Regression Methods for Spatially Extending Traffic Data

Size: px
Start display at page:

Download "Regression Methods for Spatially Extending Traffic Data"

Transcription

1 Regression Methods for Spatially Extending Traffic Data Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli Università del Sannio ABSTRACT Traffic monitoring and network state estimation represent a key issue for ITS applications. Monitoring systems usually cover a limited number of network elements and, then, it is necessary to integrate the detected information about traffic flows and their characteristics using estimation models able to provide the state of whole network. Usually, this process involves the demand matrix estimation even though the knowledge of OD flows is often not necessary for several ITS application. This paper proposes an analysis of different types of regression models for spatially extending traffic data in function of the data detected on the monitored links; in particular, the paper focuses on all link flows estimation methods based on limited measures even if the same approach can be used to estimate other traffic related measures (i.e. mean speed, travel times, pollutant emission, etc.) Keywords (traffic estimation, regression models, ITS) INTRODUCTION An effective network state estimation is a crucial component for many ITS systems. Even though some ITS systems require the knowledge of OD flows and sometimes also their dynamic evolution, for many application the estimate of link characteristics as flows, speeds and travel times allows for effective ITS strategies as flow-responsive traffic-lights or route guidance based on instantaneous travel times. In the last decade road traffic monitoring technologies have had considerable evolution [1-5] and the wide diffusion of permanent monitoring devices (inductive loop sensors, cameras, radar, etc.) has provided a lot of traffic data in a time continuous manner. Nevertheless, because of budget limitation, the number of monitored links is usually very little in respect of networks dimension. For this reason network state estimation methods represent a crucial need for ITS applications and many authors have addressed this aspect. Unfortunately the number of available measures is usually not sufficient for the full observability of the transportation system and several studies are oriented to define observability condition. The network state estimation and the prevision of its evolution over time is mainly based on OD flows estimation; indeed, the knowledge of travel demand allows for calculating all link variables (flows, speed, travel time, pollutant emission, etc.) through traffic flow assignment methods. The OD matrix estimation based on traffic measures has been widely studied in the literature both in static and in dynamic context, but, unfortunately, OD flows updating is a highly undetermined problem [6] and several hypotheses have to be introduced in order to limit the number of variables [7] or to consider more relations between variables. This is possible mainly in dynamic context where some assumptions can be introduced about time evolution of the travel demand [8][9]. Considering that in many applications the knowledge of travel demand appears not necessary, herein we assess an alternative approach for network state estimation based only on link data, without involving OD flows estimation. In particular, they will be proposed and evaluated several regression methods, both parametric and non parametric, for estimating traffic flow data on some links of the network as a function of the data measured on other road links equipped with measurement devices. PROBLEM DESCRIPTION 5 Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli

2 This paper aims to assess some parametric and non parametric regression models to spatially extend flow information on monitored links for estimating the traffic flow on other links in the network. In more detail, let: L c be the set of monitored links and f c the vector of flows on these links with dimension n lc; L r be the set of non monitored relevant links and f r the vector of flows on these links with dimension n lr; d be the origin-destination flows vector with dimension n od; M be the assignment matrix mapping the link flows from OD flows with dimension (n l x n od); M c be the assignment sub-matrix referring to the links in L c with dimension (n lc x n od) M r be the assignment sub-matrix referring to the links in L r with dimension (n lr x n od) The assignment matrix M depends on the route choices of the traveler and usually is estimated through an assignment model considering the path costs formalized by the following equation: M = AP(C) (1.a) where: A represents the link-path matrix (number of links x number of paths) with elements a i,j equal to 1 if link i belongs to the path j, 0 otherwise; P is the route choice probability matrix (number of paths x number of OD flows) calculated as function of path costs C. Assuming the assignment matrix M known, error free, and non dependent on travel demand (non-congested network), the following equations can be written: f c = M c d (2.a) f r = M r d (2.b) If the rank of matrix M c is equal to the dimension of d, equation (1.a) allows for calculating the demand vector d and the flow vector f r can be expressed as a function of the vector f c through the following equation: 1 f r = M r M C f c (2.c) Unfortunately, very often the rank of M c is significantly lower than the demand vector dimension so that the demand and the consequent relevant flows estimation requires further information; the most common strategy to estimate the travel demand is based on prior knowledge of demand to be updated by using link flow measurements through Maximum Likelihood [10], Generalized Least Squares () [11] or Bayesian [12] approach. Referring to the approach, the estimation of the travel demand d* is given by following equation: d = d p + S M c t (M c S M c t ) 1 (f c M c d p ) where: d p is a prior estimation of d; S is the dispersion matrix of d p; M c t is the transpose of the matrix M c ; By using equations (1.b) and (2.b), flows f r can be estimated by equation (2.c): f r = M r d = M r [ d p + S M c t (M c S M c t ) 1 (f c M c d p )] Equation (3.b) establishes a linear relation between measured link flow, f c, and relevant but non-measured link flows, f r, and it can be viewed as flow data spatial extension model based on formulation. Notably, equation (3.b) requires the availability of prior estimation of the travel demand and also the knowledge of its distribution and dispersion matrix. In the following, several regression models aimed to estimate the flow vector f r directly from vector f c will be tested in laboratory experiments and compared to results provided by equation (3.b), here used as benchmark. (3.a) (3.b) 6 Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli

3 In the case of congested networks, link travel costs depend on link flows and, then, the assignment model should be formalized as a fixed point problem: f = A P(A t c(f )) d (4) Equation (4) establishes a non-linear relationship between travel demand and link flows that does not allow a closed expression for the inverse problem, namely the estimation of travel demand d from link flows. Consistently with the model, demand estimation can be obtained by the solution of following optimization problem: d = argmin [(d d p ) S (d d p ) t ] (5.a) subject to: f = A P(A t c(f )) d f c = f c where f c is the vector of assigned link flows on monitored links extracted from all assigned link flows vector f. The spatial extension of f c data can be obtained by assigning the estimated travel demand d^ and considering the set of non-monitored relevant links. It is worth noting that for large networks the solution of problem (5.a)-(5.c) requires computational efforts often not compatible with real time application; for this reason, synthetic models, based on regression and preprocessed data, appear more suitable in this context. Also in the case of congested networks, several regression models will be tested in order to obtain an effective spatial extension of monitored data and the results will be compared to the formulation. (5.b) (5.c) REGRESSION MODELS FOR EXTENSION OF LINK FLOW DATA The aim of this paper is to analyze different regression methods which are able to define a relationship between the measured traffic flows and those predicted. Considering that equation (3.b), based on a formulation, provides a linear relationship between measured and predicted flow variables, the linear regression, as the simplest method, has been firstly tested. Furthermore, this paper describes the nonparametric formulations, which are models suitable to describe virtually any nonlinear relationship. In this paper we propose and test three different methods: Kernel Methods, Artificial Neural Networks (s) and Methods. A theoretical reference on the use of kernel methods is provided by [13][14] and [15], while some examples of applications of Neural Network in transportation engineering are reported in [16] and [17], while for methods the principal past studies are reported in [17] and [18]. Linear regression Regression analysis is a technique used to analyze a series of data, using a dependent variable and one or more independent variables. The aim is to estimate a possible functional relationship between the dependent variable and the independent variables. Linear regression requires a linear model. Y = b 0 + b 1 X 1 + b 2 X b nc X nc (6) where: b 0, b 1,..., b nc are the model parameters; Y is the predicted dependent variable (non measured flow in our case); X 1, X 2,..., X nc are independent variables (measured flows in our case). Consistently with minimum square error paradigm, for each relevant link flow Y r, the vector of parameters, b, is given by: b = (X t X) 1 X t Y r (7) where: 7 Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli

4 X is the independent observation matrix, with dimensions N x (n c+1), where N represents the number of observations; Y r is the vector of the observed dependent variable (i.e. the link flow of link r) with dimension N x 1; Notably, in the case of full observability (rank of M c equal to n od), equation (7) must provide the same results of equation (2.a) if the number of observations is at least equal to n c+1. Kernel regression The objective of kernel regression is to find a non-linear relation between random variables X and Y. In any nonparametric regression, the conditional expectation of a variable Y relative to a variable X may be written: E(Y X) = m(x) where m is an unknown function. Nadaraya and Watson, both in 1964, proposed to estimate m as a locally weighted average, using a kernel, as a weighting function. The Nadaraya-Watson estimator is: n m (x) = i=1 K h x (x x i )y i n = W i=1 K hx (x x i ) hx (x, x i )y i i=1 A kernel is a non-negative real-value function K mapping from X X to R satisfying the following two requirements: + K(u)du = 1; K( u) = K(u) for all values of u; with u = x x i and where the bandwidth h represents a smoothing parameter limiting the size of interval h where the values should be considered in a weighted average. The first requirement ensures that the method of kernel density estimation results in a probability density function. The second requirement ensures that the average of the corresponding distribution is equal to that of the sample used. Several types of kernel functions are commonly used: uniform, triangle, Epanechnikov, quartic (biweight), tricube, triweight, Gaussian, quadratic and cosine. Herein will be considered the Epanechnikov kernel, named K 1 in the following: K(u) = 3 4 (1 u2 )1 { u 1} and the Gaussian formulation, named K 2 in the following: K(u) = 1 2π e Neural Networks Artificial neural networks (s) are a computational model widely used in computer science and other research disciplines, which is based on a large collection of simple neural units (artificial neurons), loosely analogous to the observed behavior of a biological brain s neurons. A single artificial neuron can be implemented in many different ways. The general mathematic definition is: Where x is a neuron with n input dendrites (x 0,, x n ), that are the elements which connect the nucleus to other nucleuses, and one output neuron y(x) and (w 0,, w n ) are weights calibrated in the training procedure. Parameter g is an activation function that weights how powerful the output (if any) should be from the neuron, based on the sum of the input. If the artificial neuron should mimic a real neuron, the activation function g should be a simple threshold function returning 0 or 1. In this paper we have chosen to implement a multilayer feed-forward, which is the most common kind of. In a multilayer feed-forward, the neurons 8 Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli n n 1 2u 2 y(x) = g ( w i x i ) i=0

5 are ordered in layers, starting with an input layer and ending with an output layer. Between these two layers are a number of hidden layers. Connections in these kinds of network only go forward from one layer to the next. Implementing multilayer feed-forward s requires two different phases: a training phase (sometimes also referred to as the learning phase) and an execution phase. In the training phase the is trained to return a specific output when given a specific input, this is done by continuous training on a set of training data. In the execution phase the returns outputs on the basis of inputs. Method Support Vector Machine () analysis is a popular machine learning tool for classification and regression. methods were originally designed for nonlinear classification problems, then extended to nonlinear regression problems (SVR). Suppose we have the training data {(x 1, y 1 ),. (x n, y n )} with n patterns; in the first step we map the input pattern x into a higher dimensional feature space using a nonlinear mapping function φ. The nonlinear regression problem between x and y becomes a linear regression problem between φ(x) and y, i.e.: f(x; w) = w, φ(x) + b Where. symbolizes the inner product and b are the regression coefficients obtained minimizing the error between f and the observed values (y). The error between f and y is not evaluated by MSE norm (mean squared error), but using the ε intensive error norm: 0 if f(x; w) y < ε f(x; w) y ε = { f(x; w) y ε otherwise where small errors f(x; w) y < ε are ignored, but for large errors the above mentioned error norm approximates the mean absolute error (MAE). The regression coefficients (w and b) are estimated by minimizing the objective function: J = C n f(x i; w) y i ε w 2 i=1 where C, that controls the regularization, and ε are the hyper-parameters. n NUMERICAL EXPERIMENTS The laboratory experiments have been carried out considering a network with 16 nodes, 48 oriented links and 16 OD pairs. In the following the flows of 6 links will be assumed knows (monitored links) and the 42 link flows will represent the variables to be estimated. Starting from a prior seed matrix d p, several draws d k of OD matrix have been generated through random sampling with different levels of coefficient of variation (CV). In more detail, for each CV level, the training sets were created by assigning 200 origin destination matrices, randomly generated, to the network, so that the training dataset where composed by 200 examples with 6 independent variables (measured link flows) and the corresponding 42 dependent variables (non measured relevant link flows). Similarly the validation set was generated by the assignment of 100 origin destination matrices in order to have 100 pairs of independent variables and dependent target variables. In this first set of experiments the network has been considered uncongested and the assignment matrix M was calculated through a logit based route choice model according to assumption of link travel times independent on link flows. In Table 1 are summarized the results with reference to validation set for different models in terms of Mean Absolute Percentage Error () and Root Mean Square Error (RMSE) for overall link flows estimation. Table 1. Mean errors in uncongested network of predicted link flows for different coefficients of variation in validation dataset 9 Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli

6 CV 0,1 2,41% 2,45% 3,45% 3,51% 2,58% 2,46% 0,2 5,01% 5,06% 6,47% 6,49% 5,33% 5,09% 0,3 7,47% 7,65% 10,21% 10,24% 7,96% 7,67% 0,4 11,67% 11,96% 15,55% 15,61% 12,27% 11,99% 0,5 15,56% 16,13% 22,33% 22,41% 16,59% 16,13% 0,6 17,88% 18,04% 26,90% 27,01% 19,09% 18,04% 0,7 25,10% 26,33% 33,84% 33,92% 27,81% 25,95% 0,8 31,30% 32,21% 39,37% 39,37% 32,65% 32,35% 0,9 38,77% 37,92% 56,32% 56,66% 39,24% 37,02% 1 56,23% 58,85% 77,95% 78,02% 60,51% 57,79% RMSE CV 0,1 3,81 3,90 6,20 6,34 4,29 3,90 0,2 7,81 7,92 11,47 11,53 0,00 0,00 0,3 11,20 11,42 17,25 17,25 12,43 11,47 0,4 15,47 15,77 22,18 22,28 16,50 15,83 0,5 18,95 19,62 29,27 29,34 20,46 19,71 0,6 21,27 21,59 33,03 33,00 22,39 21,83 0,7 25,24 25,69 38,07 38,04 27,34 25,88 0,8 28,16 28,66 41,83 41,71 0,00 0,00 0,9 30,72 30,98 46,65 46,57 31,78 31, ,51 32,68 47,98 47,80 0,61 32,96 The results show that for uncongested network a classical estimation perform well till high value of the dispersion of the true OD matrix around prior OD matrix estimation d p; notably, the linear regression based on simulated data not considering explicitly prior estimation and its dispersion provides results very similar to estimation and it can be considered as an alternative approach, requiring pre-calculated dataset, to improve the computational effort for large network in real-time application. In this context, also the and the model appear suitable to be used for link flows estimation because, without making any assumption about the relationship between measured flows and predicted flows, they lead to errors not significantly different from linear models. The worst performances are given by Kernel regression models, where the error does not change significantly with the choice of kernel function and are always higher than errors provided by the other models. In Figure 1 the cumulative distributions of the of predicted link flows are reported. The results highlight that for coefficient of variation lower than 0.5 over 70% of predicted link flows shows a percentage error lower than 20% in all models, except for Kernel model where the error is nearly 25%, while this percentage quickly decreases for higher coefficient of variation leading to an unacceptable error for a significant number of links. 10 Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli

7 Cumulative Distribution - CV=0.2 Cumulative Distribution - CV=0.3 Cumulative Distribution - CV=0.4 Cumulative Distribution - CV=0.5 Cumulative Distribution - CV=0.6 Cumulative Distribution - CV=0.7 Cumulative Distribution - CV=0.8 Cumulative Distribution - CV=0.9 Fig 1: Cumulative distribution of in uncongested network among predicted link flows Similar experiments have been carried out considering congested network. In particular a BPR cost function has been considered in order to take into account the dependency of link travel time on link flows. Considering the non linear relationship between measured and predicted flows, the linear regression has not been considered in this case. As shown in Table 2 and Figure 2, the results are generally worse with respect to the uncongested network, consistently with more complex relationship involving problem variables. It is 11 Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli

8 worth noting that, both in the case of uncongested and congested network, the model and model perform very close and sometimes better than model; this aspect suggests that non-parametric regression models appear suitable for application mainly in real time application where the computation time required by formulation represents a critical issue. Therefore, the performances of as well as of non-parametric models significantly decrease as the dispersion in dataset increases. To overcome this problem the opportunity to couple a non parametric regression model with cluster preprocessing analysis has been addressed. In more detail, with reference to scenarios having coefficient of variation from 0.3 to 0.9, the training dataset has been divided into several clusters on the basis of values of independent variables (measured flows on six links). To define the clusters, a classical method based on Euclidean distance has been used and for each cluster an and a model have been trained. The validation has been carried out assigning each pair of measured and relevant link flows to the closest cluster and applying the corresponding trained models. The coming from the combination of cluster analysis and non parametric regression model with reference to overall relevant links are reported in Table 3. These results suggest that the preprocessing of data based on cluster analysis allow for splitting data in several demand scenarios leading to significantly better final estimation, particularly for model that seems performing better than approach. Table 2. Mean errors in congested network of predicted link flows for different coefficients of variation in validation dataset CV 0,1 3,52% 3,71% 3,71% 3,90% 3,72% 0,2 7,90% 7,85% 7,85% 8,08% 8,26% 0,3 12,29% 10,08% 10,10% 12,23% 13,14% 0,4 19,41% 14,96% 14,97% 18,41% 18,13% 0,5 21,71% 20,04% 20,09% 23,50% 22,72% 0,6 24,99% 24,73% 24,79% 29,70% 27,02% 0,7 38,29% 32,99% 33,03% 45,63% 44,07% 0,8 39,39% 36,54% 36,43% 47,96% 42,17% 0,9 69,82% 53,14% 53,24% 70,44% 67,12% 1 68,10% 62,97% 63,58% 77,74% 80,12% RMSE CV 0,1 3,81 6,20 6,34 7,51 6,52 0,2 13,48 13,63 13,63 13,79 13,90 0,3 20,12 19,45 19,50 23,42 21,40 0,4 28,78 25,74 25,76 32,77 31,64 0,5 34,08 32,69 32,64 40,05 36,45 0,6 38,38 36,77 36,78 45,72 42,30 0,7 44,11 44,25 44,38 56,52 50,34 0,8 45,27 46,49 46,38 60,04 51,52 0,9 54,93 51,22 51,19 61,69 56, ,69 54,70 54,72 65,53 62,16 12 Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli

9 International Journal of Engineering Technology, Management and Applied Sciences Cumulative Distribution - CV=0.2 Cumulative Distribution - CV=0.3 Cumulative Distribution - CV=0.4 Cumulative Distribution - CV=0.5 Cumulative Distribution - CV=0.6 Cumulative Distribution - CV=0.7 Cumulative Distribution - CV=0.8 Cumulative Distribution - CV=0.9 Fig 2: Cumulative distribution of in congested network among predicted link flows Table 3. in congested network of predicted link flows with combined cluster analysis and non parametric regressions. CV 0,4 10,50% 8,90% 0,5 14,45% 11,05% 0,6 17,29% 15,03% 0,7 23,45% 18,90% 0,8 29,60% 21,82% 0,9 39,83% 34,06% 13 Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli

10 CONCLUSIONS In this paper several regression models for spatial extension of traffic data has been evaluated and compared to traditional approach based on travel demand estimation. The results shows that, even though all approaches do not appear reliable when the traffic scenario get away from main pattern, the and model are suitable to be applied, mainly in real time context where the computational time required by the optimization approach represents a critical aspect. Furthermore, the combination of cluster analysis aimed to recognize different demand scenarios and non parametric regression, mainly the regression, is able to guarantee the better results. REFERENCES [1] C.-H. Chong, S.P. Kumar, Sensor networks: evolution, opportunities, and challenges, Proceedings of the IEEE, vol. 91, pp August 2003 [2] Z. Sun, G. Bebis, R. Miller, On-road vehicle detection using optical sensors: a review, 2004 IEEE Intelligent Transportation Systems, Conference, Washington, D.C., USA, pp , October 2004 [3] A. Sharma, R. Chaki, U. Bhattacharya, Applications of wireless sensor network in Intelligent Traffic System: a review, 3rd InternationalConference on Electronics Computer Technology (ICECT), pp , April 2011 [4] G.S. Tewolde, Sensor and network technology for Intelligent Transportation Systems, 2012 IEEE International Conference on Electro/Information Technology (EIT), pp. 1-7, May 2012 [5] B. Tian, B.T. Morris, M. Tang, Y. Yao, C. Gou, D. Shen, S. Tang Hierarchical and networked vehicle surveillance in ITS: a survey, IEEE Transactions on Intelligent Transportation Systems, vol. 16, pp , April 2015 [6] Marzano, V., Papola, A., Simonelli, F., Limits and perspectives of effective o-d matrix correction using traffic counts. Transportation Research Part C, 17(5), [7] Djukic, T., Flötteröd G., van Lint H., Hoogendoorn S. (2012). Efficient real time OD matrix estimation based on Principal Component Analysis. Proceedings of Intelligent Transportation Systems (ITSC), th International IEEE Conference ITS, doi /ITSC [8] Ashok, K., Ben-Akiva, M., Alternative approaches for real-time estimation and prediction of time-dependent origin-destination flows. Transportation Science 34(1), [9] Cascetta E, Papola A., Marzano V., Simonelli F., Vitiello I., Quasi-dynamic estimation of o-d flows from traffic counts: formulation, statistical validation and performance analysis on real data. Transportation research Part B, vol. 55, pp [10] M.G.H. Bell, The estimation of origin destination matrix from traffic counts, Transportation Science, 17 (2) (1983), pp [11] E. Cascetta, Estimation of trip matrices from traffic counts and survey data: a generalized least squares estimator, Transportation Research Part B, 18 (4/5) (1984), pp [12] M. Maher, Inferences on trip matrices from observations on link volumes: a Bayesian statistical approach Transportation Research Part B, 17 (6) (1983), pp [13] Blundell, R., and A. Duncan. Kernel Regressions in Empirical Microeconomics Blundell, R., and A. Duncan. Kernel Regressions in Empirical Microeconomics. Journal of Human Resources, Vol. 33, No. 1, 1995, pp [14] Nadaraya, E. A. On Estimating Regression. Theory of Probability and Its Applications, Vol. 9, No. 1, 1964, pp [15] Watson, G. S. Smooth Regression Analysis. Shankya Series A, Vol. 26 No. 4, 1964, pp [16] M. Gallo, F. Simonelli, G. De Luca, C. Della Porta, An Artificial Neural Netwprk approach for spatially extending road traffic monitoring measures, Proceedings of 2016 IEEE Workshop on Environmental, Energy, and Structural Monitoring Systems (EESMS 2016), Bari, Italy, June 2016 [17] G. De Luca; M. Gallo Artificial neural networks for forecasting user flows in transportation networks: Literature review, limits, potentialities and open challenges, th IEEE International Conference on Models and Technologies for Intelligent Transportation Systems (MT-ITS) [18] M. Sciandrone Support Vector Machines, [19] H.T. Lin, C.J. Lin A study on Sigmoid Kernels for and the Training of non-psd Kernels by SMO-type Methods Department of Computer Science and Information Engineering National Taiwan University, Roberto Iannella, Mariano Gallo, Giuseppina de Luca, Fulvio Simonelli

We can now formulate our model as showed in equation 2:

We can now formulate our model as showed in equation 2: Simulation of traffic conditions requires accurate knowledge of travel demand. In a dynamic context, this entails estimating time-dependent demand matrices, which are a discretised representation of the

More information

Improving the travel time prediction by using the real-time floating car data

Improving the travel time prediction by using the real-time floating car data Improving the travel time prediction by using the real-time floating car data Krzysztof Dembczyński Przemys law Gawe l Andrzej Jaszkiewicz Wojciech Kot lowski Adam Szarecki Institute of Computing Science,

More information

AN INTRODUCTION TO NEURAL NETWORKS. Scott Kuindersma November 12, 2009

AN INTRODUCTION TO NEURAL NETWORKS. Scott Kuindersma November 12, 2009 AN INTRODUCTION TO NEURAL NETWORKS Scott Kuindersma November 12, 2009 SUPERVISED LEARNING We are given some training data: We must learn a function If y is discrete, we call it classification If it is

More information

18.6 Regression and Classification with Linear Models

18.6 Regression and Classification with Linear Models 18.6 Regression and Classification with Linear Models 352 The hypothesis space of linear functions of continuous-valued inputs has been used for hundreds of years A univariate linear function (a straight

More information

Data Mining Part 5. Prediction

Data Mining Part 5. Prediction Data Mining Part 5. Prediction 5.5. Spring 2010 Instructor: Dr. Masoud Yaghini Outline How the Brain Works Artificial Neural Networks Simple Computing Elements Feed-Forward Networks Perceptrons (Single-layer,

More information

Neural Networks Lecture 4: Radial Bases Function Networks

Neural Networks Lecture 4: Radial Bases Function Networks Neural Networks Lecture 4: Radial Bases Function Networks H.A Talebi Farzaneh Abdollahi Department of Electrical Engineering Amirkabir University of Technology Winter 2011. A. Talebi, Farzaneh Abdollahi

More information

Bidirectional Representation and Backpropagation Learning

Bidirectional Representation and Backpropagation Learning Int'l Conf on Advances in Big Data Analytics ABDA'6 3 Bidirectional Representation and Bacpropagation Learning Olaoluwa Adigun and Bart Koso Department of Electrical Engineering Signal and Image Processing

More information

TRAFFIC FLOW MODELING AND FORECASTING THROUGH VECTOR AUTOREGRESSIVE AND DYNAMIC SPACE TIME MODELS

TRAFFIC FLOW MODELING AND FORECASTING THROUGH VECTOR AUTOREGRESSIVE AND DYNAMIC SPACE TIME MODELS TRAFFIC FLOW MODELING AND FORECASTING THROUGH VECTOR AUTOREGRESSIVE AND DYNAMIC SPACE TIME MODELS Kamarianakis Ioannis*, Prastacos Poulicos Foundation for Research and Technology, Institute of Applied

More information

Artificial Neural Networks. Edward Gatt

Artificial Neural Networks. Edward Gatt Artificial Neural Networks Edward Gatt What are Neural Networks? Models of the brain and nervous system Highly parallel Process information much more like the brain than a serial computer Learning Very

More information

Artificial Neural Network

Artificial Neural Network Artificial Neural Network Contents 2 What is ANN? Biological Neuron Structure of Neuron Types of Neuron Models of Neuron Analogy with human NN Perceptron OCR Multilayer Neural Network Back propagation

More information

Pattern Recognition and Machine Learning. Perceptrons and Support Vector machines

Pattern Recognition and Machine Learning. Perceptrons and Support Vector machines Pattern Recognition and Machine Learning James L. Crowley ENSIMAG 3 - MMIS Fall Semester 2016 Lessons 6 10 Jan 2017 Outline Perceptrons and Support Vector machines Notation... 2 Perceptrons... 3 History...3

More information

NONLINEAR CLASSIFICATION AND REGRESSION. J. Elder CSE 4404/5327 Introduction to Machine Learning and Pattern Recognition

NONLINEAR CLASSIFICATION AND REGRESSION. J. Elder CSE 4404/5327 Introduction to Machine Learning and Pattern Recognition NONLINEAR CLASSIFICATION AND REGRESSION Nonlinear Classification and Regression: Outline 2 Multi-Layer Perceptrons The Back-Propagation Learning Algorithm Generalized Linear Models Radial Basis Function

More information

A GENERAL FORMULATION FOR SUPPORT VECTOR MACHINES. Wei Chu, S. Sathiya Keerthi, Chong Jin Ong

A GENERAL FORMULATION FOR SUPPORT VECTOR MACHINES. Wei Chu, S. Sathiya Keerthi, Chong Jin Ong A GENERAL FORMULATION FOR SUPPORT VECTOR MACHINES Wei Chu, S. Sathiya Keerthi, Chong Jin Ong Control Division, Department of Mechanical Engineering, National University of Singapore 0 Kent Ridge Crescent,

More information

Pattern Recognition Prof. P. S. Sastry Department of Electronics and Communication Engineering Indian Institute of Science, Bangalore

Pattern Recognition Prof. P. S. Sastry Department of Electronics and Communication Engineering Indian Institute of Science, Bangalore Pattern Recognition Prof. P. S. Sastry Department of Electronics and Communication Engineering Indian Institute of Science, Bangalore Lecture - 27 Multilayer Feedforward Neural networks with Sigmoidal

More information

Research Article Accurate Multisteps Traffic Flow Prediction Based on SVM

Research Article Accurate Multisteps Traffic Flow Prediction Based on SVM Mathematical Problems in Engineering Volume 2013, Article ID 418303, 8 pages http://dx.doi.org/10.1155/2013/418303 Research Article Accurate Multisteps Traffic Flow Prediction Based on SVM Zhang Mingheng,

More information

ARTIFICIAL NEURAL NETWORK PART I HANIEH BORHANAZAD

ARTIFICIAL NEURAL NETWORK PART I HANIEH BORHANAZAD ARTIFICIAL NEURAL NETWORK PART I HANIEH BORHANAZAD WHAT IS A NEURAL NETWORK? The simplest definition of a neural network, more properly referred to as an 'artificial' neural network (ANN), is provided

More information

Artificial Neural Networks (ANN) Xiaogang Su, Ph.D. Department of Mathematical Science University of Texas at El Paso

Artificial Neural Networks (ANN) Xiaogang Su, Ph.D. Department of Mathematical Science University of Texas at El Paso Artificial Neural Networks (ANN) Xiaogang Su, Ph.D. Department of Mathematical Science University of Texas at El Paso xsu@utep.edu Fall, 2018 Outline Introduction A Brief History ANN Architecture Terminology

More information

A Support Vector Regression Model for Forecasting Rainfall

A Support Vector Regression Model for Forecasting Rainfall A Support Vector Regression for Forecasting Nasimul Hasan 1, Nayan Chandra Nath 1, Risul Islam Rasel 2 Department of Computer Science and Engineering, International Islamic University Chittagong, Bangladesh

More information

Support Vector Regression (SVR) Descriptions of SVR in this discussion follow that in Refs. (2, 6, 7, 8, 9). The literature

Support Vector Regression (SVR) Descriptions of SVR in this discussion follow that in Refs. (2, 6, 7, 8, 9). The literature Support Vector Regression (SVR) Descriptions of SVR in this discussion follow that in Refs. (2, 6, 7, 8, 9). The literature suggests the design variables should be normalized to a range of [-1,1] or [0,1].

More information

Artificial Neural Networks

Artificial Neural Networks Introduction ANN in Action Final Observations Application: Poverty Detection Artificial Neural Networks Alvaro J. Riascos Villegas University of los Andes and Quantil July 6 2018 Artificial Neural Networks

More information

Artificial Intelligence

Artificial Intelligence Artificial Intelligence Jeff Clune Assistant Professor Evolving Artificial Intelligence Laboratory Announcements Be making progress on your projects! Three Types of Learning Unsupervised Supervised Reinforcement

More information

Neural Networks: Introduction

Neural Networks: Introduction Neural Networks: Introduction Machine Learning Fall 2017 Based on slides and material from Geoffrey Hinton, Richard Socher, Dan Roth, Yoav Goldberg, Shai Shalev-Shwartz and Shai Ben-David, and others 1

More information

Lecture 4: Feed Forward Neural Networks

Lecture 4: Feed Forward Neural Networks Lecture 4: Feed Forward Neural Networks Dr. Roman V Belavkin Middlesex University BIS4435 Biological neurons and the brain A Model of A Single Neuron Neurons as data-driven models Neural Networks Training

More information

Nearest Neighbor. Machine Learning CSE546 Kevin Jamieson University of Washington. October 26, Kevin Jamieson 2

Nearest Neighbor. Machine Learning CSE546 Kevin Jamieson University of Washington. October 26, Kevin Jamieson 2 Nearest Neighbor Machine Learning CSE546 Kevin Jamieson University of Washington October 26, 2017 2017 Kevin Jamieson 2 Some data, Bayes Classifier Training data: True label: +1 True label: -1 Optimal

More information

EEE 241: Linear Systems

EEE 241: Linear Systems EEE 4: Linear Systems Summary # 3: Introduction to artificial neural networks DISTRIBUTED REPRESENTATION An ANN consists of simple processing units communicating with each other. The basic elements of

More information

22c145-Fall 01: Neural Networks. Neural Networks. Readings: Chapter 19 of Russell & Norvig. Cesare Tinelli 1

22c145-Fall 01: Neural Networks. Neural Networks. Readings: Chapter 19 of Russell & Norvig. Cesare Tinelli 1 Neural Networks Readings: Chapter 19 of Russell & Norvig. Cesare Tinelli 1 Brains as Computational Devices Brains advantages with respect to digital computers: Massively parallel Fault-tolerant Reliable

More information

Simulation on a partitioned urban network: an approach based on a network fundamental diagram

Simulation on a partitioned urban network: an approach based on a network fundamental diagram The Sustainable City IX, Vol. 2 957 Simulation on a partitioned urban network: an approach based on a network fundamental diagram A. Briganti, G. Musolino & A. Vitetta DIIES Dipartimento di ingegneria

More information

Real-Time Travel Time Prediction Using Multi-level k-nearest Neighbor Algorithm and Data Fusion Method

Real-Time Travel Time Prediction Using Multi-level k-nearest Neighbor Algorithm and Data Fusion Method 1861 Real-Time Travel Time Prediction Using Multi-level k-nearest Neighbor Algorithm and Data Fusion Method Sehyun Tak 1, Sunghoon Kim 2, Kiate Jang 3 and Hwasoo Yeo 4 1 Smart Transportation System Laboratory,

More information

Course 10. Kernel methods. Classical and deep neural networks.

Course 10. Kernel methods. Classical and deep neural networks. Course 10 Kernel methods. Classical and deep neural networks. Kernel methods in similarity-based learning Following (Ionescu, 2018) The Vector Space Model ò The representation of a set of objects as vectors

More information

Artificial Neural Networks" and Nonparametric Methods" CMPSCI 383 Nov 17, 2011!

Artificial Neural Networks and Nonparametric Methods CMPSCI 383 Nov 17, 2011! Artificial Neural Networks" and Nonparametric Methods" CMPSCI 383 Nov 17, 2011! 1 Todayʼs lecture" How the brain works (!)! Artificial neural networks! Perceptrons! Multilayer feed-forward networks! Error

More information

Artificial Neural Networks Examination, June 2005

Artificial Neural Networks Examination, June 2005 Artificial Neural Networks Examination, June 2005 Instructions There are SIXTY questions. (The pass mark is 30 out of 60). For each question, please select a maximum of ONE of the given answers (either

More information

From CDF to PDF A Density Estimation Method for High Dimensional Data

From CDF to PDF A Density Estimation Method for High Dimensional Data From CDF to PDF A Density Estimation Method for High Dimensional Data Shengdong Zhang Simon Fraser University sza75@sfu.ca arxiv:1804.05316v1 [stat.ml] 15 Apr 2018 April 17, 2018 1 Introduction Probability

More information

MINIMUM EXPECTED RISK PROBABILITY ESTIMATES FOR NONPARAMETRIC NEIGHBORHOOD CLASSIFIERS. Maya Gupta, Luca Cazzanti, and Santosh Srivastava

MINIMUM EXPECTED RISK PROBABILITY ESTIMATES FOR NONPARAMETRIC NEIGHBORHOOD CLASSIFIERS. Maya Gupta, Luca Cazzanti, and Santosh Srivastava MINIMUM EXPECTED RISK PROBABILITY ESTIMATES FOR NONPARAMETRIC NEIGHBORHOOD CLASSIFIERS Maya Gupta, Luca Cazzanti, and Santosh Srivastava University of Washington Dept. of Electrical Engineering Seattle,

More information

ECE521 Lectures 9 Fully Connected Neural Networks

ECE521 Lectures 9 Fully Connected Neural Networks ECE521 Lectures 9 Fully Connected Neural Networks Outline Multi-class classification Learning multi-layer neural networks 2 Measuring distance in probability space We learnt that the squared L2 distance

More information

Econ 582 Nonparametric Regression

Econ 582 Nonparametric Regression Econ 582 Nonparametric Regression Eric Zivot May 28, 2013 Nonparametric Regression Sofarwehaveonlyconsideredlinearregressionmodels = x 0 β + [ x ]=0 [ x = x] =x 0 β = [ x = x] [ x = x] x = β The assume

More information

Computational statistics

Computational statistics Computational statistics Lecture 3: Neural networks Thierry Denœux 5 March, 2016 Neural networks A class of learning methods that was developed separately in different fields statistics and artificial

More information

Machine Learning Linear Classification. Prof. Matteo Matteucci

Machine Learning Linear Classification. Prof. Matteo Matteucci Machine Learning Linear Classification Prof. Matteo Matteucci Recall from the first lecture 2 X R p Regression Y R Continuous Output X R p Y {Ω 0, Ω 1,, Ω K } Classification Discrete Output X R p Y (X)

More information

Announcements. Proposals graded

Announcements. Proposals graded Announcements Proposals graded Kevin Jamieson 2018 1 Bayesian Methods Machine Learning CSE546 Kevin Jamieson University of Washington November 1, 2018 2018 Kevin Jamieson 2 MLE Recap - coin flips Data:

More information

CS:4420 Artificial Intelligence

CS:4420 Artificial Intelligence CS:4420 Artificial Intelligence Spring 2018 Neural Networks Cesare Tinelli The University of Iowa Copyright 2004 18, Cesare Tinelli and Stuart Russell a a These notes were originally developed by Stuart

More information

Neural Networks Lecturer: J. Matas Authors: J. Matas, B. Flach, O. Drbohlav

Neural Networks Lecturer: J. Matas Authors: J. Matas, B. Flach, O. Drbohlav Neural Networks 30.11.2015 Lecturer: J. Matas Authors: J. Matas, B. Flach, O. Drbohlav 1 Talk Outline Perceptron Combining neurons to a network Neural network, processing input to an output Learning Cost

More information

Introduction to Natural Computation. Lecture 9. Multilayer Perceptrons and Backpropagation. Peter Lewis

Introduction to Natural Computation. Lecture 9. Multilayer Perceptrons and Backpropagation. Peter Lewis Introduction to Natural Computation Lecture 9 Multilayer Perceptrons and Backpropagation Peter Lewis 1 / 25 Overview of the Lecture Why multilayer perceptrons? Some applications of multilayer perceptrons.

More information

ANN and Statistical Theory Based Forecasting and Analysis of Power System Variables

ANN and Statistical Theory Based Forecasting and Analysis of Power System Variables ANN and Statistical Theory Based Forecasting and Analysis of Power System Variables Sruthi V. Nair 1, Poonam Kothari 2, Kushal Lodha 3 1,2,3 Lecturer, G. H. Raisoni Institute of Engineering & Technology,

More information

A Reservoir Sampling Algorithm with Adaptive Estimation of Conditional Expectation

A Reservoir Sampling Algorithm with Adaptive Estimation of Conditional Expectation A Reservoir Sampling Algorithm with Adaptive Estimation of Conditional Expectation Vu Malbasa and Slobodan Vucetic Abstract Resource-constrained data mining introduces many constraints when learning from

More information

Improving forecasting under missing data on sparse spatial networks

Improving forecasting under missing data on sparse spatial networks Improving forecasting under missing data on sparse spatial networks J. Haworth 1, T. Cheng 1, E. J. Manley 1 1 SpaceTimeLab, Department of Civil, Environmental and Geomatic Engineering, University College

More information

Neural Networks. Chapter 18, Section 7. TB Artificial Intelligence. Slides from AIMA 1/ 21

Neural Networks. Chapter 18, Section 7. TB Artificial Intelligence. Slides from AIMA   1/ 21 Neural Networks Chapter 8, Section 7 TB Artificial Intelligence Slides from AIMA http://aima.cs.berkeley.edu / 2 Outline Brains Neural networks Perceptrons Multilayer perceptrons Applications of neural

More information

Feedforward Neural Nets and Backpropagation

Feedforward Neural Nets and Backpropagation Feedforward Neural Nets and Backpropagation Julie Nutini University of British Columbia MLRG September 28 th, 2016 1 / 23 Supervised Learning Roadmap Supervised Learning: Assume that we are given the features

More information

Portugaliae Electrochimica Acta 26/4 (2008)

Portugaliae Electrochimica Acta 26/4 (2008) Portugaliae Electrochimica Acta 6/4 (008) 6-68 PORTUGALIAE ELECTROCHIMICA ACTA Comparison of Regression Model and Artificial Neural Network Model for the Prediction of Volume Percent of Diamond Deposition

More information

Gaussian Processes. Le Song. Machine Learning II: Advanced Topics CSE 8803ML, Spring 2012

Gaussian Processes. Le Song. Machine Learning II: Advanced Topics CSE 8803ML, Spring 2012 Gaussian Processes Le Song Machine Learning II: Advanced Topics CSE 8803ML, Spring 01 Pictorial view of embedding distribution Transform the entire distribution to expected features Feature space Feature

More information

22/04/2014. Economic Research

22/04/2014. Economic Research 22/04/2014 Economic Research Forecasting Models for Exchange Rate Tuesday, April 22, 2014 The science of prognostics has been going through a rapid and fruitful development in the past decades, with various

More information

Neural networks (not in book)

Neural networks (not in book) (not in book) Another approach to classification is neural networks. were developed in the 1980s as a way to model how learning occurs in the brain. There was therefore wide interest in neural networks

More information

Linear & nonlinear classifiers

Linear & nonlinear classifiers Linear & nonlinear classifiers Machine Learning Hamid Beigy Sharif University of Technology Fall 1394 Hamid Beigy (Sharif University of Technology) Linear & nonlinear classifiers Fall 1394 1 / 34 Table

More information

Design Collocation Neural Network to Solve Singular Perturbed Problems with Initial Conditions

Design Collocation Neural Network to Solve Singular Perturbed Problems with Initial Conditions Article International Journal of Modern Engineering Sciences, 204, 3(): 29-38 International Journal of Modern Engineering Sciences Journal homepage:www.modernscientificpress.com/journals/ijmes.aspx ISSN:

More information

Supplementary Technical Details and Results

Supplementary Technical Details and Results Supplementary Technical Details and Results April 6, 2016 1 Introduction This document provides additional details to augment the paper Efficient Calibration Techniques for Large-scale Traffic Simulators.

More information

CSE 352 (AI) LECTURE NOTES Professor Anita Wasilewska. NEURAL NETWORKS Learning

CSE 352 (AI) LECTURE NOTES Professor Anita Wasilewska. NEURAL NETWORKS Learning CSE 352 (AI) LECTURE NOTES Professor Anita Wasilewska NEURAL NETWORKS Learning Neural Networks Classifier Short Presentation INPUT: classification data, i.e. it contains an classification (class) attribute.

More information

An Adaptive Neural Network Scheme for Radar Rainfall Estimation from WSR-88D Observations

An Adaptive Neural Network Scheme for Radar Rainfall Estimation from WSR-88D Observations 2038 JOURNAL OF APPLIED METEOROLOGY An Adaptive Neural Network Scheme for Radar Rainfall Estimation from WSR-88D Observations HONGPING LIU, V.CHANDRASEKAR, AND GANG XU Colorado State University, Fort Collins,

More information

Bayesian Machine Learning

Bayesian Machine Learning Bayesian Machine Learning Andrew Gordon Wilson ORIE 6741 Lecture 2: Bayesian Basics https://people.orie.cornell.edu/andrew/orie6741 Cornell University August 25, 2016 1 / 17 Canonical Machine Learning

More information

Mining Classification Knowledge

Mining Classification Knowledge Mining Classification Knowledge Remarks on NonSymbolic Methods JERZY STEFANOWSKI Institute of Computing Sciences, Poznań University of Technology SE lecture revision 2013 Outline 1. Bayesian classification

More information

Should all Machine Learning be Bayesian? Should all Bayesian models be non-parametric?

Should all Machine Learning be Bayesian? Should all Bayesian models be non-parametric? Should all Machine Learning be Bayesian? Should all Bayesian models be non-parametric? Zoubin Ghahramani Department of Engineering University of Cambridge, UK zoubin@eng.cam.ac.uk http://learning.eng.cam.ac.uk/zoubin/

More information

Introduction to Neural Networks

Introduction to Neural Networks Introduction to Neural Networks What are (Artificial) Neural Networks? Models of the brain and nervous system Highly parallel Process information much more like the brain than a serial computer Learning

More information

Feed-forward Network Functions

Feed-forward Network Functions Feed-forward Network Functions Sargur Srihari Topics 1. Extension of linear models 2. Feed-forward Network Functions 3. Weight-space symmetries 2 Recap of Linear Models Linear Models for Regression, Classification

More information

Direct Method for Training Feed-forward Neural Networks using Batch Extended Kalman Filter for Multi- Step-Ahead Predictions

Direct Method for Training Feed-forward Neural Networks using Batch Extended Kalman Filter for Multi- Step-Ahead Predictions Direct Method for Training Feed-forward Neural Networks using Batch Extended Kalman Filter for Multi- Step-Ahead Predictions Artem Chernodub, Institute of Mathematical Machines and Systems NASU, Neurotechnologies

More information

An artificial neural networks (ANNs) model is a functional abstraction of the

An artificial neural networks (ANNs) model is a functional abstraction of the CHAPER 3 3. Introduction An artificial neural networs (ANNs) model is a functional abstraction of the biological neural structures of the central nervous system. hey are composed of many simple and highly

More information

Hierarchical models for the rainfall forecast DATA MINING APPROACH

Hierarchical models for the rainfall forecast DATA MINING APPROACH Hierarchical models for the rainfall forecast DATA MINING APPROACH Thanh-Nghi Do dtnghi@cit.ctu.edu.vn June - 2014 Introduction Problem large scale GCM small scale models Aim Statistical downscaling local

More information

Introduction to Neural Networks

Introduction to Neural Networks Introduction to Neural Networks Philipp Koehn 4 April 205 Linear Models We used before weighted linear combination of feature values h j and weights λ j score(λ, d i ) = j λ j h j (d i ) Such models can

More information

Classification and Pattern Recognition

Classification and Pattern Recognition Classification and Pattern Recognition Léon Bottou NEC Labs America COS 424 2/23/2010 The machine learning mix and match Goals Representation Capacity Control Operational Considerations Computational Considerations

More information

Lecture 16: Introduction to Neural Networks

Lecture 16: Introduction to Neural Networks Lecture 16: Introduction to Neural Networs Instructor: Aditya Bhasara Scribe: Philippe David CS 5966/6966: Theory of Machine Learning March 20 th, 2017 Abstract In this lecture, we consider Bacpropagation,

More information

A multiple testing procedure for input variable selection in neural networks

A multiple testing procedure for input variable selection in neural networks A multiple testing procedure for input variable selection in neural networks MicheleLaRoccaandCiraPerna Department of Economics and Statistics - University of Salerno Via Ponte Don Melillo, 84084, Fisciano

More information

COMS 4771 Introduction to Machine Learning. Nakul Verma

COMS 4771 Introduction to Machine Learning. Nakul Verma COMS 4771 Introduction to Machine Learning Nakul Verma Announcements HW1 due next lecture Project details are available decide on the group and topic by Thursday Last time Generative vs. Discriminative

More information

Multivariate Data Analysis and Machine Learning in High Energy Physics (III)

Multivariate Data Analysis and Machine Learning in High Energy Physics (III) Multivariate Data Analysis and Machine Learning in High Energy Physics (III) Helge Voss (MPI K, Heidelberg) Graduierten-Kolleg, Freiburg, 11.5-15.5, 2009 Outline Summary of last lecture 1-dimensional Likelihood:

More information

ECE521: Inference Algorithms and Machine Learning University of Toronto. Assignment 1: k-nn and Linear Regression

ECE521: Inference Algorithms and Machine Learning University of Toronto. Assignment 1: k-nn and Linear Regression ECE521: Inference Algorithms and Machine Learning University of Toronto Assignment 1: k-nn and Linear Regression TA: Use Piazza for Q&A Due date: Feb 7 midnight, 2017 Electronic submission to: ece521ta@gmailcom

More information

Midterm Review CS 6375: Machine Learning. Vibhav Gogate The University of Texas at Dallas

Midterm Review CS 6375: Machine Learning. Vibhav Gogate The University of Texas at Dallas Midterm Review CS 6375: Machine Learning Vibhav Gogate The University of Texas at Dallas Machine Learning Supervised Learning Unsupervised Learning Reinforcement Learning Parametric Y Continuous Non-parametric

More information

Multi-Plant Photovoltaic Energy Forecasting Challenge: Second place solution

Multi-Plant Photovoltaic Energy Forecasting Challenge: Second place solution Multi-Plant Photovoltaic Energy Forecasting Challenge: Second place solution Clément Gautrais 1, Yann Dauxais 1, and Maël Guilleme 2 1 University of Rennes 1/Inria Rennes clement.gautrais@irisa.fr 2 Energiency/University

More information

Pattern Recognition and Machine Learning. Bishop Chapter 6: Kernel Methods

Pattern Recognition and Machine Learning. Bishop Chapter 6: Kernel Methods Pattern Recognition and Machine Learning Chapter 6: Kernel Methods Vasil Khalidov Alex Kläser December 13, 2007 Training Data: Keep or Discard? Parametric methods (linear/nonlinear) so far: learn parameter

More information

Neural Networks and the Back-propagation Algorithm

Neural Networks and the Back-propagation Algorithm Neural Networks and the Back-propagation Algorithm Francisco S. Melo In these notes, we provide a brief overview of the main concepts concerning neural networks and the back-propagation algorithm. We closely

More information

Statistical Techniques in Robotics (16-831, F12) Lecture#21 (Monday November 12) Gaussian Processes

Statistical Techniques in Robotics (16-831, F12) Lecture#21 (Monday November 12) Gaussian Processes Statistical Techniques in Robotics (16-831, F12) Lecture#21 (Monday November 12) Gaussian Processes Lecturer: Drew Bagnell Scribe: Venkatraman Narayanan 1, M. Koval and P. Parashar 1 Applications of Gaussian

More information

Last update: October 26, Neural networks. CMSC 421: Section Dana Nau

Last update: October 26, Neural networks. CMSC 421: Section Dana Nau Last update: October 26, 207 Neural networks CMSC 42: Section 8.7 Dana Nau Outline Applications of neural networks Brains Neural network units Perceptrons Multilayer perceptrons 2 Example Applications

More information

Discriminative Models

Discriminative Models No.5 Discriminative Models Hui Jiang Department of Electrical Engineering and Computer Science Lassonde School of Engineering York University, Toronto, Canada Outline Generative vs. Discriminative models

More information

SINGLE-TASK AND MULTITASK SPARSE GAUSSIAN PROCESSES

SINGLE-TASK AND MULTITASK SPARSE GAUSSIAN PROCESSES SINGLE-TASK AND MULTITASK SPARSE GAUSSIAN PROCESSES JIANG ZHU, SHILIANG SUN Department of Computer Science and Technology, East China Normal University 500 Dongchuan Road, Shanghai 20024, P. R. China E-MAIL:

More information

A Hybrid ARIMA and Neural Network Model to Forecast Particulate. Matter Concentration in Changsha, China

A Hybrid ARIMA and Neural Network Model to Forecast Particulate. Matter Concentration in Changsha, China A Hybrid ARIMA and Neural Network Model to Forecast Particulate Matter Concentration in Changsha, China Guangxing He 1, Qihong Deng 2* 1 School of Energy Science and Engineering, Central South University,

More information

Relevance Vector Machines for Earthquake Response Spectra

Relevance Vector Machines for Earthquake Response Spectra 2012 2011 American American Transactions Transactions on on Engineering Engineering & Applied Applied Sciences Sciences. American Transactions on Engineering & Applied Sciences http://tuengr.com/ateas

More information

Artificial Neural Networks. MGS Lecture 2

Artificial Neural Networks. MGS Lecture 2 Artificial Neural Networks MGS 2018 - Lecture 2 OVERVIEW Biological Neural Networks Cell Topology: Input, Output, and Hidden Layers Functional description Cost functions Training ANNs Back-Propagation

More information

STA 4273H: Statistical Machine Learning

STA 4273H: Statistical Machine Learning STA 4273H: Statistical Machine Learning Russ Salakhutdinov Department of Statistics! rsalakhu@utstat.toronto.edu! http://www.utstat.utoronto.ca/~rsalakhu/ Sidney Smith Hall, Room 6002 Lecture 3 Linear

More information

Learning Gaussian Process Models from Uncertain Data

Learning Gaussian Process Models from Uncertain Data Learning Gaussian Process Models from Uncertain Data Patrick Dallaire, Camille Besse, and Brahim Chaib-draa DAMAS Laboratory, Computer Science & Software Engineering Department, Laval University, Canada

More information

Neural Networks with Applications to Vision and Language. Feedforward Networks. Marco Kuhlmann

Neural Networks with Applications to Vision and Language. Feedforward Networks. Marco Kuhlmann Neural Networks with Applications to Vision and Language Feedforward Networks Marco Kuhlmann Feedforward networks Linear separability x 2 x 2 0 1 0 1 0 0 x 1 1 0 x 1 linearly separable not linearly separable

More information

Outliers Treatment in Support Vector Regression for Financial Time Series Prediction

Outliers Treatment in Support Vector Regression for Financial Time Series Prediction Outliers Treatment in Support Vector Regression for Financial Time Series Prediction Haiqin Yang, Kaizhu Huang, Laiwan Chan, Irwin King, and Michael R. Lyu Department of Computer Science and Engineering

More information

Article from. Predictive Analytics and Futurism. July 2016 Issue 13

Article from. Predictive Analytics and Futurism. July 2016 Issue 13 Article from Predictive Analytics and Futurism July 2016 Issue 13 Regression and Classification: A Deeper Look By Jeff Heaton Classification and regression are the two most common forms of models fitted

More information

Practicals 5 : Perceptron

Practicals 5 : Perceptron Université Paul Sabatier M2 SE Data Mining Practicals 5 : Perceptron Framework The aim of this last session is to introduce the basics of neural networks theory through the special case of the perceptron.

More information

Exploiting k-nearest Neighbor Information with Many Data

Exploiting k-nearest Neighbor Information with Many Data Exploiting k-nearest Neighbor Information with Many Data 2017 TEST TECHNOLOGY WORKSHOP 2017. 10. 24 (Tue.) Yung-Kyun Noh Robotics Lab., Contents Nonparametric methods for estimating density functions Nearest

More information

DEPARTMENT OF COMPUTER SCIENCE Autumn Semester MACHINE LEARNING AND ADAPTIVE INTELLIGENCE

DEPARTMENT OF COMPUTER SCIENCE Autumn Semester MACHINE LEARNING AND ADAPTIVE INTELLIGENCE Data Provided: None DEPARTMENT OF COMPUTER SCIENCE Autumn Semester 203 204 MACHINE LEARNING AND ADAPTIVE INTELLIGENCE 2 hours Answer THREE of the four questions. All questions carry equal weight. Figures

More information

Automatic Rank Determination in Projective Nonnegative Matrix Factorization

Automatic Rank Determination in Projective Nonnegative Matrix Factorization Automatic Rank Determination in Projective Nonnegative Matrix Factorization Zhirong Yang, Zhanxing Zhu, and Erkki Oja Department of Information and Computer Science Aalto University School of Science and

More information

CSC242: Intro to AI. Lecture 21

CSC242: Intro to AI. Lecture 21 CSC242: Intro to AI Lecture 21 Administrivia Project 4 (homeworks 18 & 19) due Mon Apr 16 11:59PM Posters Apr 24 and 26 You need an idea! You need to present it nicely on 2-wide by 4-high landscape pages

More information

Machine Learning for Large-Scale Data Analysis and Decision Making A. Neural Networks Week #6

Machine Learning for Large-Scale Data Analysis and Decision Making A. Neural Networks Week #6 Machine Learning for Large-Scale Data Analysis and Decision Making 80-629-17A Neural Networks Week #6 Today Neural Networks A. Modeling B. Fitting C. Deep neural networks Today s material is (adapted)

More information

Preferred citation style. Schüssler, N. (2009) Challenges of route choice models derived from GPS, Workshop on Discrete Choice Models, EPF

Preferred citation style. Schüssler, N. (2009) Challenges of route choice models derived from GPS, Workshop on Discrete Choice Models, EPF Preferred citation style Schüssler, N. (2009) Challenges of route choice models derived from GPS, Workshop on Discrete Choice Models, EPF Lausanne, Lausanne, August 2009. Challenges of route choice models

More information

Sections 18.6 and 18.7 Artificial Neural Networks

Sections 18.6 and 18.7 Artificial Neural Networks Sections 18.6 and 18.7 Artificial Neural Networks CS4811 - Artificial Intelligence Nilufer Onder Department of Computer Science Michigan Technological University Outline The brain vs. artifical neural

More information

APPLICATION OF RADIAL BASIS FUNCTION NEURAL NETWORK, TO ESTIMATE THE STATE OF HEALTH FOR LFP BATTERY

APPLICATION OF RADIAL BASIS FUNCTION NEURAL NETWORK, TO ESTIMATE THE STATE OF HEALTH FOR LFP BATTERY International Journal of Electrical and Electronics Engineering (IJEEE) ISSN(P): 2278-9944; ISSN(E): 2278-9952 Vol. 7, Issue 1, Dec - Jan 2018, 1-6 IASET APPLICATION OF RADIAL BASIS FUNCTION NEURAL NETWORK,

More information

Learning from Examples

Learning from Examples Learning from Examples Data fitting Decision trees Cross validation Computational learning theory Linear classifiers Neural networks Nonparametric methods: nearest neighbor Support vector machines Ensemble

More information

Neural networks. Chapter 20, Section 5 1

Neural networks. Chapter 20, Section 5 1 Neural networks Chapter 20, Section 5 Chapter 20, Section 5 Outline Brains Neural networks Perceptrons Multilayer perceptrons Applications of neural networks Chapter 20, Section 5 2 Brains 0 neurons of

More information

Bits of Machine Learning Part 1: Supervised Learning

Bits of Machine Learning Part 1: Supervised Learning Bits of Machine Learning Part 1: Supervised Learning Alexandre Proutiere and Vahan Petrosyan KTH (The Royal Institute of Technology) Outline of the Course 1. Supervised Learning Regression and Classification

More information

Artificial Neural Networks Examination, June 2004

Artificial Neural Networks Examination, June 2004 Artificial Neural Networks Examination, June 2004 Instructions There are SIXTY questions (worth up to 60 marks). The exam mark (maximum 60) will be added to the mark obtained in the laborations (maximum

More information

Statistical Inference

Statistical Inference Statistical Inference Liu Yang Florida State University October 27, 2016 Liu Yang, Libo Wang (Florida State University) Statistical Inference October 27, 2016 1 / 27 Outline The Bayesian Lasso Trevor Park

More information