Geometric Approach Geometric Interpretation:

Slides:



Advertisements
Similar presentations
Modeling of Data. Basic Bayes theorem Bayes theorem relates the conditional probabilities of two events A, and B: A might be a hypothesis and B might.
Advertisements

State Space Models. Let { x t :t T} and { y t :t T} denote two vector valued time series that satisfy the system of equations: y t = A t x t + v t (The.
Fast Algorithms For Hierarchical Range Histogram Constructions
A KTEC Center of Excellence 1 Pattern Analysis using Convex Optimization: Part 2 of Chapter 7 Discussion Presenter: Brian Quanz.
Visual Recognition Tutorial
Regression Analysis. Unscheduled Maintenance Issue: l 36 flight squadrons l Each experiences unscheduled maintenance actions (UMAs) l UMAs costs $1000.
SEBD Tutorial, June Monitoring Distributed Streams Joint works with Tsachi Scharfman, Daniel Keren.
OBBTree: A Hierarchical Structure for Rapid Interference Detection Gottschalk, M. C. Lin and D. ManochaM. C. LinD. Manocha Department of Computer Science,
EE 685 presentation Optimization Flow Control, I: Basic Algorithm and Convergence By Steven Low and David Lapsley Asynchronous Distributed Algorithm Proof.
Single Point of Contact Manipulation of Unknown Objects Stuart Anderson Advisor: Reid Simmons School of Computer Science Carnegie Mellon University.
Efficient Distance Computation between Non-Convex Objects By Sean Quinlan Presented by Sean Augenstein and Nicolas Lee.
Bump Hunting The objective PRIM algorithm Beam search References: Feelders, A.J. (2002). Rule induction by bump hunting. In J. Meij (Ed.), Dealing with.
PIC: Practical Internet Coordinates for Distance Estimation Manuel Costa joint work with Miguel Castro, Ant Rowstron, Peter Key Microsoft Research Cambridge.
10/5/ Geometric Approach Geometric Interpretation: Geometric Interpretation: Each node holds a statistics vector Each node holds a statistics vector.
Chapter 20 Linear Regression. What if… We believe that an important relation between two measures exists? For example, we ask 5 people about their salary.
CIS 350 – I Game Programming Instructor: Rolf Lakaemper.
A Passive Approach to Sensor Network Localization Rahul Biswas and Sebastian Thrun International Conference on Intelligent Robots and Systems 2004 Presented.
CpSc 881: Machine Learning Evaluating Hypotheses.
ECE-7000: Nonlinear Dynamical Systems Overfitting and model costs Overfitting  The more free parameters a model has, the better it can be adapted.
Neural Nets: Something you can use and something to think about Cris Koutsougeras What are Neural Nets What are they good for Pointers to some models and.
Copyright © 2014 by McGraw-Hill Higher Education. All rights reserved. Essentials of Business Statistics: Communicating with Numbers By Sanjiv Jaggia and.
Intelligent Database Systems Lab 國立雲林科技大學 National Yunlin University of Science and Technology Efficient Optimal Linear Boosting of a Pair of Classifiers.
Local Prediction of a Spatio-Temporal Process with Application to Wet Sulfate Deposition Presented by Isin OZAKSOY.
Additional Topics in Prediction Methodology. Introduction Predictive distribution for random variable Y 0 is meant to capture all the information about.
Spectral Partitioning: One way to slice a problem in half C B A.
Giansalvo EXIN Cirrincione unit #4 Single-layer networks They directly compute linear discriminant functions using the TS without need of determining.
Mingze Zhang, Mun Choon Chan and A. L. Ananda School of Computing
Support vector machines
Chapter 8: Estimating with Confidence
Chapter 8: Estimating with Confidence
STATISTICAL ORBIT DETERMINATION Kalman (sequential) filter
Deep Feedforward Networks
Some General Concepts of Point Estimation
Core-Sets and Geometric Optimization problems.
CH 5: Multivariate Methods
Department of Civil and Environmental Engineering
Multiple Regression.
Privacy and Fault-Tolerance in Distributed Optimization Nitin Vaidya University of Illinois at Urbana-Champaign.
Additive Combinatorics and its Applications in Theoretical CS
Data Analysis and Statistical Software I ( ) Quarter: Autumn 02/03
Sampling Distribution of the Sample Mean
Modelling data and curve fitting
Research: algorithmic solutions for networking
Ying shen Sse, tongji university Sep. 2016
Experiments in Machine Learning
Aviv Rosenberg 10/01/18 Seminar on Experts and Bandits
Large Scale Support Vector Machines
Introduction to Predictive Modeling
On the effect of randomness on planted 3-coloring models
Curve fit metrics When we fit a curve to data we ask:
Chapter 8: Estimating with Confidence
Curve fit metrics When we fit a curve to data we ask:
Chapter 8: Estimating with Confidence
Support vector machines
Chapter 8: Estimating with Confidence
Learning From Observed Data
Chapter 8: Estimating with Confidence
Chapter 8: Estimating with Confidence
Chapter 8: Estimating with Confidence
Essential Statistics Sampling Distributions
Chapter 8: Estimating with Confidence
Chapter 8: Estimating with Confidence
Chapter 8: Estimating with Confidence
Chapter 8: Estimating with Confidence
Chapter 8: Estimating with Confidence
Chapter 8: Estimating with Confidence
Image Stitching Linda Shapiro ECE/CSE 576.
Some General Concepts of Point Estimation
Introduction to Machine Learning
Image Stitching Linda Shapiro ECE P 596.
Presentation transcript:

Geometric Approach Geometric Interpretation: Each node holds a statistics vector Coloring the vector space Grey:: function > threshold White:: function <= threshold Goal: determine color of global data vector (average). 12/29/2017

Bounding the Convex Hull Observation: average is in the convex hull  If convex hull monochromatic then average too But – convex hull may become large 12/29/2017

Drift Vectors Periodically calculate an estimate vector - the current global Each node maintains a drift vector – the change in the local statistics vector since the last time the estimate vector was calculated Global average statistics vector is also the average of the drift vectors 12/29/2017

The Bounding Theorem [SIGMOD’06] A reference point is known to all nodes Each vertex constructs a sphere Theorem: convex hull is bounded by the union of spheres  Local constraints! 12/29/2017

Proofs of the bounding theorem: SIGMOD06 – induction on the dimension. Micha Sharir – induction on number of points. Yuri Rabinovich – uses the following observation: z is not in the sphere supported by x,y iff (x-z,y-z)>0.

Basic Algorithm An initial estimate vector is calculated Nodes check color of drift spheres Drift vector is the diameter of the drift sphere If any sphere non monochromatic: node triggers re-calculation of estimate vector 12/29/2017

Reuters Corpus (RCV1-v2) 800,000+ news stories Aug 20 1996 -- Aug 19 1997 Corporate/Industrial tagging n=10 12/29/2017 10 nodes, random data distribution

Trade-off: Accuracy vs. Performance Inefficiency: value of function on average is close to the threshold Performance can be enhanced at the cost of less accurate result: Set error margin around the threshold value 12/29/2017

Performance Analysis 12/29/2017

Performance Analysis (cntd.) Change dist(…,f,r) with D_global 12/29/2017

Balancing Globally calculating average is costly Often possible to average only some of the data vectors. 12/29/2017

Shape Sensitivity [PODS’08] Fitting cover to Data Fitting cover to threshold surface Specific function classes 12/29/2017

Fitting Cover to Data (using the covariance matrix) 12/29/2017

Fitting Cover to Threshold Surface -- Reference Vector Selection 12/29/2017

Distance Fields Skeleton, Medial Axis 12/29/2017

Results – Shape Sensitivity 12/29/2017

Prediction-Based Geometric Monitoring [SIGMOD’12] ΔV1 ΔV2 ΔV3 ΔV4 ΔV5 ep ΔVp1 ΔVp2 ΔVp3 ΔVp4 ΔVp5 f(v(t)) > T v(t) Instead of drift vectors which expressed the change of the local vectors since the last contact with the coordinating source, we now have prediction deviation vectors which denote how much accurate are the predictions provided by the adopted estimators. As long as local predictors remain good, the convex hull formed by the prediction deviation vectors will be tighter and local constraints monitored at each site will be stricter. click Moreover, notice that, together with the prediction deviations, the common reference vector (e^p) changes positions following the predicted v(t) movement. Stricter local constraints if local predictions remain accurate Keeping up with v(t) movement

Let the nodes communicate only when “something happens” Local Constraints Safe Zones! Let the nodes communicate only when “something happens” Send me your current measurements! Tell me only if your measurement is larger than 50!                            

These Safe Zones save more communication! Local Distributions   Reasonable to assume future data will behave similarly…       58 45 10 66 44 20 43 50 15 78 17 85 30 21 70 47 11 76 25 12 65 5 56 75 34 16                   These Safe Zones save more communication!

Optimal Safe Zones   1. Legal / Safe   2. Large: Minimize Communication

Example: Air quality monitoring       What are the optimal Safe Zones…?              

The Optimization Problem   Is this Convex?     Is this Linear? How many constraints are these? BAD NEWS: This problem is NP-hard.

The Optimization Problem             X   Step 3: Use non-convex optimization toolboxes (e.g. Matlab’s “fmincon”). These toolboxes use sophisticated Gradient Descent algorithms and return close-to-optimal results.

Data Set   How the data looks like

Ratio Queries           Example of triangular Safe Zones

Improvement over convex-hull cover method 5’000 hours Up to 200 nodes were involved in the experiment. The average improvement was by a factor of 17.5 Why do we improve so much?

Higher Dimensions        

Chi-Square Monitoring (5D)       Examples of axis aligned boxes as Safe Zones

Improvement over GM 1’000 hours 90 nodes The improvement over the Geometric Method gets more substantial in higher dimensions.

Safe Zones - Example

Biclique: Non-Convex Safe Zones Safe Zone Algorithm (for 2 nodes): Take the data points, build a bipartite graph(how?), find the maximal Biclique, these are your Safe Zones!

Conclusions Local filtering for large-scale distributed data systems Saving in communication is unlimited Bounded only by the aggregate over system lifetime Saving bandwidth, central resources, power. Not necessary to sacrifice precision and latency Less communication  more Privacy 12/29/2017