniel Kržanović

More information

Lecture Slides for INTRODUCTION TO. Machine Learning. ETHEM ALPAYDIN The MIT Press,

Lecture Slides for INTRODUCTION TO. Machine Learning. ETHEM ALPAYDIN The MIT Press, Lecture Slides for INTRODUCTION TO Machine Learning ETHEM ALPAYDIN The MIT Press, 2004 alpaydin@boun.edu.tr http://www.cmpe.boun.edu.tr/~ethem/i2ml CHAPTER 14: Assessing and Comparing Classification Algorithms

More information

Least Squares Classification

Least Squares Classification Least Squares Classification Stephen Boyd EE103 Stanford University November 4, 2017 Outline Classification Least squares classification Multi-class classifiers Classification 2 Classification data fitting

More information

Dynamic Clustering-Based Estimation of Missing Values in Mixed Type Data

Dynamic Clustering-Based Estimation of Missing Values in Mixed Type Data Dynamic Clustering-Based Estimation of Missing Values in Mixed Type Data Vadim Ayuyev, Joseph Jupin, Philip Harris and Zoran Obradovic Temple University, Philadelphia, USA 2009 Real Life Data is Often

More information

ANALYTICAL AND NUMERICAL PREDICTION OF SPRINGBACK IN SHEET METAL BENDING

ANALYTICAL AND NUMERICAL PREDICTION OF SPRINGBACK IN SHEET METAL BENDING ANALYTICAL AND NUMERICAL PREDICTION OF SPRINGBACK IN SHEET METAL BENDING Slota Ján, Jurčišin Miroslav Department of Technologies and Materials, Faculty of Mechanical Engineering, Technical University of

More information

On the relation between Zenkevich and Wiener indices of alkanes

On the relation between Zenkevich and Wiener indices of alkanes J.Serb.Chem.Soc. 69(4)265 271(2004) UDC 547.21:54 12+539.6 JSCS 3152 Original scientific paper On the relation between Zenkevich and Wiener indices of alkanes IVAN GUTMAN a*, BORIS FURTULA a, BILJANA ARSI]

More information

ANALYSIS OF THE RELIABILITY OF THE "ALTERNATOR- ALTERNATOR BELT" SYSTEM

ANALYSIS OF THE RELIABILITY OF THE ALTERNATOR- ALTERNATOR BELT SYSTEM I. Mavrin, D. Kovacevic, B. Makovic: Analysis of the Reliability of the "Alternator- Alternator Belt" System IVAN MAVRIN, D.Sc. DRAZEN KOVACEVIC, B.Eng. BRANKO MAKOVIC, B.Eng. Fakultet prometnih znanosti,

More information

Osobine metode rezolucije: zaustavlja se, pouzdanost i kompletnost. Iskazna logika 4

Osobine metode rezolucije: zaustavlja se, pouzdanost i kompletnost. Iskazna logika 4 Matematička logika u računarstvu Department of Mathematics and Informatics, Faculty of Science,, Serbia novembar 2012 Rezolucija 1 Metod rezolucije je postupak za dokazivanje da li je neka iskazna (ili

More information

DETERMINATION OF THE EFFECTIVE STRAIN FLOW IN COLD FORMED MATERIAL

DETERMINATION OF THE EFFECTIVE STRAIN FLOW IN COLD FORMED MATERIAL DETERMINATION OF THE EFFECTIVE STRAIN FLOW IN COLD FORMED MATERIAL Leo Gusel University of Maribor, Faculty of Mechanical Engineering Smetanova 17, SI 000 Maribor, Slovenia ABSTRACT In the article the

More information

MACHINE LEARNING ADVANCED MACHINE LEARNING

MACHINE LEARNING ADVANCED MACHINE LEARNING MACHINE LEARNING ADVANCED MACHINE LEARNING Recap of Important Notions on Estimation of Probability Density Functions 2 2 MACHINE LEARNING Overview Definition pdf Definition joint, condition, marginal,

More information

AN EXPERIMENTAL METHOD FOR DETERMINATION OF NATURAL CIRCULAR FREQUENCY OF HELICAL TORSIONAL SPRINGS UDC:

AN EXPERIMENTAL METHOD FOR DETERMINATION OF NATURAL CIRCULAR FREQUENCY OF HELICAL TORSIONAL SPRINGS UDC: UNIVERSITY OF NIŠ The scientific journal FACTA UNIVERSITATIS Series: Mechanical Engineering Vol.1, N o 5, 1998 pp. 547-554 Editor of series: Nenad Radojković, e-mail: radojkovic@ni.ac.yu Address: Univerzitetski

More information

BANA 7046 Data Mining I Lecture 4. Logistic Regression and Classications 1

BANA 7046 Data Mining I Lecture 4. Logistic Regression and Classications 1 BANA 7046 Data Mining I Lecture 4. Logistic Regression and Classications 1 Shaobo Li University of Cincinnati 1 Partially based on Hastie, et al. (2009) ESL, and James, et al. (2013) ISLR Data Mining I

More information

Machine Learning for natural language processing

Machine Learning for natural language processing Machine Learning for natural language processing Classification: Naive Bayes Laura Kallmeyer Heinrich-Heine-Universität Düsseldorf Summer 2016 1 / 20 Introduction Classification = supervised method for

More information

APPLICATION OF THOMAS-FERMI MODEL TO FULLERENE MOLECULE AND NANOTUBE UDC 547. Yuri Kornyushin

APPLICATION OF THOMAS-FERMI MODEL TO FULLERENE MOLECULE AND NANOTUBE UDC 547. Yuri Kornyushin FACTA UNIVERSITATIS Series: Physics, Chemistry and Technology Vol. 5, N o 1, 2007, pp. 11-18 DOI: 10.2298/FUPCT0701011K APPLICATION OF THOMAS-FERMI MODEL TO FULLERENE MOLECULE AND NANOTUBE UDC 547 Yuri

More information

Data Mining and Knowledge Discovery. Petra Kralj Novak. 2011/11/29

Data Mining and Knowledge Discovery. Petra Kralj Novak. 2011/11/29 Data Mining and Knowledge Discovery Petra Kralj Novak Petra.Kralj.Novak@ijs.si 2011/11/29 1 Practice plan 2011/11/08: Predictive data mining 1 Decision trees Evaluating classifiers 1: separate test set,

More information

Metode izračunavanja determinanti matrica n-tog reda

Metode izračunavanja determinanti matrica n-tog reda Osječki matematički list 10(2010), 31 42 31 STUDENTSKA RUBRIKA Metode izračunavanja determinanti matrica n-tog reda Damira Keček Sažetak U članku su opisane metode izračunavanja determinanti matrica n-tog

More information

Određivanje koncentracija dve reaktivne boje u bojenom pamučnom materijalu

Određivanje koncentracija dve reaktivne boje u bojenom pamučnom materijalu Određivanje koncentracija dve reaktivne boje u bojenom pamučnom materijalu Milena Miljković 1, Milovan Purenović 1, Miodrag Stamenković, Milica Petrović 1 1 Univerzitet u Nišu, Prirodno matematički fakultet,

More information

GIS AND REMOTE SENSING APPLICATION IN GEOLOGICAL MAPPING AND 3D TERRAIN MODELING: A CASE STUDY IN EGHEI UPLIFT, LIBYA

GIS AND REMOTE SENSING APPLICATION IN GEOLOGICAL MAPPING AND 3D TERRAIN MODELING: A CASE STUDY IN EGHEI UPLIFT, LIBYA Geographic information systems SYNTHESIS 2015 International Scientific Conference of IT and Business-Related Research GIS AND REMOTE SENSING APPLICATION IN GEOLOGICAL MAPPING AND 3D TERRAIN MODELING: A

More information

KINETIKA UMREŽAVANJA SMEŠA ALKID/MELAMINSKA SMOLA

KINETIKA UMREŽAVANJA SMEŠA ALKID/MELAMINSKA SMOLA MIRJANA C. JOVIČIĆ RADMILA Ž. RADIČEVIĆ Tehnološki fakultet, Univerzitet u Novom Sadu, Novi Sad, Srbija NAUČNI RAD UDK 667.633.26:665.944:66.09 DOI: 10.2298/HEMIND0906629J KINETIKA UMREŽAVANJA SMEŠA ALKID/MELAMINSKA

More information

15-388/688 - Practical Data Science: Nonlinear modeling, cross-validation, regularization, and evaluation

15-388/688 - Practical Data Science: Nonlinear modeling, cross-validation, regularization, and evaluation 15-388/688 - Practical Data Science: Nonlinear modeling, cross-validation, regularization, and evaluation J. Zico Kolter Carnegie Mellon University Fall 2016 1 Outline Example: return to peak demand prediction

More information

IMPROVEMENT OF HIPPARCOS PROPER MOTIONS IN DECLINATION

IMPROVEMENT OF HIPPARCOS PROPER MOTIONS IN DECLINATION Serb. Astron. J. 172 (2006), 41-51 UDC 521.96 DOI: 10.2298/SAJ0672041D Preliminary report IMPROVEMENT OF HIPPARCOS PROPER MOTIONS IN DECLINATION G. Damljanović 1, N. Pejović 2 and B. Jovanović 1 1 Astronomical

More information

Asian Journal of Science and Technology Vol. 4, Issue 08, pp , August, 2013 RESEARCH ARTICLE

Asian Journal of Science and Technology Vol. 4, Issue 08, pp , August, 2013 RESEARCH ARTICLE Available Online at http://www.journalajst.com ASIAN JOURNAL OF SCIENCE AND TECHNOLOGY ISSN: 0976-3376 Asian Journal of Science and Technology Vol. 4, Issue 08, pp.037-041, August, 2013 RESEARCH ARTICLE

More information

CSC314 / CSC763 Introduction to Machine Learning

CSC314 / CSC763 Introduction to Machine Learning CSC314 / CSC763 Introduction to Machine Learning COMSATS Institute of Information Technology Dr. Adeel Nawab More on Evaluating Hypotheses/Learning Algorithms Lecture Outline: Review of Confidence Intervals

More information

Uvod u analizu (M3-02) 05., 07. i 12. XI dr Nenad Teofanov. principle) ili Dirihleov princip (engl. Dirichlet box principle).

Uvod u analizu (M3-02) 05., 07. i 12. XI dr Nenad Teofanov. principle) ili Dirihleov princip (engl. Dirichlet box principle). Uvod u analizu (M-0) 0., 07. i. XI 0. dr Nenad Teofanov. Kardinalni broj skupa R U ovom predavanju se razmatra veličina skupa realnih brojeva. Jasno, taj skup ima beskonačno mnogo elemenata. Pokazaće se,

More information