2011 COURSE IN NEUROINFORMATICS MARINE BIOLOGICAL LABORATORY WOODS HOLE, MA Introduction to Spline Models or Advanced Connect-the-Dots Uri Eden BU Department.

Slides:



Advertisements
Similar presentations
Interpolating curves.
Advertisements

Computer Graphics (Spring 2008) COMS 4160, Lecture 6: Curves 1
Polynomial Curve Fitting BITS C464/BITS F464 Navneet Goyal Department of Computer Science, BITS-Pilani, Pilani Campus, India.
CS 445 / 645 Introduction to Computer Graphics Lecture 23 Bézier Curves Lecture 23 Bézier Curves.
CS 445 / 645 Introduction to Computer Graphics Lecture 22 Hermite Splines Lecture 22 Hermite Splines.
March 1, 2009Dr. Muhammed Al-Mulhem1 ICS 415 Computer Graphics Hermite Splines Dr. Muhammed Al-Mulhem March 1, 2009 Dr. Muhammed Al-Mulhem March 1, 2009.
Flexible smoothing with B-splines and Penalties or P-splines P-splines = B-splines + Penalization Applications : Generalized Linear and non linear Modelling.
Pattern Recognition and Machine Learning
CS 445/645 Fall 2001 Hermite and Bézier Splines. Specifying Curves Control Points –A set of points that influence the curve’s shape Knots –Control points.
Data mining and statistical learning - lecture 6
MATH 685/ CSI 700/ OR 682 Lecture Notes
Selected from presentations by Jim Ramsay, McGill University, Hongliang Fei, and Brian Quanz Basis Basics.
Basis Expansion and Regularization Presenter: Hongliang Fei Brian Quanz Brian Quanz Date: July 03, 2008.
Chapter 18 Interpolation The Islamic University of Gaza
1 Chapter 4 Interpolation and Approximation Lagrange Interpolation The basic interpolation problem can be posed in one of two ways: The basic interpolation.
Dr. S.M. Malaek Assistant: M. Younesi
08/30/00 Dinesh Manocha, COMP258 Hermite Curves A mathematical representation as a link between the algebraic & geometric form Defined by specifying the.
Slide 127 October 1999CS Computer Graphics (Top Changwatchai) Review of Spline Concepts Sections 10-6 to in Hearn & Baker Splines can be 2D.
Splines II – Interpolating Curves
Advanced Computer Graphics (Spring 2005) COMS 4162, Lecture 12: Spline Curves (review) Ravi Ramamoorthi Most material.
CSE 221: Probabilistic Analysis of Computer Systems Topics covered: Statistical inference (Sec. )
Drawing Parametric Curves Jean-Paul Mueller. Curves - The parametric form of a curve expresses the value of each spatial variable for points on the curve.
Modelling: Curves Week 11, Wed Mar 23
ECIV 301 Programming & Graphics Numerical Methods for Engineers Lecture 24 Regression Analysis-Chapter 17.
University of British Columbia CPSC 414 Computer Graphics © Tamara Munzner 1 Curves Week 13, Mon 24 Nov 2003.
ECIV 301 Programming & Graphics Numerical Methods for Engineers Lecture 21 CURVE FITTING Chapter 18 Function Interpolation and Approximation.
CpE- 310B Engineering Computation and Simulation Dr. Manal Al-Bzoor
CMPS1371 Introduction to Computing for Engineers NUMERICAL METHODS.
A D V A N C E D C O M P U T E R G R A P H I C S CMSC 635 January 15, 2013 Spline curves 1/23 Curves and Surfaces.
CS 376 Introduction to Computer Graphics 04 / 23 / 2007 Instructor: Michael Eckmann.
V. Space Curves Types of curves Explicit Implicit Parametric.
Geographic Information Science
INTERPOLATION & APPROXIMATION. Curve algorithm General curve shape may be generated using method of –Interpolation (also known as curve fitting) Curve.
Course 13 Curves and Surfaces. Course 13 Curves and Surface Surface Representation Representation Interpolation Approximation Surface Segmentation.
June D Object Representation Shmuel Wimer Bar Ilan Univ., School of Engineering.
Computer Programming (TKK-2144) 13/14 Semester 1 Instructor: Rama Oktavian Office Hr.: M.13-15, W Th , F
Introduction to Numerical Analysis I MATH/CMPSC 455 Splines.
Greg Humphreys CS445: Intro Graphics University of Virginia, Fall 2003 Parametric Curves & Surfaces Greg Humphreys University of Virginia CS 445, Spring.
CSCE 441: Keyframe Animation/Smooth Curves (Cont.) Jinxiang Chai.
CSCE 441: Keyframe Animation/Smooth Curves (Cont.) Jinxiang Chai.
Keyframe Interpolation and Speed Control Jehee Lee Seoul National University.
By: Mark Coose Joetta Swift Micah Weiss. What Problems Can Interpolation Solve? Given a table of values, find a simple function that passes through the.
Gaussian Process and Prediction. (C) 2001 SNU CSE Artificial Intelligence Lab (SCAI)2 Outline Gaussian Process and Bayesian Regression  Bayesian regression.
Review of statistical modeling and probability theory Alan Moses ML4bio.
Ship Computer Aided Design
Foundations of Computer Graphics (Spring 2012) CS 184, Lecture 12: Curves 1
1 Lecture 8 Post-Graduate Students Advanced Programming (Introduction to MATLAB) Code: ENG 505 Dr. Basheer M. Nasef Computers & Systems Dept.
Curves University of British Columbia CPSC 314 Computer Graphics Jan-Apr 2013 Tamara Munzner.
Piecewise Polynomial Parametric Curves Sun-Jeong Kim.
Object Modeling: Curves and Surfaces CEng 477 Introduction to Computer Graphics.
Introduction to Parametric Curve and Surface Modeling.
Interpolation - Introduction
Markov Chain Monte Carlo in R
Parametric Curves cgvr.korea.ac.kr.
Chapter 10-2: Curves.
Interpolation Estimation of intermediate values between precise data points. The most common method is: Although there is one and only one nth-order.
Ch3: Model Building through Regression
CSE 167 [Win 17], Lecture 9: Curves 1 Ravi Ramamoorthi
Final Project CSCE 790E (Medical Image Processing)
© University of Wisconsin, CS559 Spring 2004
Chapter XVII Parametric Curves and Surfaces
a kind of filtering that leads to useful features
Spline Interpolation Class XVII.
Implicit Functions Some surfaces can be represented as the vanishing points of functions (defined over 3D space) Places where a function f(x,y,z)=0 Some.
a kind of filtering that leads to useful features
CSCE 441: Keyframe Animation/Smooth Curves (Cont.)
Introduction to Parametric Curve and Surface Modeling
Type to enter a caption. Computer Graphics Week 10 Lecture 1.
Spline representation. ❖ A spline is a flexible strip used to produce a smooth curve through a designated set of points. ❖ Mathematically describe such.
Presentation transcript:

2011 COURSE IN NEUROINFORMATICS MARINE BIOLOGICAL LABORATORY WOODS HOLE, MA Introduction to Spline Models or Advanced Connect-the-Dots Uri Eden BU Department of Mathematics and Statistics August 16, 2011 Slide acknowledgements: D. Keren, Dept. of Computer Sci, Haifa U.

Motivating Problem Compute a smooth firing rate function from repeated spike train data Time (ms) Trial

Motivating Problem Compute a smooth firing rate function from repeated spike train data Firing Rate (Hz) Time (sec)

Motivating Problem Compute a smooth firing rate function from repeated spike train data Time (sec) Firing Rate (Hz)

Motivating Problem Compute a smooth firing rate function from repeated spike train data Time (sec) Firing Rate (Hz)

Problem Statement Given a set of control points, P i, known to be on a curve, find a parametric function that interpolates or approximates the curve.

For interpolating 2 points we need a polynomial of degree 1. Polynomial interpolation (x 1,y 1 ) (x 2,y 2 )

For interpolating 2 points we need a polynomial of degree 1. Polynomial interpolation (x 1,y 1 ) (x 2,y 2 )

For interpolating 2 points we need a polynomial of degree 1. For interpolating 3 points we need a polynomial of degree 2. In general, for interpolating n points, we need a polynomial of degree n-1. Polynomial interpolation (x 1,y 1 ) (x 2,y 2 )

Polynomial interpolation Let’s try it on our example control points: Time (sec) Firing Rate (Hz)

Polynomial interpolation Let’s try it on our example control points: Time (sec) Firing Rate (Hz)

Polynomial interpolation Disadvantages: –Coefficients are difficult to interpret geometrically –No local operations – adjusting parameters to change one region will change all others –If polynomial degree is high: Resulting function will have unwanted fluctuations Numerical precision problems may arise. –If polynomial degree is low, the resulting curve will not interpolate the points

Spline interpolation Another approach: Polynomial splines –Advantages: Rich, flexible representation Geometrically meaningful coefficients –Local, piecewise, low-degree, polynomial curves with continuous, smooth joints

Spline interpolation Simplest Example: Zeroth order ‘spline’ –Constant function between control points –Simple mathematical representation Local (x i+1,y i+1 ) (x i,y i )

Motivating Problem Compute a smooth firing rate function from repeated spike train data Time (sec) Firing Rate (Hz)

Motivating Problem Compute a smooth firing rate function from repeated spike train data Time (sec) Firing Rate (Hz) Local but not continuous

Spline interpolation Simple Example: Linear ‘spline’ –Linear combination of the two nearest control points –Simple mathematical representation where

Spline interpolation Simple Example: Linear ‘spline’

Spline interpolation Simple Example: Linear ‘spline’

Spline interpolation Simple Example: Linear ‘spline’

Spline interpolation Simple Example: Linear ‘spline’

Spline interpolation Simple Example: Linear ‘spline’

Motivating Problem Compute a smooth firing rate function from repeated spike train data Time (sec) Firing Rate (Hz)

Motivating Problem Compute a smooth firing rate function from repeated spike train data Time (sec) Firing Rate (Hz) Local and continuous, but not smooth

Spline interpolation Cubic Splines –Let be cubic functions. –Produces continuous, differentiable curves –Need to decide how to model derivatives at control points Natural Splines: Derivative is zero at endpoints Cardinal Splines: Derivative determined by nearby control points Hermitian Splines: Derivatives determined by separate parameters

Spline interpolation Cardinal splines –Linear combination of 4 nearest control points –Features: Derivatives determined by adjacent points Shape controlled by tension parameter,

Spline interpolation Cardinal splines

Spline interpolation Cardinal splines –Basis/blending functions for cardinal spline with s = 0.5.

Spline interpolation Cardinal splines

Spline interpolation Cardinal splines

Spline interpolation Cardinal splines

Spline interpolation Cardinal splines

Motivating Problem Compute a smooth firing rate function from repeated spike train data Time (sec) Firing Rate (Hz)

Motivating Problem Compute a smooth firing rate function from repeated spike train data Time (sec) Firing Rate (Hz)

Fitting Spline Models To define a spline model, we must specify –Control point locations (x-coordinates) –Tension parameter Often these are selected ad-hoc, but adaptive Bayesian procedures (e.g. BARS) to determine optimal values typically produce better results.

Fitting Spline Models Under a specified noise model, the spline parameters (heights of control points) can be estimated by maximum likelihood. –E.g. If, where is Gaussian white noise, and is a spline function, the ML estimator,, can be found by simple linear regression. –For spike data, ML spline parameters can be estimated using point process models.

Conclusions Splines offer one approach to modeling smooth functional relationships in neuroscience data. Cubic splines are local 3 rd degree polynomials that are piecewise continuous and differentiable. Under common noise distributions, spline models can be fit easily using ML methods. Goodness of spline model fits depends on selection of control point locations and tension parameters.