The Hadamard transform is a popular orthogonal transform with a low-complexity algorithm with O(N log N) complexity. In this presentation, we describe a new sub-linear complexity algorithm to compute the Hadamard transform of signals whose Hadamard transform coefficients are sparse - that is very few are non-zero.
IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com
DPPs everywhere: repulsive point processes for Monte Carlo integration, signa...Advanced-Concepts-Team
Determinantal point processes (DPPs) are specific repulsive point processes, which were introduced in the 1970s by Macchi to model fermion beams in quantum optics. More recently, they have been studied as models and sampling tools by statisticians and machine learners. Important statistical quantities associated to DPPs have geometric and algebraic interpretations, which makes them a fun object to study and a powerful algorithmic building block.
After a quick introduction to determinantal point processes, I will discuss some of our recent statistical applications of DPPs. First, we used DPPs to sample nodes in numerical integration, resulting in Monte Carlo integration with fast convergence with respect to the number of integrand evaluations. Second, we used DPP machinery to characterize the distribution of the zeros of time-frequency transforms of white noise, a recent challenge in signal processing. Third, we turned DPPs into low-error variable selection procedures in linear regression.
Towards a stable definition of Algorithmic RandomnessHector Zenil
Although information content is invariant up to an additive constant, the range of possible additive constants applicable to programming languages is so large that in practice it plays a major role in the actual evaluation of K(s), the Kolmogorov complexity of a string s. We present a summary of the approach we've developed to overcome the problem by calculating its algorithmic probability and evaluating the algorithmic complexity via the coding theorem, thereby providing a stable framework for Kolmogorov complexity even for short strings. We also show that reasonable formalisms produce reasonable complexity classifications.
NOW A DAYS VISIBLE LIGHT COMMUNICATION IS MORE POPULAR. The visible light communication (VLC) refers to the communication technology which utilizes the visible light source as a signal transmitter, the air as a transmission medium, and the appropriate photodiode as a signal receiving component.Visible light is thus by definition comprised
of visually-perceivable electromagnetic waves.
The visible spectrum covers wave lengths
from 380 nm to 750 nm. The Visible Light Communications Consortium (VLCC) which is mainly comprised of Japanese technology companies was founded in November 2003. It increases the data speed compared to any communication
Position
IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com
DPPs everywhere: repulsive point processes for Monte Carlo integration, signa...Advanced-Concepts-Team
Determinantal point processes (DPPs) are specific repulsive point processes, which were introduced in the 1970s by Macchi to model fermion beams in quantum optics. More recently, they have been studied as models and sampling tools by statisticians and machine learners. Important statistical quantities associated to DPPs have geometric and algebraic interpretations, which makes them a fun object to study and a powerful algorithmic building block.
After a quick introduction to determinantal point processes, I will discuss some of our recent statistical applications of DPPs. First, we used DPPs to sample nodes in numerical integration, resulting in Monte Carlo integration with fast convergence with respect to the number of integrand evaluations. Second, we used DPP machinery to characterize the distribution of the zeros of time-frequency transforms of white noise, a recent challenge in signal processing. Third, we turned DPPs into low-error variable selection procedures in linear regression.
Towards a stable definition of Algorithmic RandomnessHector Zenil
Although information content is invariant up to an additive constant, the range of possible additive constants applicable to programming languages is so large that in practice it plays a major role in the actual evaluation of K(s), the Kolmogorov complexity of a string s. We present a summary of the approach we've developed to overcome the problem by calculating its algorithmic probability and evaluating the algorithmic complexity via the coding theorem, thereby providing a stable framework for Kolmogorov complexity even for short strings. We also show that reasonable formalisms produce reasonable complexity classifications.
NOW A DAYS VISIBLE LIGHT COMMUNICATION IS MORE POPULAR. The visible light communication (VLC) refers to the communication technology which utilizes the visible light source as a signal transmitter, the air as a transmission medium, and the appropriate photodiode as a signal receiving component.Visible light is thus by definition comprised
of visually-perceivable electromagnetic waves.
The visible spectrum covers wave lengths
from 380 nm to 750 nm. The Visible Light Communications Consortium (VLCC) which is mainly comprised of Japanese technology companies was founded in November 2003. It increases the data speed compared to any communication
Position
PROGRAMMA ATTIVITA’ DIDATTICA A.A. 2016/17
DOTTORATO DI RICERCA IN INGEGNERIA STRUTTURALE E GEOTECNICA
____________________________________________________________
STOCHASTIC DYNAMICS AND MONTE CARLO SIMULATION IN EARTHQUAKE ENGINEERING APPLICATIONS
Lecture Series by
Agathoklis Giaralis, Ph.D., M.ASCE., P.E. City, University of London
Visiting Professor Sapienza University of Rome
FPGA Implementation of Large Area Efficient and Low Power Geortzel Algorithm ...IDES Editor
Spectrum analysis is very essential requirement in
instrumentation and communication signal interception
.Spectrum analysis is normally carried out by online or offline
FFT processing. But the FFT being highly mathematical
intensive, is not suitable for low area and low power
applications. Offline FFT processing can’t give the real time
spectrum estimation which is essential in communication
signal interception. Online FFT computation takes very high
resources, which makes the system costly and power hungry.
The Goertzel algorithm is a digital signal processing (DSP)
technique for identifying frequency components of a signal,
published by Dr. Gerald Goertzel in 1958. While the general
Fast Fourier transform (FFT) algorithm computes evenly
across the bandwidth of the incoming signal, the Goertzel
algorithm looks at specific, predetermined frequency. However
the implementation of Goertzel algorithm for spectrum
computation is not explored for FPGA implementation. The
FPGA being capable of offering high frequency data paths in
them become suitable for realizing high speed spectrum
analysis algorithms.
"Evaluation of the Hilbert Huang transformation of transient signals for brid...TRUSS ITN
Abstract: The assessment of bridge condition from vibration measurements has generally been determined via the monitoring of modal parameters determined though adaptations of the standard Fast Fourier Transform (FFT) or other stationary time-series based transformations. However, the non-stationary nature of measured vibration signals from damaged structures can limit the quality of frequency content information estimated by such methods. The Hilbert–Huang Transform’s (HHT) ability to decompose non-stationary measured vibration data into a time-frequency-energy representation allows signal variations to be identified sooner than other stationary-based transformations, thus potentially allowing early detection of damage. The present study uses data obtained from a progressive damage test conducted on a real bridge subjected to excitation from a double axle passing vehicle as a test subject. Decomposed vibration signals from the HHT and associated marginal spectrums are assessed to determine structural condition for various damage states and different locations along the bridge.
IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com
IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com
DESIGN OF QUATERNARY LOGICAL CIRCUIT USING VOLTAGE AND CURRENT MODE LOGICVLSICS Design
In VLSI technology, designers main concentration were on area required and on performance of the
device. In VLSI design power consumption is one of the major concerns due to continuous increase in chip
density and decline in size of CMOS circuits and frequency at which circuits are operating. By considering
these parameter logical circuits are designed using quaternary voltage mode logic and quaternary current
mode logic. Power consumption required for quaternary voltage mode logic is 51.78 % less as compared
to binary . Area in terms of number of transistor required for quaternary voltage mode logic is 3 times
more as compared to binary. As quaternary voltage mode circuit required large area as compared to
quaternary current mode circuit but power consumption required in quaternary voltage mode circuit is less
than that required in quaternary current mode circuit .
Integration with kernel methods, Transported meshfree methodsMercier Jean-Marc
I made public for discussion a first version (subject to changes) of a talk that will be given at the Particles 2019 conference.
The bold points of this presentation are the following :
1) We present new to my knowledge, sharp, estimations for Monte-Carlo type methods.
2) These estimations can be used in a wide variety of context to perform a sharp error analysis.
3) We present a class of numerical methods that we refer to as Transported Meshfree Methods. This class of methods can be used for a wide variety of problems based on Partial Differential Equations, among which Artificial Intelligence problems belongs.
4) We can guarantee, thanks to the error analysis, a worst-error while computing with transported meshfree methods. We can also check that this error matches optimal convergence rate.
Cosmetic shop management system project report.pdfKamal Acharya
Buying new cosmetic products is difficult. It can even be scary for those who have sensitive skin and are prone to skin trouble. The information needed to alleviate this problem is on the back of each product, but it's thought to interpret those ingredient lists unless you have a background in chemistry.
Instead of buying and hoping for the best, we can use data science to help us predict which products may be good fits for us. It includes various function programs to do the above mentioned tasks.
Data file handling has been effectively used in the program.
The automated cosmetic shop management system should deal with the automation of general workflow and administration process of the shop. The main processes of the system focus on customer's request where the system is able to search the most appropriate products and deliver it to the customers. It should help the employees to quickly identify the list of cosmetic product that have reached the minimum quantity and also keep a track of expired date for each cosmetic product. It should help the employees to find the rack number in which the product is placed.It is also Faster and more efficient way.
Final project report on grocery store management system..pdfKamal Acharya
In today’s fast-changing business environment, it’s extremely important to be able to respond to client needs in the most effective and timely manner. If your customers wish to see your business online and have instant access to your products or services.
Online Grocery Store is an e-commerce website, which retails various grocery products. This project allows viewing various products available enables registered users to purchase desired products instantly using Paytm, UPI payment processor (Instant Pay) and also can place order by using Cash on Delivery (Pay Later) option. This project provides an easy access to Administrators and Managers to view orders placed using Pay Later and Instant Pay options.
In order to develop an e-commerce website, a number of Technologies must be studied and understood. These include multi-tiered architecture, server and client-side scripting techniques, implementation technologies, programming language (such as PHP, HTML, CSS, JavaScript) and MySQL relational databases. This is a project with the objective to develop a basic website where a consumer is provided with a shopping cart website and also to know about the technologies used to develop such a website.
This document will discuss each of the underlying technologies to create and implement an e- commerce website.
Saudi Arabia stands as a titan in the global energy landscape, renowned for its abundant oil and gas resources. It's the largest exporter of petroleum and holds some of the world's most significant reserves. Let's delve into the top 10 oil and gas projects shaping Saudi Arabia's energy future in 2024.
Welcome to WIPAC Monthly the magazine brought to you by the LinkedIn Group Water Industry Process Automation & Control.
In this month's edition, along with this month's industry news to celebrate the 13 years since the group was created we have articles including
A case study of the used of Advanced Process Control at the Wastewater Treatment works at Lleida in Spain
A look back on an article on smart wastewater networks in order to see how the industry has measured up in the interim around the adoption of Digital Transformation in the Water Industry.
Sachpazis:Terzaghi Bearing Capacity Estimation in simple terms with Calculati...Dr.Costas Sachpazis
Terzaghi's soil bearing capacity theory, developed by Karl Terzaghi, is a fundamental principle in geotechnical engineering used to determine the bearing capacity of shallow foundations. This theory provides a method to calculate the ultimate bearing capacity of soil, which is the maximum load per unit area that the soil can support without undergoing shear failure. The Calculation HTML Code included.
Sachpazis:Terzaghi Bearing Capacity Estimation in simple terms with Calculati...
A Fast Hadamard Transform for Signals with Sub-linear Sparsity
1. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
A Fast Hadamard Transform for Signals with
Sub-linear Sparsity
Robin Scheibler Saeid Haghighatshoar Martin Vetterli
School of Computer and Communication Sciences
École Polytechnique Fédérale de Lausanne, Switzerland
October 28, 2013
SparseFHT 1 / 20 EPFL
2. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Why the Hadamard transform ?
I Historically, low computation
approximation to DFT.
I Coding, 1969 Mariner Mars probe.
I Communication, orthogonal codes
in WCDMA.
I Compressed sensing, maximally
incoherent with Dirac basis.
I Spectroscopy, design of
instruments with lower noise.
I Recent advances in sparse FFT.
16 ⇥ 16 Hadamard matrix
Mariner probe
SparseFHT 2 / 20 EPFL
3. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Fast Hadamard transform
I Butterfly structure similar to FFT.
I Time complexity O(N log2 N).
I Sample complexity N.
+ Universal, i.e. works for all signals.
Does not exploit signal structure
(e.g. sparsity).
Can we do better ?
SparseFHT 3 / 20 EPFL
4. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Fast Hadamard transform
I Butterfly structure similar to FFT.
I Time complexity O(N log2 N).
I Sample complexity N.
+ Universal, i.e. works for all signals.
Does not exploit signal structure
(e.g. sparsity).
Can we do better ?
SparseFHT 3 / 20 EPFL
5. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Fast Hadamard transform
I Butterfly structure similar to FFT.
I Time complexity O(N log2 N).
I Sample complexity N.
+ Universal, i.e. works for all signals.
Does not exploit signal structure
(e.g. sparsity).
Can we do better ?
SparseFHT 3 / 20 EPFL
6. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Fast Hadamard transform
I Butterfly structure similar to FFT.
I Time complexity O(N log2 N).
I Sample complexity N.
+ Universal, i.e. works for all signals.
Does not exploit signal structure
(e.g. sparsity).
Can we do better ?
SparseFHT 3 / 20 EPFL
7. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Contribution: Sparse fast Hadamard transform
Assumptions
I The signal is exaclty K-sparse in the transform domain.
I Sub-linear sparsity regime K = O(N↵), 0 < ↵ < 1.
I Support of the signal is uniformly random.
Contribution
An algorithm computing the K non-zero coefficients with:
I Time complexity O(K log2 K log2
N
K ).
I Sample complexity O(K log2
N
K ).
I Probability of failure asymptotically vanishes.
SparseFHT 4 / 20 EPFL
8. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Contribution: Sparse fast Hadamard transform
Assumptions
I The signal is exaclty K-sparse in the transform domain.
I Sub-linear sparsity regime K = O(N↵), 0 < ↵ < 1.
I Support of the signal is uniformly random.
Contribution
An algorithm computing the K non-zero coefficients with:
I Time complexity O(K log2 K log2
N
K ).
I Sample complexity O(K log2
N
K ).
I Probability of failure asymptotically vanishes.
SparseFHT 4 / 20 EPFL
9. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Outline
1. Sparse FHT algorithm
2. Analysis of probability of failure
3. Empirical results
SparseFHT 5 / 20 EPFL
10. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Another look at the Hadamard transform
I Consider indices of x 2 RN,
N = 2n.
I Take the binary expansion of
indices.
I Represent signal on hypercube.
I Take DFT in every direction.
I = {0, . . . , 23
1}
SparseFHT 6 / 20 EPFL
11. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Another look at the Hadamard transform
I Consider indices of x 2 RN,
N = 2n.
I Take the binary expansion of
indices.
I Represent signal on hypercube.
I Take DFT in every direction.
I = {(0, 0, 0), . . . , (1, 1, 1)}
SparseFHT 6 / 20 EPFL
12. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Another look at the Hadamard transform
I Consider indices of x 2 RN,
N = 2n.
I Take the binary expansion of
indices.
I Represent signal on hypercube.
I Take DFT in every direction.
(0,0,0) (0,1,0)
(1,0,1) (1,1,1)
(1,1,0)(1,0,0)
(0,0,1) (0,1,1)
SparseFHT 6 / 20 EPFL
13. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Another look at the Hadamard transform
I Consider indices of x 2 RN,
N = 2n.
I Take the binary expansion of
indices.
I Represent signal on hypercube.
I Take DFT in every direction.
(0,0,0) (0,1,0)
(1,0,1) (1,1,1)
(1,1,0)(1,0,0)
(0,0,1) (0,1,1)
SparseFHT 6 / 20 EPFL
14. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Another look at the Hadamard transform
I Consider indices of x 2 RN,
N = 2n.
I Take the binary expansion of
indices.
I Represent signal on hypercube.
I Take DFT in every direction.
(0,0,0) (0,1,0)
(1,0,1) (1,1,1)
(1,1,0)(1,0,0)
(0,0,1) (0,1,1)
SparseFHT 6 / 20 EPFL
15. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Another look at the Hadamard transform
I Consider indices of x 2 RN,
N = 2n.
I Take the binary expansion of
indices.
I Represent signal on hypercube.
I Take DFT in every direction.
(0,0,0) (0,1,0)
(1,0,1) (1,1,1)
(1,1,0)(1,0,0)
(0,0,1) (0,1,1)
SparseFHT 6 / 20 EPFL
16. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Another look at the Hadamard transform
I Consider indices of x 2 RN,
N = 2n.
I Take the binary expansion of
indices.
I Represent signal on hypercube.
I Take DFT in every direction.
(0,0,0) (0,1,0)
(1,0,1) (1,1,1)
(1,1,0)(1,0,0)
(0,0,1) (0,1,1)
Xk0,...,kn 1
=
1X
m0=0
· · ·
1X
mn 1=0
( 1)k0m0+···+kn 1mn 1
xm0,...,mn 1
,
SparseFHT 6 / 20 EPFL
17. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Another look at the Hadamard transform
I Consider indices of x 2 RN,
N = 2n.
I Take the binary expansion of
indices.
I Represent signal on hypercube.
I Take DFT in every direction.
(0,0,0) (0,1,0)
(1,0,1) (1,1,1)
(1,1,0)(1,0,0)
(0,0,1) (0,1,1)
Xk =
X
m2Fn
2
( 1)hk , mi
xm, k, m 2 Fn
2, hk , mi =
n 1X
i=0
ki mi.
Treat indices as binary vectors.
SparseFHT 6 / 20 EPFL
18. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Another look at the Hadamard transform
I Consider indices of x 2 RN,
N = 2n.
I Take the binary expansion of
indices.
I Represent signal on hypercube.
I Take DFT in every direction.
(0,0,0) (0,1,0)
(1,0,1) (1,1,1)
(1,1,0)(1,0,0)
(0,0,1) (0,1,1)
Xk =
X
m2Fn
2
( 1)hk , mi
xm, k, m 2 Fn
2, hk , mi =
n 1X
i=0
ki mi.
Treat indices as binary vectors.
SparseFHT 6 / 20 EPFL
19. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Hadamard property I: downsampling/aliasing
Given B = 2b, a divider of N = 2n, and H 2 Fb⇥n
2 ,
where rows of H are a subset of rows of identity matrix,
xHT m
WHT
!
X
i2N(H)
XHT k+i, m, k 2 Fb
2.
e.g. H =
⇥
0b⇥(n b) Ib
⇤
selects the b high order bits.
SparseFHT 7 / 20 EPFL
20. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Hadamard property I: downsampling/aliasing
Given B = 2b, a divider of N = 2n, and H 2 Fb⇥n
2 ,
where rows of H are a subset of rows of identity matrix,
xHT m
WHT
!
X
i2N(H)
XHT k+i, m, k 2 Fb
2.
e.g. H =
⇥
0b⇥(n b) Ib
⇤
selects the b high order bits.
SparseFHT 7 / 20 EPFL
21. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Hadamard property I: downsampling/aliasing
Given B = 2b, a divider of N = 2n, and H 2 Fb⇥n
2 ,
where rows of H are a subset of rows of identity matrix,
xHT m
WHT
!
X
i2N(H)
XHT k+i, m, k 2 Fb
2.
e.g. H =
⇥
0b⇥(n b) Ib
⇤
selects the b high order bits.
(0,0,0) (0,1,0)
(1,0,1)
(1,1,1)(0,1,1)(0,0,1)
(1,0,0) (1,1,0)
(0,0,0) (0,1,0)
(1,0,1) (1,1,1)
(0,1,1)
(0,0,1)
(1,0,0) (1,1,0)
WHT
SparseFHT 7 / 20 EPFL
22. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Hadamard property I: downsampling/aliasing
Given B = 2b, a divider of N = 2n, and H 2 Fb⇥n
2 ,
where rows of H are a subset of rows of identity matrix,
xHT m
WHT
!
X
i2N(H)
XHT k+i, m, k 2 Fb
2.
e.g. H =
⇥
0b⇥(n b) Ib
⇤
selects the b high order bits.
(0,0,0) (0,1,0)
(0,1,1)(0,0,1)
(0,0,0) (0,1,0)
(1,0,1) (1,1,1)
(0,1,1)
(0,0,1)
(1,0,0) (1,1,0)
(1,0,1)
(1,1,1)
(1,0,0) (1,1,0)
WHT
SparseFHT 7 / 20 EPFL
23. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Aliasing induced bipartite graph
time-domain
Hadamard-domain
I Downsampling induces an aliasing pattern.
I Different downsamplings produce different patterns.
SparseFHT 8 / 20 EPFL
24. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Aliasing induced bipartite graph
time-domain
Hadamard-domain
4-WHT
I Downsampling induces an aliasing pattern.
I Different downsamplings produce different patterns.
SparseFHT 8 / 20 EPFL
25. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Aliasing induced bipartite graph
time-domain
Hadamard-domain
4-WHT
I Downsampling induces an aliasing pattern.
I Different downsamplings produce different patterns.
SparseFHT 8 / 20 EPFL
26. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Aliasing induced bipartite graph
time-domain
Hadamard-domain
4-WHT 4-WHT
I Downsampling induces an aliasing pattern.
I Different downsamplings produce different patterns.
SparseFHT 8 / 20 EPFL
27. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Aliasing induced bipartite graph
time-domain
Hadamard-domain
4-WHT 4-WHT
I Downsampling induces an aliasing pattern.
I Different downsamplings produce different patterns.
SparseFHT 8 / 20 EPFL
28. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Aliasing induced bipartite graph
time-domain
Hadamard-domain
4-WHT 4-WHT
Checks
Variables
I Downsampling induces an aliasing pattern.
I Different downsamplings produce different patterns.
SparseFHT 8 / 20 EPFL
29. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Genie-aided peeling decoder
A genie indicates us
I if a check is connected to only one variable (singleton),
I in that case, the genie also gives the index of that variable.
Peeling decoder algorithm:
1. Find a singleton check: {X1, X8, X11}
2. Peel it off.
3. Repeat until nothing left.
SparseFHT 9 / 20 EPFL
30. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Genie-aided peeling decoder
A genie indicates us
I if a check is connected to only one variable (singleton),
I in that case, the genie also gives the index of that variable.
Peeling decoder algorithm:
1. Find a singleton check: {X1, X8, X11}
2. Peel it off.
3. Repeat until nothing left.
SparseFHT 9 / 20 EPFL
31. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Genie-aided peeling decoder
A genie indicates us
I if a check is connected to only one variable (singleton),
I in that case, the genie also gives the index of that variable.
Peeling decoder algorithm:
1. Find a singleton check: {X1, X8, X11}
2. Peel it off.
3. Repeat until nothing left.
SparseFHT 9 / 20 EPFL
32. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Genie-aided peeling decoder
A genie indicates us
I if a check is connected to only one variable (singleton),
I in that case, the genie also gives the index of that variable.
Peeling decoder algorithm:
1. Find a singleton check: {X1, X8, X11}
2. Peel it off.
3. Repeat until nothing left.
SparseFHT 9 / 20 EPFL
33. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Genie-aided peeling decoder
A genie indicates us
I if a check is connected to only one variable (singleton),
I in that case, the genie also gives the index of that variable.
Peeling decoder algorithm:
1. Find a singleton check: {X1, X8, X11}
2. Peel it off.
3. Repeat until nothing left.
SparseFHT 9 / 20 EPFL
34. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Genie-aided peeling decoder
A genie indicates us
I if a check is connected to only one variable (singleton),
I in that case, the genie also gives the index of that variable.
Peeling decoder algorithm:
1. Find a singleton check: {X1, X8, X11}
2. Peel it off.
3. Repeat until nothing left.
SparseFHT 9 / 20 EPFL
35. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Genie-aided peeling decoder
A genie indicates us
I if a check is connected to only one variable (singleton),
I in that case, the genie also gives the index of that variable.
Success
Peeling decoder algorithm:
1. Find a singleton check: {X1, X8, X11}
2. Peel it off.
3. Repeat until nothing left.
SparseFHT 9 / 20 EPFL
36. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Hadamard property II: shift/modulation
Theorem (shift/modulation)
Given p 2 Fn
2,
xm+p
WHT
! Xk ( 1)hp , ki
.
Consequence
The signal can be modulated in frequency by manipulating the
time-domain samples.
SparseFHT 10 / 20 EPFL
37. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Hadamard property II: shift/modulation
Theorem (shift/modulation)
Given p 2 Fn
2,
xm+p
WHT
! Xk ( 1)hp , ki
.
Consequence
The signal can be modulated in frequency by manipulating the
time-domain samples.
SparseFHT 10 / 20 EPFL
38. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
How to construct the Genie
Non-modulated Modulated
I Collision: if 2 variables connected to same check.
I
Xi ( 1)hp , ii+Xj ( 1)hp , ji
Xi +Xj
6= ±1, (mild assumption on distribution of X).
SparseFHT 11 / 20 EPFL
39. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
How to construct the Genie
Non-modulated Modulated
I Collision: if 2 variables connected to same check.
I
Xi ( 1)hp , ii+Xj ( 1)hp , ji
Xi +Xj
6= ±1, (mild assumption on distribution of X).
SparseFHT 11 / 20 EPFL
40. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
How to construct the Genie
Non-modulated Modulated
I Collision: if 2 variables connected to same check.
I
Xi ( 1)hp , ii+Xj ( 1)hp , ji
Xi +Xj
6= ±1, (mild assumption on distribution of X).
SparseFHT 11 / 20 EPFL
41. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
How to construct the Genie
Non-modulated Modulated
I Singleton: only one variable connected to check.
I
Xi ( 1)hp , ii
Xi
= ( 1)hp , ii = ±1. We can know hp , ii!
I O(log2
N
K ) measurements sufficient to recover index i,
(dimension of null-space of downsampling matrix H).
SparseFHT 11 / 20 EPFL
42. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
How to construct the Genie
Non-modulated Modulated
I Singleton: only one variable connected to check.
I
Xi ( 1)hp , ii
Xi
= ( 1)hp , ii = ±1. We can know hp , ii!
I O(log2
N
K ) measurements sufficient to recover index i,
(dimension of null-space of downsampling matrix H).
SparseFHT 11 / 20 EPFL
43. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
How to construct the Genie
Non-modulated Modulated
I Singleton: only one variable connected to check.
I
Xi ( 1)hp , ii
Xi
= ( 1)hp , ii = ±1. We can know hp , ii!
I O(log2
N
K ) measurements sufficient to recover index i,
(dimension of null-space of downsampling matrix H).
SparseFHT 11 / 20 EPFL
44. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
How to construct the Genie
Non-modulated Modulated
I Singleton: only one variable connected to check.
I
Xi ( 1)hp , ii
Xi
= ( 1)hp , ii = ±1. We can know hp , ii!
I O(log2
N
K ) measurements sufficient to recover index i,
(dimension of null-space of downsampling matrix H).
SparseFHT 11 / 20 EPFL
45. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Sparse fast Hadamard transform
Algorithm
1. Set number of checks per downsampling B = O(K).
2. Choose C downsampling matrices H1, . . . , HC.
3. Compute C(log2 N/K + 1) size-K fast Hadamard
transform, each takes O(K log2 K).
4. Decode non-zero coefficients using peeling decoder.
Performance
I Time complexity – O(K log2 K log2 N/K).
I Sample complexity – O(K log2
N
K ).
I How to construct H1, . . . , HC ? Probability of success ?
SparseFHT 12 / 20 EPFL
46. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Sparse fast Hadamard transform
Algorithm
1. Set number of checks per downsampling B = O(K).
2. Choose C downsampling matrices H1, . . . , HC.
3. Compute C(log2 N/K + 1) size-K fast Hadamard
transform, each takes O(K log2 K).
4. Decode non-zero coefficients using peeling decoder.
Performance
I Time complexity – O(K log2 K log2 N/K).
I Sample complexity – O(K log2
N
K ).
I How to construct H1, . . . , HC ? Probability of success ?
SparseFHT 12 / 20 EPFL
47. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Sparse fast Hadamard transform
Algorithm
1. Set number of checks per downsampling B = O(K).
2. Choose C downsampling matrices H1, . . . , HC.
3. Compute C(log2 N/K + 1) size-K fast Hadamard
transform, each takes O(K log2 K).
4. Decode non-zero coefficients using peeling decoder.
Performance
I Time complexity – O(K log2 K log2 N/K).
I Sample complexity – O(K log2
N
K ).
I How to construct H1, . . . , HC ? Probability of success ?
SparseFHT 12 / 20 EPFL
48. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Very sparse regime
Setting
I K = O(N↵), 0 < ↵ < 1/3.
I Uniformly random support.
I Study asymptotic probability of failure as n ! 1.
Downsampling matrices construction
I Achieves values ↵ = 1
C , i.e. b = n
C .
I Deterministic downsampling matrices H1, . . . , HC,
SparseFHT 13 / 20 EPFL
49. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Very sparse regime
Setting
I K = O(N↵), 0 < ↵ < 1/3.
I Uniformly random support.
I Study asymptotic probability of failure as n ! 1.
Downsampling matrices construction
I Achieves values ↵ = 1
C , i.e. b = n
C .
I Deterministic downsampling matrices H1, . . . , HC,
SparseFHT 13 / 20 EPFL
50. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Very sparse regime
Setting
I K = O(N↵), 0 < ↵ < 1/3.
I Uniformly random support.
I Study asymptotic probability of failure as n ! 1.
Downsampling matrices construction
I Achieves values ↵ = 1
C , i.e. b = n
C .
I Deterministic downsampling matrices H1, . . . , HC,
SparseFHT 13 / 20 EPFL
51. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Very sparse regime
Setting
I K = O(N↵), 0 < ↵ < 1/3.
I Uniformly random support.
I Study asymptotic probability of failure as n ! 1.
Downsampling matrices construction
I Achieves values ↵ = 1
C , i.e. b = n
C .
I Deterministic downsampling matrices H1, . . . , HC,
SparseFHT 13 / 20 EPFL
52. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Very sparse regime
Setting
I K = O(N↵), 0 < ↵ < 1/3.
I Uniformly random support.
I Study asymptotic probability of failure as n ! 1.
Downsampling matrices construction
I Achieves values ↵ = 1
C , i.e. b = n
C .
I Deterministic downsampling matrices H1, . . . , HC,
SparseFHT 13 / 20 EPFL
53. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Very sparse regime
Setting
I K = O(N↵), 0 < ↵ < 1/3.
I Uniformly random support.
I Study asymptotic probability of failure as n ! 1.
Downsampling matrices construction
I Achieves values ↵ = 1
C , i.e. b = n
C .
I Deterministic downsampling matrices H1, . . . , HC,
SparseFHT 13 / 20 EPFL
54. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Uniformly random support model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
55. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Uniformly random support model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
56. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Uniformly random support model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
57. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Uniformly random support model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
58. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Uniformly random support model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
59. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Balls-and-bins model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
60. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Balls-and-bins model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
61. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Balls-and-bins model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
62. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Balls-and-bins model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
63. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Balls-and-bins model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
64. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Balls-and-bins model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
65. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Balls-and-bins model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
66. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Balls-and-bins model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
67. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Balls-and-bins model
Balls-and-bins model
I Theorem: Both constructions are equivalent.
Proof: By construction, all rows of Hi are linearly independent.
I Reduces to LDPC decoding analysis.
I Error correcting code design (Luby et al. 2001).
I FFAST (Sparse FFT algorithm) (Pawar & Ramchandran 2013).
SparseFHT 14 / 20 EPFL
68. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Extension to less-sparse regime
I K = O(N↵), 2/3 ↵ < 1.
I Balls-and-bins model not equivalent anymore.
I Let ↵ = 1 1
C . Construct H1, . . . , HC,
I By construction: N(Hi)
T
N(Hj) = 0.
SparseFHT 15 / 20 EPFL
69. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Extension to less-sparse regime
I K = O(N↵), 2/3 ↵ < 1.
I Balls-and-bins model not equivalent anymore.
I Let ↵ = 1 1
C . Construct H1, . . . , HC,
I By construction: N(Hi)
T
N(Hj) = 0.
SparseFHT 15 / 20 EPFL
70. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Extension to less-sparse regime
I K = O(N↵), 2/3 ↵ < 1.
I Balls-and-bins model not equivalent anymore.
I Let ↵ = 1 1
C . Construct H1, . . . , HC,
I By construction: N(Hi)
T
N(Hj) = 0.
SparseFHT 15 / 20 EPFL
71. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Extension to less-sparse regime
I K = O(N↵), 2/3 ↵ < 1.
I Balls-and-bins model not equivalent anymore.
I Let ↵ = 1 1
C . Construct H1, . . . , HC,
I By construction: N(Hi)
T
N(Hj) = 0.
SparseFHT 15 / 20 EPFL
72. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Extension to less-sparse regime
I K = O(N↵), 2/3 ↵ < 1.
I Balls-and-bins model not equivalent anymore.
I Let ↵ = 1 1
C . Construct H1, . . . , HC,
I By construction: N(Hi)
T
N(Hj) = 0.
SparseFHT 15 / 20 EPFL
73. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Extension to less-sparse regime
I K = O(N↵), 2/3 ↵ < 1.
I Balls-and-bins model not equivalent anymore.
I Let ↵ = 1 1
C . Construct H1, . . . , HC,
I By construction: N(Hi)
T
N(Hj) = 0.
SparseFHT 15 / 20 EPFL
74. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Extension to less-sparse regime
I K = O(N↵), 2/3 ↵ < 1.
I Balls-and-bins model not equivalent anymore.
I Let ↵ = 1 1
C . Construct H1, . . . , HC,
I By construction: N(Hi)
T
N(Hj) = 0.
SparseFHT 15 / 20 EPFL
75. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Extension to less-sparse regime
I K = O(N↵), 2/3 ↵ < 1.
I Balls-and-bins model not equivalent anymore.
I Let ↵ = 1 1
C . Construct H1, . . . , HC,
I By construction: N(Hi)
T
N(Hj) = 0.
SparseFHT 15 / 20 EPFL
76. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Extension to less-sparse regime
I K = O(N↵), 2/3 ↵ < 1.
I Balls-and-bins model not equivalent anymore.
I Let ↵ = 1 1
C . Construct H1, . . . , HC,
I By construction: N(Hi)
T
N(Hj) = 0.
SparseFHT 15 / 20 EPFL
77. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Extension to less-sparse regime
I K = O(N↵), 2/3 ↵ < 1.
I Balls-and-bins model not equivalent anymore.
I Let ↵ = 1 1
C . Construct H1, . . . , HC,
I By construction: N(Hi)
T
N(Hj) = 0.
SparseFHT 15 / 20 EPFL
78. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
SparseFHT – Probability of success
Probability of success - N = 222
0 1/3 2/3 1
0
0.2
0.4
0.6
0.8
1
↵
SparseFHT 16 / 20 EPFL
79. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
SparseFHT vs. FHT
Runtime [µs] – N = 215
0 1/3 2/3 1
0
200
400
600
800
1000
Sparse FHT
FHT
↵
SparseFHT 17 / 20 EPFL
80. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Conclusion
Contribution
I Sparse fast Hadamard algorithm.
I Time complexity O(K log2 K log2
N
K ).
I Sample complexity O(K log2
N
K ).
I Probability of success asymptotically equal to 1.
What’s next ?
I Investigate noisy case.
SparseFHT 18 / 20 EPFL
81. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Conclusion
Contribution
I Sparse fast Hadamard algorithm.
I Time complexity O(K log2 K log2
N
K ).
I Sample complexity O(K log2
N
K ).
I Probability of success asymptotically equal to 1.
What’s next ?
I Investigate noisy case.
SparseFHT 18 / 20 EPFL
82. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Thanks for your attention!
Code and figures available at
http://lcav.epfl.ch/page-99903.html
SparseFHT 19 / 20 EPFL
83. Introduction Sparse FHT algorithm Analysis of probability of failure Empirical results Conclusion
Reference
[1] M.G. Luby, M. Mitzenmacher, M.A. Shokrollahi, and
D.A. Spielman,
Efficient erasure correcting codes,
IEEE Trans. Inform. Theory, vol. 47, no. 2,
pp. 569–584, 2001.
[2] S. Pawar and K. Ramchandran,
Computing a k-sparse n-length discrete Fourier transform
using at most 4k samples and O(k log k) complexity,
arXiv.org, vol. cs.DS. 04-May-2013.
SparseFHT 20 / 20 EPFL