DOCSLIB.ORG

Learning of Decision Fusion Mappings for Pattern Recognition

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Learning of Decision Fusion Mappings for Pattern Recognition Learning of Decision Fusion Mappings for Pattern Recognition Friedhelm Schwenker, Christian R. Dietrich, Christian Thiel, Gunther¨ Palm Department of Neural Information Processing Universitat¨ Ulm, 89069 Ulm, Germany fi[email protected] Abstract 2. Training of the Fusion Layer performing a mapping of the classifier outputs (soft or crisp decisions) into the Different learning algorithms for the decision fusion set of desired class labels. mapping of a multiple classifier system are compared in this This two-phase MCS training is very similar to RBF learn- paper. It is very well known that the confusion matrices ing procedures, where the network is learned over two or of the individual classifiers are utilised in the naive Bayes even three learning phases: combination of classifier outputs. After a brief review of the decision templates, the linear associative memory and the 1. Training of the RBF kernel parameters (centres and pseudoinverse matrix approaches it is demonstrated that all widths) of the first layer through clustering or vector four adaptive decision fusion mappings share the confusion quantization. matrices as the essential ingredient. 2. Supervised learning of the output layer (second layer) of the network. 1 Introduction 3. Training both layers of the RBF network through a backpropagation-like optimisation procedure. During the last years, a community has emerged that is Typically, for these three-phase learning scheme a labeled interested in the fusion of classifiers in the context of so- training set is used to calculate the radial basis function called multiple classifier systems (MCS). One goal is to centres, the radial basis function widths and the weights of design classifier systems with a high classification accu- the output layer, but it should be noticed that other learning racy ([7]...[13],[11]).The reason behind this new approach schemes may be applied. For instance, if a large set of unla- is that in traditional pattern recognition, the best individ- belled data points is available this data can be used to train ual classifier has to be found, which is difficult to identity. the parameters of the first RBF layer (centres and widths) Also, a single classifier can not make use of information through unsupervised clustering [17]. from other discrimination functions. For the training of the two layer fusion architecture two Typical fusion strategies for the combination of classi- labeled data sets may be used. It is assumed that I classifier fiers in MCS are based on fixed fusion mappings, for in- decisions, here calculated on I different features, have to be stance averaging or multiplying the classifier outputs [1, 6]. combined. First, the classifiers Ci, i = 1, ..., I, are trained, In this paper we consider supervised learning methods to by utilising a training set, and then another labeled training train the fusion mapping of the MCS. Whereas the MCS ar- set R, which is not necessarily completely disjunct to the chitecture (see Figure 1) is very similar to layered artificial first, is sent to the previously trained first level classifiers to neural networks, like radial basis function networks (RBFs) calculate the individual classifier outputs. These classifier or multilayer perceptrons (MLPs), the training is rather dif- i outputs C (xi), i = 1, ..., I together with the desired class ferent. For most artificial neural network architectures, e.g. labels are then used to train the decision fusion mapping F. MLPs, the parameters are typically trained simultaneously To train this fusion mapping different algorithms have been through a backpropagation-like training algorithm by a one- proposed. phase learning procedure. On the other hand a MCS is typ- In this paper decision templates [9, 10], naive Bayes ically trained by two completely different training phases: decision fusion [19], linear matrix memory [4], and Pseu- 1. Building the Classifier Layer consisting of a set of clas- doinverse matrix [4] methods are studied and links between sifiers where each classifier is trained separately, e.g. these fusion mappings are shown. The mappings are intro- each classifier is trained on a specific feature subset. duced in Section 2, while Section 3 presents experimental results on their performance. The closing Section explores stored classifier outputs Ci are adapted through a Hebbian the close links between the fusion mappings. learning rule [15] and V i is given as the product of the clas- sifier outputs Ci and the desired classifier outputs Y : 2 Adaptive fusion mappings i T V := YCi . (4) | {z } i In this Section we briefly describe methods to train adap- W tive fusion mappings. It is assumed that a set of first level In the case of crisp classifiers the matrix V i is equal to the classifiers C1, ..., CI has been trained through a supervised i i confusion matrix of classifier C , where Vω,ω∗ is equal to learning procedure using a labeled training set. This corre- the number of samples of class ω in the training set which sponds to phase 1 of the two-phase MCS learning scheme. were assigned by Ci to class ω∗ [19]. For soft classifiers the To fix the notation, let Ω = {1, ..., L} be a set of class ω-th row of V i contains the accummulated soft classifier labels and Ci(xµ) ∈ ∆ be the probabilistic classifier output i µ ω 1 i decisions of C for the feature vectors xi ∈ R . i µ of the i-th classifier C given the input feature vector xi ∈ In the classification phase these matrices are then used to R, with ∆ defined as combine the individual classifier decisions to calculate the overall classifier decision (see Eq 3). For a feature vector L X X = (x , ..., x ) Ci(x ) ∆ := {(y , ..., y ) ∈ [0, 1]L| y = 1}. (1) 1 I , the classifier outputs i are applied 1 L l to the matrices V i, i = 1, ..., I and the outputs zi ∈ IRL are l=1 given by i i i T Then for each classifier the outputs for the training set R z := V (C (xi)) . (5) are given by a (L × M)-matrix C , i = 1, ..., I where M = i The combined class membership estimate based on I fea- |R| and the µ-th column of Ci contains the classifier output i µ T ture vectors is then calculated as the average of the individ- C (xi ) . Here the superscript T denotes the transposition. µ µ ual outputs of the second level classifiers The desired classifier outputs ω ∈ Ω for inputs xi ∈ R are given by the (L × M)-matrix Y defined by the 1 of L I I X i X i i T encoding scheme for class labels z := z = V C (xi) (6) i=1 i=1 ( 1, l = ωµ Yl,µ = . (2) and the final class label based on the combined class mem- 0, otherwise bership estimate z is then determined by the maximum membership rule That is, corresponding to Ci the µ-th column Y·,µ ∈ ∆ contains the binary coded target output of feature vector ω := argmax(zl). (7) µ l∈Ω xi ∈ R. In the following we discuss different approaches to com- i 2.2 Decision Templates bine the classifier outputs C (xi), i = 1, ..., I into an overall classifier decision The concept of decision templates is a simple, intuitive, 1 I z := F(C (x1), ..., C (xI )). (3) and robust aggregation idea that evolved from the fuzzy tem- plate which was introduced by KUNCHEVA, see [10, 11]. The four different learning schemes, namely linear associa- Decision templates are calculated as the mean of the classi- µ tive memory, decision template, pseudoinverse matrix and fier outputs for inputs xi of class ω ∈ Ω naive Bayes are introduced to calculate the decision fusion 1 X mapping F : ∆I → ∆ (see Eq. 3). In all these methods T ω := C(xµ). (8) i |Rω| i the fusion mapping F is realised through (L × L)-matrices µ ω xi ∈R V 1, ..., V I , calculated through a certain training algorithm. The overall classifier system is depicted in Figure 1. In the case of I input features for each class the decision template T ω is given by the (I × L)-matrix 2.1 Linear Associative Memory ω T1 T ω := . ∈ ∆I . (9) A linear decision fusion mapping F may be re- . ω alised through an associative matrix memory whose error- TI correcting properties have been shown in several numerical 1Independent from the classifier type (soft or crisp) we consider W i a experiments and theoretical investigations [8]. In order to confusion matrix. It is defined to compare the individual fusion schemes calculate the memory matrix V i for each classifier Ci the (see Eq. 4,11,12 and 14). Classifier Layer Fusion Layer 1 C 1 (x1) 1 Feature 1 V z : : : i : Decision C (x ) : i Fusion i i z Feature i V : I : Final : : : I : C (xI) z Classification Feature I V I z Classification Transformation Figure 1. Two layer MCS architecture consisting of a classifier layer and an additional fusion layer. In i general the combination of the classifier outputs C (xi), i = 1, ..., I is accomplished through a fusion 1 I mapping F(C (x1), ..., C (xI )). In this paper we restrict ourselves to separable linear fusion mappings, i i where the classifier outputs C (xi) are multiplied with matrices V , i = 1, ..., I. The resulting decisions z1, ..., zI are combined with decision fusion. To produce independent answers, the classifiers in this example work on different features. In order to align the decision template combining scheme a set of input vectors X = (x1, ..., xI ) and a set of classifier i in such a way that the combination is done as proposed in outputs C (xi), i = 1, ..., I the combined class membership Figure 1, for each feature space a linear matrix operator V i estimate is given by Eq.
Load more

Recommended publications
  • CMPSCI 585 Programming Assignment 1 Spam Filtering Using Naive Bayes
    CMPSCI 585 Programming Assignment 1 Out: February 5, 2003 Due: February 17, 2003 Spam Filtering using Naive Bayes Naive Bayes is a simple, effective machine learning solution to the problem of document classification. For this assignment, you will implement a naive Bayes classifier to classify email messages as either spam (junk mail) or ham (legitimate messages). Data The data was collected from http://spamassassin.org/publiccorpus/. For this assignment, use the slightly modified version found at http://canberra.cs.umass.edu/~culotta/cs585/ass1-data.tgz (8.3M). The data consists of 4150 ham messages and 1897 spam messages, with original header information intact. Tasks Read data: Read in all messages and store them in some efficient manner. To do this you must decide how to tokenize each message – i.e. designate which characters indicate the start of a new “word.” See http://www.paulgrahm.com/spam.html for one way of doing this. (You will probably also assign each unique word an integer index you can use as an index into arrays of word counts and word proabilities.) Split Data: Randomly split the messages into a training set (70% of the messages) and a testing set (30%). Train Classifier: Using the training set only, estimate and store the prior class distributions P (spam) and P (ham), as well as the conditional probability distributions P (w|spam) and P (w|ham). Test Classifier: Classify each message in the testing set as spam or ham according to the Naive Bayes formulation. Note that you will most likely run out of floating-point resolution unless you do the product over words 1 Spam Ham Spam TP FP Ham FN TN Table 1: Confusion Matrix: TP = “true positive”, TN = “true negative”, FP = “false positive”, ‘FN = “false negative”.
    [Show full text]
  • Using Machine Learning to Improve Dense and Sparse Matrix Multiplication Kernels
    Iowa State University Capstones, Theses and Graduate Theses and Dissertations Dissertations 2019 Using machine learning to improve dense and sparse matrix multiplication kernels Brandon Groth Iowa State University Follow this and additional works at: https://lib.dr.iastate.edu/etd Part of the Applied Mathematics Commons, and the Computer Sciences Commons Recommended Citation Groth, Brandon, "Using machine learning to improve dense and sparse matrix multiplication kernels" (2019). Graduate Theses and Dissertations. 17688. https://lib.dr.iastate.edu/etd/17688 This Dissertation is brought to you for free and open access by the Iowa State University Capstones, Theses and Dissertations at Iowa State University Digital Repository. It has been accepted for inclusion in Graduate Theses and Dissertations by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected]. Using machine learning to improve dense and sparse matrix multiplication kernels by Brandon Micheal Groth A dissertation submitted to the graduate faculty in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Major: Applied Mathematics Program of Study Committee: Glenn R. Luecke, Major Professor James Rossmanith Zhijun Wu Jin Tian Kris De Brabanter The student author, whose presentation of the scholarship herein was approved by the program of study committee, is solely responsible for the content of this dissertation. The Graduate College will ensure this dissertation is globally accessible and will not permit alterations after a degree is conferred. Iowa State University Ames, Iowa 2019 Copyright c Brandon Micheal Groth, 2019. All rights reserved. ii DEDICATION I would like to dedicate this thesis to my wife Maria and our puppy Tiger.
    [Show full text]
  • Arxiv:1902.03680V3 [Cs.LG] 17 Jun 2019 1
    Learning From Noisy Labels By Regularized Estimation Of Annotator Confusion Ryutaro Tanno1 ∗ Ardavan Saeedi2 Swami Sankaranarayanan2 Daniel C. Alexander1 Nathan Silberman2 1University College London, UK 2Butterfly Network, New York, USA 1 2 fr.tanno, [email protected] fasaeedi,swamiviv,[email protected] Abstract tations is modest or the tasks are ambiguous. For example, The predictive performance of supervised learning algo- vision applications in medical image analysis [2] require an- rithms depends on the quality of labels. In a typical label notations from clinical experts, which incur high costs and collection process, multiple annotators provide subjective often suffer from high inter-reader variability [3, 4, 5, 6]. noisy estimates of the “truth” under the influence of their However, if the exact process by which each annotator varying skill-levels and biases. Blindly treating these noisy generates the labels was known, we could correct the an- labels as the ground truth limits the accuracy of learning al- notations accordingly and thus train our model on a cleaner gorithms in the presence of strong disagreement. This prob- set of data. Furthermore, this additional knowledge of the lem is critical for applications in domains such as medical annotators’ skills can be utilized to decide on which exam- imaging where both the annotation cost and inter-observer ples to be labeled by which annotators [7, 8, 9]. Therefore, variability are high. In this work, we present a method for methods that can accurately model the label noise of anno- simultaneously learning the individual annotator model and tators are useful for improving not only the accuracy of the the underlying true label distribution, using only noisy ob- trained model, but also the quality of labels in the future.
    [Show full text]
  • Performance Metric Elicitation from Pairwise Classifier Comparisons Arxiv:1806.01827V2 [Stat.ML] 18 Jan 2019
    Performance Metric Elicitation from Pairwise Classifier Comparisons Gaurush Hiranandani† Shant Boodaghians† Ruta Mehta† [email protected] [email protected] [email protected] Oluwasanmi Koyejo† [email protected] January 21, 2019 Abstract Given a binary prediction problem, which performance metric should the classifier optimize? We address this question by formalizing the problem of Metric Elicitation. The goal of metric elicitation is to discover the performance metric of a practitioner, which reflects her innate rewards (costs) for correct (incorrect) classification. In particular, we focus on eliciting binary classification performance metrics from pairwise feedback, where a practitioner is queried to provide relative preference between two classifiers. By exploiting key geometric properties of the space of confusion matrices, we obtain provably query efficient algorithms for eliciting linear and linear-fractional performance metrics. We further show that our method is robust to feedback and finite sample noise. 1 Introduction Selecting an appropriate performance metric is crucial to the real-world utility of predictive machine learning. Specialized teams of statisticians and economists are routinely hired in the arXiv:1806.01827v2 [stat.ML] 18 Jan 2019 industry to monitor many metrics { since optimizing the wrong metric directly translates into lost revenue [6]. Medical predictions are another important application, where ignoring cost sensitive trade-offs can directly impact lives [23]. Unfortunately, there is scant formal guidance
    [Show full text]
  • MASTER TP^ Ostrlbunon of THIS DOCUMENT IS UNL.M.TED Kaon Content of Three-Prong Decays of the Tan Lepton
    LBL--30035 DE91 007647 Kaon Content of Three-Prong Decays of the Tau Lepton* James Jackson Eastman Ph.D. Thesis "This work was supported by the Director, Office of Energy Research, Office of High Energy and Nuclear Physics, Division of High Energy Physics, of the U.S. Department of Energy under Contract No. DE-AC03-76SF00098. MASTER TP^ OSTRlBUnON OF THIS DOCUMENT IS UNL.M.TED Kaon Content of Three-Prong Decays of the Tan Lepton James Jackson Eastman Ph.D. Thesis Department of Physics University of California, Berkeley, CA 94720 and Lawrence Berkeley Laboratory, University of California, Berkeley, CA 94720 December 2, 1990 Abstract We present a series of measurements involving the production of charged kaons in three-prong hadronic decays of the T lepton. The data sample was obtained with the TPC/Two-Gamma detector facility at PEP. We set a limit on the branch­ ing fraction BR(T~ -> vrK~K°) < 0.26% at the 95% confidence level. The pro­ cess T~ —» v,K~Ka is related via SU(3) to the second-class current decay r~ —» vTit~t). We also present new measurements of the three-prong branching frac­ + + tions BR{T~ -» vTK~ir Tr- + neutrals) = 0.70 ±8:i?% and BR(T~ -> t/TK-K ir + neutrals) = 0.16 +g:J?%. This work is supported by the United States Department of Energy under Contract DE-AC03-76SF00098. ii FOR Ann. Acknowledgements HIGH-ENERGY PHYSICS EXPERIMENTS are great big projects, and I want to ex­ press my gratitude to the many scientists, engineers, technicians and programmers who built and ran the PEP4/9 experiment.
    [Show full text]
  • Designing Alternative Representations of Confusion Matrices to Support Non-Expert Public Understanding of Algorithm Performance
    153 Designing Alternative Representations of Confusion Matrices to Support Non-Expert Public Understanding of Algorithm Performance HONG SHEN, Carnegie Mellon University, USA HAOJIAN JIN, Carnegie Mellon University, USA ÁNGEL ALEXANDER CABRERA, Carnegie Mellon University, USA ADAM PERER, Carnegie Mellon University, USA HAIYI ZHU, Carnegie Mellon University, USA JASON I. HONG, Carnegie Mellon University, USA Ensuring effective public understanding of algorithmic decisions that are powered by machine learning techniques has become an urgent task with the increasing deployment of AI systems into our society. In this work, we present a concrete step toward this goal by redesigning confusion matrices for binary classification to support non-experts in understanding the performance of machine learning models. Through interviews (n=7) and a survey (n=102), we mapped out two major sets of challenges lay people have in understanding standard confusion matrices: the general terminologies and the matrix design. We further identified three sub-challenges regarding the matrix design, namely, confusion about the direction of reading the data, layered relations and quantities involved. We then conducted an online experiment with 483 participants to evaluate how effective a series of alternative representations target each of those challenges in the context of an algorithm for making recidivism predictions. We developed three levels of questions to evaluate users’ objective understanding. We assessed the effectiveness of our alternatives for accuracy in answering those questions, completion time, and subjective understanding. Our results suggest that (1) only by contextualizing terminologies can we significantly improve users’ understanding and (2) flow charts, which help point out the direction ofreading the data, were most useful in improving objective understanding.
    [Show full text]
  • A Generalized Hierarchical Approach for Data Labeling Zachary D. Blanks
    A Generalized Hierarchical Approach for Data Labeling by Zachary D. Blanks B.S. Operations Research, United States Air Force Academy Submitted to the Sloan School of Management in partial fulfillment of the requirements for the degree of Master of Science in Operations Research at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June, 2019 © Zachary D. Blanks, 2019. All rights reserved. The author hereby grants to MIT and DRAPER permission to reproduce and to distribute publicly paper and electronic copies of this thesis document in whole or in part in any medium now known or hereafter created. Author....................................................................................... Sloan School of Management May 17, 2019 Certifiedby.................................................................................. Dr. Troy M. Lau The Charles Stark Draper Laboratory Technical Supervisor Certifiedby.................................................................................. Prof. Rahul Mazumder Assistant Professor of Operations Research and Statistics Thesis Supervisor Acceptedby.................................................................................. Prof. Dimitris Bertsimas Boeing Professor of Operations Research Co-Director, Operations Research Center 1 THIS PAGE INTENTIONALLY LEFT BLANK 2 A Generalized Hierarchical Approach for Data Labeling by Zachary D. Blanks Submitted to the Sloan School of Management on May 17, 2019 in partial fulfillment of the requirements for the degree of Master of Science in Operations Research Abstract The goal of this thesis was to develop a data type agnostic classification algorithm best suited for problems where there are a large number of similar labels (e.g., classifying a port versus a shipyard). The most common approach to this issue is to simply ignore it, and attempt to fit a classifier against all targets at once (a “flat” classifier). The problem with this technique is that it tends to do poorly due to label similarity.
    [Show full text]
  • PREDICTING ACADEMIC PERFORMANCE of POTENTIAL ELECTRICAL ENGINEERING MAJORS a Thesis by SANJHANA SUNDARARAJ Submitted to the Offi
    PREDICTING ACADEMIC PERFORMANCE OF POTENTIAL ELECTRICAL ENGINEERING MAJORS A Thesis by SANJHANA SUNDARARAJ Submitted to the Office of Graduate and Professional Studies of Texas A&M University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Chair of Committee, Tie Liu Co-Chair, Xiaoning Qian Ulisses Braga Neto Xia Hu Head of Department, Miroslav Begovic August 2017 Major Subject: Electrical Engineering Copyright 2017 Sanjhana Sundararaj ABSTRACT Data Analytics for education is fast growing into an important part of higher learning institutions, which helps to improve student success rate and decision-making with regards to teaching methods, course selection, and student retention. The undergraduate program at Texas A&M University requires students to take up a general engineering program during their freshman and sophomore years. During the course of this period, student academic performance, abilities and participation is assessed. As per the Entry-to-a-Major policy, departments place the students in the best possible ma- jor based on their displayed capacities and in alignment with their goals. Our focus is on the Electrical Engineering department and the success rate of students with aspirations and background in this major. An approach to improve student retention rate is to predict beforehand the performance of students in specific course disciplines based on the infor- mation that is mined from their previous records. Based on the outcome, decisions can be made in advance regarding their further enrollment in the area and need for specific attention in certain aspects to get students up to the benchmark. In this thesis, we put together a set attributes related to students in the general pro- gram and with an electrical engineering aligned background.
    [Show full text]
  • Jets + Missing Energy Signatures at the Large Hadron Collider
    Jets + Missing Energy Signatures At The Large Hadron Collider DISSERTATION Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University By Khalida S. Hendricks, M.S. Graduate Program in Physics The Ohio State University 2019 Dissertation Committee: Linda Carpenter, Advisor Amy Connolly Annika Peter Antonio Boveia c Copyright by Khalida S. Hendricks 2019 Abstract In this work we consider new ways to use jets plus missing energy signatures in searches at the Large Hadron Collider. We study the Higgs boson (h) decay to two light jets at the 14 TeV High-Luminosity- LHC (HL-LHC), where a light jet (j) represents any non-flavor tagged jet from the obser- vational point of view. We estimate the achievable bounds on the decay product branching fractions through the associated production V h (V = W ±;Z). As a reasonable estimation, we only focus on the boosted region of high pT (h) and the three leptonic decay channels of the vector boson. We find that with 3000 fb−1 data at the HL-LHC, we should expect approximately 1σ statistical significance on the SM V h(gg) signal in this channel. This cor- responds to a reachable upper bound BR(h jj) 4 BRSM (h gg) at 95% confidence ! ≤ ! level. A consistency fit also leads to an upper bound of BR(h cc) < 15 BRSM (h cc) ! ! at 95% confidence level. The estimated bound may be further strengthened by adopting multiple variable analyses, or adding other production channels. We then consider some simple machine learning techniques applied to the same channels.
    [Show full text]
  • Multiple Classifier Systems Incorporating Uncertainty
    Christian Thiel Multiple Classifier Systems Incorporating Uncertainty The title page and cover show the painting doctors by Michael S. Parker (www.michaelsparker.com), 2003. Used with friendly permission of the artist. Cover layout by Kreativfrosch Grafik + Design (www.kreativfrosch.eu), Wegscheid bei Passau, Germany. Universit¨atUlm | 89069 Ulm | Germany Dissertation zur Erlangung des Doktorgrades Doktor der Naturwissenschaften (Dr. rer. nat.) der Fakult¨at f¨urIngenieurwissenschaften und Informatik der Universit¨at Ulm vorgelegt von Christian Andr´eThiel Angefertigt am Institut f¨urNeuroinformatik Einreichung der Dissertation: Dezember 2009 Tag der Promotion: 10. Juni 2010 Gutachter: Prof. Dr. G¨unther Palm Gutachter: Prof. Dr. Heiko Neumann Gutachter: Prof. Dr. Barbara Hammer Amtierender Dekan der Fakult¨at: Prof. Dr.-Ing. Michael Weber www.ChristianThiel.com Abstract The inclusion of uncertain class information into multi classifier systems (MCS) is the central theme in this thesis. A multi classifier system relies on multiple base classifiers, each of which is trained on a separate view of the problem at hand. Combining their answers will lead to a more accurate final decision. An example would be emotion recognition, with decisions based on observations of the mouth, the eyes or the pitch of the voice. Traditionally in classification one sample is associated with exactly one class, for example anger. But, in practical applications, such a hard distinction is not warranted; instead a sample should have soft class memberships, thus being associated fuzzily with multiple classes at the same time. The inclusion of this uncertain information into various, but isolated building blocks of a MCS has been tackled by a great many researchers.
    [Show full text]
  • Classifier Uncertainty: Evidence, Potential Impact, and Probabilistic Treatment
    Classifier uncertainty: evidence, potential impact, and probabilistic treatment Niklas Tötscha,1 and Daniel Hoffmanna aBioinformatics and Computational Biophysics, Universität Duisburg-Essen, 45141 Essen, Germany Classifiers are often tested on relatively small data sets, which CM entries cannot be neglected, which in turn makes all should lead to uncertain performance metrics. Nevertheless, these performance metrics derived from the CM uncertain, too. In metrics are usually taken at face value. We present an approach to the light of the ongoing replication crisis (1), it is plausible quantify the uncertainty of classification performance metrics, based that negligence of the metric uncertainty impedes reproducible on a probability model of the confusion matrix. Application of our ap- classification experiments. proach to classifiers from the scientific literature and a classification There is a lack of awareness of this problem, especially competition shows that uncertainties can be surprisingly large and outside the machine learning community. One often encoun- limit performance evaluation. In fact, some published classifiers are ters discussions of classifier performance lacking any statistical likely to be misleading. The application of our approach is simple analysis of the validity in the literature. If there is a statistical and requires only the confusion matrix. It is agnostic of the under- analysis it usually relies on frequentist methods such as confi- lying classifier. Our method can also be used for the estimation of dence intervals for the metrics or null hypothesis significance sample sizes that achieve a desired precision of a performance met- testing (NHST) to determine if a classifier is truly better than ric. random guessing. NHST “must be viewed as approximate, heuristic tests, rather than as rigorously correct statistical Classification Machine Learning Uncertainty Bayesian Modeling | | | | methods” (2).
    [Show full text]
  • Evaluating Prediction Strategies in an Enhanced Meta-Learning Framework
    EVALUATING PREDICTION STRATEGIES IN AN ENHANCED META-LEARNING FRAMEWORK Silviu Cacoveanu, Camelia Vidrighin and Rodica Potolea Technical University of Cluj-Napoca, Cluj-Napoca, Romania Keywords: Meta-learning Framework, Data Set Features, Performance Metrics, Prediction Strategies. Abstract: Finding the best learning strategy for a new domain/problem can prove to be an expensive and time- consuming process even for the experienced analysts. This paper presents several enhancements to a meta- learning framework we have previously designed and implemented. Its main goal is to automatically identify the most reliable learning schemes for a particular problem, based on the knowledge acquired about existing data sets, while minimizing the work done by the user but still offering flexibility. The main enhancements proposed here refer to the addition of several classifier performance metrics, including two original metrics, for widening the evaluation criteria, the addition of several new benchmark data sets for improving the outcome of the neighbor estimation step, and the integration of complex prediction strategies. Systematic evaluations have been performed to validate the new context of the framework. The analysis of the results revealed new research perspectives in the meta-learning area. 1 INTRODUCTION consuming since one has to decide which of the learning strategies is most suited given the context. Ever since the beginning of the information age, No definite way of discovering the best learning companies, organizations, universities all around the algorithm for a new problem has been devised yet, world have been taking advantage of technological but many proposals for selecting a good technique advances to store large amounts of data.
    [Show full text]