The document discusses Hermite integrators and Riordan arrays. It provides:
1) An overview of a general form for the correctors of a family of 2-step Hermite integrators that achieve 2(p+1)-th order accuracy by directly calculating up to the p-th order derivative of the force.
2) Details on constructing Hermite integrators by solving a linear equation to determine the coefficients, with an example of a 6th order integrator.
3) An outline of the proof for the general form of the coefficients, which involves setting up a differential recurrence relation and solving a linear system using LU decomposition, with the proof of the inverse matrices later shown using Riordan arrays.
Using blurred images to assess damage in bridge structures?Alessandro Palmeri
Faster trains and augmented traffic have significantly increased the number and amplitude of loading cycles experienced on a daily basis by composite steel-concrete bridges. This higher demand accelerates the occurrence of damage in the shear connectors between the two materials, which in turn can severely affect performance and reliability of these structures. The aim of this talk is to present the preliminary results of theoretical and experimental investigations undertaken to assess the feasibility of using the envelope of deflections and rotations induced by moving loads as a practical and cost-effective alternative to traditional methods of health monitoring for composite bridges. Both analytical and numerical formulations for this dynamic problem are presented and the results of a parametric study are discussed. A novel photogrammetric approach is also introduced, which allows identifying vibration patterns in civil engineering structures by analysing blurred targets in long-exposure digital images. The initial experimental validation of this approach is presented and further challenges are highlighted.
On maximal and variational Fourier restrictionVjekoslavKovac1
Workshop talk slides, Follow-up workshop to trimester program "Harmonic Analysis and Partial Differential Equations", Hausdorff Institute, Bonn, May 2019.
Second part of Matrices at undergraduate in science (math, physics, engineering) level.
Please send comments and suggestions to solo.hermelin@gmail.com.
For more presentations visit my website at
http://www.solohermelin.com.
Response Surface in Tensor Train format for Uncertainty QuantificationAlexander Litvinenko
We apply low-rank Tensor Train format to solve PDEs with uncertain coefficients. First, we approximate uncertain permeability coefficient in TT format, then the operator and then apply iterations to solve stochastic Galerkin system.
The lattice Boltzmann equation: background, boundary conditions, and Burnett-...Tim Reis
An overview of the lattice Boltzmann equation and a discussion of moment-based boundary conditions. Includes applications to the slip flow regime and the Burnett stress. Some analysis sheds insight into the physical and numerical behaviour of the algorithm.
Gamma Function mathematics and history.
Please send comments and suggestions for improvements to solo.hermelin@gmail.com. Thanks.
More presentations on different subjects can be found on my website at http://www.solohermelin.com.
Using blurred images to assess damage in bridge structures?Alessandro Palmeri
Faster trains and augmented traffic have significantly increased the number and amplitude of loading cycles experienced on a daily basis by composite steel-concrete bridges. This higher demand accelerates the occurrence of damage in the shear connectors between the two materials, which in turn can severely affect performance and reliability of these structures. The aim of this talk is to present the preliminary results of theoretical and experimental investigations undertaken to assess the feasibility of using the envelope of deflections and rotations induced by moving loads as a practical and cost-effective alternative to traditional methods of health monitoring for composite bridges. Both analytical and numerical formulations for this dynamic problem are presented and the results of a parametric study are discussed. A novel photogrammetric approach is also introduced, which allows identifying vibration patterns in civil engineering structures by analysing blurred targets in long-exposure digital images. The initial experimental validation of this approach is presented and further challenges are highlighted.
On maximal and variational Fourier restrictionVjekoslavKovac1
Workshop talk slides, Follow-up workshop to trimester program "Harmonic Analysis and Partial Differential Equations", Hausdorff Institute, Bonn, May 2019.
Second part of Matrices at undergraduate in science (math, physics, engineering) level.
Please send comments and suggestions to solo.hermelin@gmail.com.
For more presentations visit my website at
http://www.solohermelin.com.
Response Surface in Tensor Train format for Uncertainty QuantificationAlexander Litvinenko
We apply low-rank Tensor Train format to solve PDEs with uncertain coefficients. First, we approximate uncertain permeability coefficient in TT format, then the operator and then apply iterations to solve stochastic Galerkin system.
The lattice Boltzmann equation: background, boundary conditions, and Burnett-...Tim Reis
An overview of the lattice Boltzmann equation and a discussion of moment-based boundary conditions. Includes applications to the slip flow regime and the Burnett stress. Some analysis sheds insight into the physical and numerical behaviour of the algorithm.
Gamma Function mathematics and history.
Please send comments and suggestions for improvements to solo.hermelin@gmail.com. Thanks.
More presentations on different subjects can be found on my website at http://www.solohermelin.com.
MATHS SYMBOLS - OTHER OPERATIONS (2) - SUMMATION - Pi (PRODUCT) - FACTORIAL - SUM of FACTORIALS - SUBTRACTION of FACTORIALS - PRODUCT of FACTORIALS - DIVISION of FACTORIALS - POWER of FACTORIALS - PERMUTATIONS - POSSIBILITIES of COMBINATIONS - COMBINATIONS - BINOMIAL COEFFICIENT - BINOMIAL FORMULA - TETRATION - PENTATION - EXAMPLES and CALCULATIONS STEP by STEP
Find the compact trigonometric Fourier series for the periodic signal.pdfarihantelectronics
Find the compact trigonometric Fourier series for the periodic signal x(t) and sketch the
amplitude and phase spectrum for first 4 frequency components. By inspection of the spectra,
sketch the exponential Fourier spectra. By inspection of spectra in part b), write the exponential
Fourier series for x(t)
Solution
ECE 3640 Lecture 4 – Fourier series: expansions of periodic functions. Objective: To build upon
the ideas from the previous lecture to learn about Fourier series, which are series representations
of periodic functions. Periodic signals and representations From the last lecture we learned how
functions can be represented as a series of other functions: f(t) = Xn k=1 ckik(t). We discussed
how certain classes of things can be built using certain kinds of basis functions. In this lecture we
will consider specifically functions that are periodic, and basic functions which are
trigonometric. Then the series is said to be a Fourier series. A signal f(t) is said to be periodic
with period T0 if f(t) = f(t + T0) for all t. Diagram on board. Note that this must be an everlasting
signal. Also note that, if we know one period of the signal we can find the rest of it by periodic
extension. The integral over a single period of the function is denoted by Z T0 f(t)dt. When
integrating over one period of a periodic function, it does not matter when we start. Usually it is
convenient to start at the beginning of a period. The building block functions that can be used to
build up periodic functions are themselves periodic: we will use the set of sinusoids. If the period
of f(t) is T0, let 0 = 2/T0. The frequency 0 is said to be the fundamental frequency; the
fundamental frequency is related to the period of the function. Furthermore, let F0 = 1/T0. We
will represent the function f(t) using the set of sinusoids i0(t) = cos(0t) = 1 i1(t) = cos(0t + 1)
i2(t) = cos(20t + 2) . . . Then, f(t) = C0 + X n=1 Cn cos(n0t + n) The frequency n0 is said to be
the nth harmonic of 0. Note that for each basis function associated with f(t) there are actually two
parameters: the amplitude Cn and the phase n. It will often turn out to be more useful to
represent the function using both sines and cosines. Note that we can write Cn cos(n0t + n) = Cn
cos(n) cos(n0t) Cn sin(n)sin(n0t). ECE 3640: Lecture 4 – Fourier series: expansions of periodic
functions. 2 Now let an = Cn cos n bn = Cn sin n Then Cn cos(n0t + n) = an cos(n0t) + bn
sin(n0t) Then the series representation can be f(t) = C0 + X n=1 Cn cos(n0t + n) = a0 + X n=1 an
cos(n0t) + bn sin(n0t) The first of these is the compact trigonometric Fourier series. The second
is the trigonometric Fourier series.. To go from one to the other use C0 = a0 Cn = p a 2 n + b 2 n
n = tan1 (bn/an). To complete the representation we must be able to compute the coefficients.
But this is the same sort of thing we did before. If we can show that the set of functions
{cos(n0t),sin(n0t)} is in fact an orthogonal set, then we can use the same.
Top Features to Include in Your Winzo Clone App for Business Growth (4).pptxrickgrimesss22
Discover the essential features to incorporate in your Winzo clone app to boost business growth, enhance user engagement, and drive revenue. Learn how to create a compelling gaming experience that stands out in the competitive market.
Gamify Your Mind; The Secret Sauce to Delivering Success, Continuously Improv...Shahin Sheidaei
Games are powerful teaching tools, fostering hands-on engagement and fun. But they require careful consideration to succeed. Join me to explore factors in running and selecting games, ensuring they serve as effective teaching tools. Learn to maintain focus on learning objectives while playing, and how to measure the ROI of gaming in education. Discover strategies for pitching gaming to leadership. This session offers insights, tips, and examples for coaches, team leads, and enterprise leaders seeking to teach from simple to complex concepts.
In 2015, I used to write extensions for Joomla, WordPress, phpBB3, etc and I ...Juraj Vysvader
In 2015, I used to write extensions for Joomla, WordPress, phpBB3, etc and I didn't get rich from it but it did have 63K downloads (powered possible tens of thousands of websites).
In the ever-evolving landscape of technology, enterprise software development is undergoing a significant transformation. Traditional coding methods are being challenged by innovative no-code solutions, which promise to streamline and democratize the software development process.
This shift is particularly impactful for enterprises, which require robust, scalable, and efficient software to manage their operations. In this article, we will explore the various facets of enterprise software development with no-code solutions, examining their benefits, challenges, and the future potential they hold.
Launch Your Streaming Platforms in MinutesRoshan Dwivedi
The claim of launching a streaming platform in minutes might be a bit of an exaggeration, but there are services that can significantly streamline the process. Here's a breakdown:
Pros of Speedy Streaming Platform Launch Services:
No coding required: These services often use drag-and-drop interfaces or pre-built templates, eliminating the need for programming knowledge.
Faster setup: Compared to building from scratch, these platforms can get you up and running much quicker.
All-in-one solutions: Many services offer features like content management systems (CMS), video players, and monetization tools, reducing the need for multiple integrations.
Things to Consider:
Limited customization: These platforms may offer less flexibility in design and functionality compared to custom-built solutions.
Scalability: As your audience grows, you might need to upgrade to a more robust platform or encounter limitations with the "quick launch" option.
Features: Carefully evaluate which features are included and if they meet your specific needs (e.g., live streaming, subscription options).
Examples of Services for Launching Streaming Platforms:
Muvi [muvi com]
Uscreen [usencreen tv]
Alternatives to Consider:
Existing Streaming platforms: Platforms like YouTube or Twitch might be suitable for basic streaming needs, though monetization options might be limited.
Custom Development: While more time-consuming, custom development offers the most control and flexibility for your platform.
Overall, launching a streaming platform in minutes might not be entirely realistic, but these services can significantly speed up the process compared to building from scratch. Carefully consider your needs and budget when choosing the best option for you.
Software Engineering, Software Consulting, Tech Lead, Spring Boot, Spring Cloud, Spring Core, Spring JDBC, Spring Transaction, Spring MVC, OpenShift Cloud Platform, Kafka, REST, SOAP, LLD & HLD.
Unleash Unlimited Potential with One-Time Purchase
BoxLang is more than just a language; it's a community. By choosing a Visionary License, you're not just investing in your success, you're actively contributing to the ongoing development and support of BoxLang.
Globus Compute wth IRI Workflows - GlobusWorld 2024Globus
As part of the DOE Integrated Research Infrastructure (IRI) program, NERSC at Lawrence Berkeley National Lab and ALCF at Argonne National Lab are working closely with General Atomics on accelerating the computing requirements of the DIII-D experiment. As part of the work the team is investigating ways to speedup the time to solution for many different parts of the DIII-D workflow including how they run jobs on HPC systems. One of these routes is looking at Globus Compute as a way to replace the current method for managing tasks and we describe a brief proof of concept showing how Globus Compute could help to schedule jobs and be a tool to connect compute at different facilities.
First Steps with Globus Compute Multi-User EndpointsGlobus
In this presentation we will share our experiences around getting started with the Globus Compute multi-user endpoint. Working with the Pharmacology group at the University of Auckland, we have previously written an application using Globus Compute that can offload computationally expensive steps in the researcher's workflows, which they wish to manage from their familiar Windows environments, onto the NeSI (New Zealand eScience Infrastructure) cluster. Some of the challenges we have encountered were that each researcher had to set up and manage their own single-user globus compute endpoint and that the workloads had varying resource requirements (CPUs, memory and wall time) between different runs. We hope that the multi-user endpoint will help to address these challenges and share an update on our progress here.
Globus Connect Server Deep Dive - GlobusWorld 2024Globus
We explore the Globus Connect Server (GCS) architecture and experiment with advanced configuration options and use cases. This content is targeted at system administrators who are familiar with GCS and currently operate—or are planning to operate—broader deployments at their institution.
We describe the deployment and use of Globus Compute for remote computation. This content is aimed at researchers who wish to compute on remote resources using a unified programming interface, as well as system administrators who will deploy and operate Globus Compute services on their research computing infrastructure.
Code reviews are vital for ensuring good code quality. They serve as one of our last lines of defense against bugs and subpar code reaching production.
Yet, they often turn into annoying tasks riddled with frustration, hostility, unclear feedback and lack of standards. How can we improve this crucial process?
In this session we will cover:
- The Art of Effective Code Reviews
- Streamlining the Review Process
- Elevating Reviews with Automated Tools
By the end of this presentation, you'll have the knowledge on how to organize and improve your code review proces
Software Engineering, Software Consulting, Tech Lead.
Spring Boot, Spring Cloud, Spring Core, Spring JDBC, Spring Security,
Spring Transaction, Spring MVC,
Log4j, REST/SOAP WEB-SERVICES.
Cyaniclab : Software Development Agency Portfolio.pdfCyanic lab
CyanicLab, an offshore custom software development company based in Sweden,India, Finland, is your go-to partner for startup development and innovative web design solutions. Our expert team specializes in crafting cutting-edge software tailored to meet the unique needs of startups and established enterprises alike. From conceptualization to execution, we offer comprehensive services including web and mobile app development, UI/UX design, and ongoing software maintenance. Ready to elevate your business? Contact CyanicLab today and let us propel your vision to success with our top-notch IT solutions.
In software engineering, the right architecture is essential for robust, scalable platforms. Wix has undergone a pivotal shift from event sourcing to a CRUD-based model for its microservices. This talk will chart the course of this pivotal journey.
Event sourcing, which records state changes as immutable events, provided robust auditing and "time travel" debugging for Wix Stores' microservices. Despite its benefits, the complexity it introduced in state management slowed development. Wix responded by adopting a simpler, unified CRUD model. This talk will explore the challenges of event sourcing and the advantages of Wix's new "CRUD on steroids" approach, which streamlines API integration and domain event management while preserving data integrity and system resilience.
Participants will gain valuable insights into Wix's strategies for ensuring atomicity in database updates and event production, as well as caching, materialization, and performance optimization techniques within a distributed system.
Join us to discover how Wix has mastered the art of balancing simplicity and extensibility, and learn how the re-adoption of the modest CRUD has turbocharged their development velocity, resilience, and scalability in a high-growth environment.
1. Hermite integrators and Riordan arrays
When one feels tired in round-off error
Keigo Nitadori
keigo@riken.jp
Co-Design Team, Flagship 2020 Project
RIKEN Advanced Institute for Computational Science
January 7, 2016
2. Today I talk about
A general form of the correctors for the family of 2-step Hermite
integrators
Up to p-th order derivative of the force is calculated directly to
obtain 2(p + 1)-th order accuracy
A mathematical proof in elementary algebra is presented
In my last talk in MODEST 15-S, a simple and beautiful proof
made by Satoko Yamamoto were mostly omitted, but today I
would like you to see the full story.
3. References
I appreciate for many useful online documents:
1. R. Sprugnoli (2006). An Introduction to Mathematical Methods
in Combinatorics.
http://www.dsi.unifi.it/~resp/Handbook.pdf
2. R. Sprugnoli (2006). Riordan Array Proofs of Identities in
Gould’s Book.
http://www.dsi.unifi.it/~resp/GouldBK.pdf
3. D. Merlini (2011). A survey on Riordan arrays.
https://lipn.univ-paris13.fr/~banderier/
Seminaires/Slides/merlini.pdf
4. P. Barry (2005) A Catalan Transform and Related
Transformations on Integer Sequences. Journal of Integer
Sequences 8. https://cs.uwaterloo.ca/journals/JIS/
VOL8/Barry/barry84.html
4. Polynomial shift by Pascal matrix I
We consider to shift a finite order polynomial f (t), of which up to the
p-th derivatives are f (n) (t) = dn
dtn f (t) (0 ≤ n ≤ p). Adjusting the
dimension of the derivatives by a step size h, a vector
F(t) =
f (t)
h f (1) (t)
h2 f (2) (t)/2!
...
hn f (p) (t)/n!
. (1)
obeys a differential equation
d
dt
F(t) =
1
h
0 1 0 · · · 0
0 0 2 · · · 0
0 0 0
...
...
...
...
...
... p
0 0 0 · · · 0
F(t). (2)
5. Polynomial shift by Pascal matrix II
A formal solution at t + h is,
F(t + h) = exp
0 1 0 · · · 0
0 0 2 · · · 0
0 0 0
...
...
...
...
...
... p
0 0 0 · · · 0
F(t)
=
0
0
1
0
2
0 · · · p
0
0 1
1
2
1 · · · p
1
0 0 2
2 · · · p
2
...
...
...
...
...
0 0 0 · · · p
p
F(t). (3)
This matrix is referred to as upper triangle Pascal matrix.
7. Construction of the Hermite integrators I
For simplicity, we integrate from t = −h to t = h, thus a timestep
becomes ∆t = 2h (not h).
The inputs are additions and subtractions
F+
n =
1
2
hn
n!
f (n)
(h) + f (n)
(−h) , (4)
F−
n =
1
2
hn
n!
f (n)
(h) − f (n)
(−h) , (5)
for 0 ≤ n ≤ p. The outputs are fitting polynomial at t = 0,
Fn =
hn
n!
f (n)
(0), (6)
for 0 ≤ n ≤ 2p + 1.
The integral includes only the even order terms:
∆v =
h
−h
f (t)dt = F0 +
1
3
F2 +
1
5
F4 + · · · +
1
2p + 1
F2p, (7)
for a 2(p + 1)-th order integrator.
8. Construction of the Hermite integrators II
A linear equation we want to solve is
F+
0
F−
1
F+
2
F−
3
...
=
0
0
2
0
4
0
6
0 . . .
0 2
1
4
1
6
1 . . .
0 2
2
4
2
6
2 . . .
0 0 4
3
6
3 . . .
...
...
...
...
...
F0
F2
F4
F6
...
, (8)
of size (p + 1). The element of matrix is 2j
i (counted from (0, 0)),
which means we have picked up the even number columns from the
Pascal matrix.
9. Construction of the Hermite integrators III
Example
By solving
F+
0
F−
1
F+
2
=
1 1 1
0 2 4
0 1 6
F0
F2
F4
, (9)
we have
F0
F2
F4
=
1
8
8 −5 2
0 6 −4
0 −1 2
F+
0
F−
1
F+
2
. (10)
The sixth-order integrator is thus
∆v =2 F0 +
1
3
F2 +
1
5
F4 h
= F+
0 −
2
5
F−
1 +
2
15
F+
2 ∆t (11)
11. General form
For the 2(p + 1)-th order integrator (p ≥ 0), the k-th term
(0 ≤ k ≤ p) is,
c(p)
k =
1
(−2)k
(2k)!!
(2k + 1)!!
p − k + m
p − k
(2k + 1)!!
(2k + 1 − 2m)!!
(2p + 1 − 2m)!!
(2p + 1)!!
,
(12)
with m = (k + 1)/2 . A recurrent form starting from the diagonal
element to the bottom is
c(p)
k =
1
(−2)p
(2p)!!
(2p + 1)!!
(p = k)
p − k + m
p − k
2p + 1 − 2m
2p + 1
c(p−1)
k (p > k)
. (13)
It is a rational recurrent form, however, we still seek for a differential
recurrent form in c(p)
k − c(p−1)
k for our proof.
12. Coffee break (double factorial)
Definition:
n!! = n · (n − 2)!! (for n ≥ 2), 1!! = 0!! = 1. (14)
(There are products of odd numbers, and even numbers)
Properties:
(2n)!! =2n
n!, (15)
(2n + 1)!!(2n)!! =(2n + 1)!, (16)
(2n + 1)!! =
(2n + 1)!
2nn!
=
(n + 1)!
2n
2n + 1
n
, (17)
(2n)!!(2n − 1)!! =(2n)!, (18)
(2n − 1)!! =
(2n)!
2nn!
=
n!
2n
2n
n
=
n!
2n
2n
n
2n − 1
n − 1
, (19)
etc.
13. Outline of the proof
1. We wrote an expected form of the coefficients, c(p)
k
2. Then calculate a differential recurrence, c(p)
k − c(p−1)
k
3. We set a linear equation Ax = b, of which solution x should
correspond to the expected coefficients
4. LU decomposition, A = LU
5. The proof for the form of inverse matrices L−1 and U−1 will be
shown later, by using of a modern tool Riordan arrays
6. Calculate L−1b
7. Finally we see that the solution x = U−1L−1b obeys the same
recursion to c(p)
k , given a matrix size (p + 1)
14. Differential recurrence I
This is a hand exercise (remember m = (k + 1)/2 ):
c(p)
k − c(p−1)
k =c(p)
k 1 −
p − k + m
p − k
2p + 1 − 2m
2p + 1
=c(p)
k
m(2k + 1 − 2m)
(p − k + m)(2p + 1 − 2m)
(20)
Now, the difference is simplified as in
1
(−2)k
(2k)!!
(2k + 1)!!
p − k + m
m
(2k + 1)!!
(2k + 1 − 2m)!!
(2p + 1 − 2m)!!
(2p + 1)!!
×
m(2k + 1 − 2m)
(p − k + m)(2p + 1 − 2m)
=
(−1)k k!
(2p + 1)!!
p − k + m − 1
m − 1
(2p − 1 − 2m)!!
(2k − 1 − 2m)!!
=
(−1)k k!
(2p + 1)!!
(p − k + m − 1)!
(p − k)!(m − 1)!
2k−m
2p−m
(k − m)!
(p − m)!
(2p − 2m)!
(2k − 2m)!
(21)
15. Differential recurrence II
Let us now introduce b = k mod 2 ∈ {0, 1}, and ¯b = 1 − b, hence
k = 2m − b. Some useful properties are:
(n + b)!
n!
= (n + 1)b
,
(n − b)!
(n − 1)!
= n
¯b
, (n − 1)b
a
¯b
= n − b,
(−1)k
2p−k
(2m − b)!
(2p + 1)!!
(p − m − 1 + b)!
(p − 2m + b)!(m − 1)!
(m − b)!
(p − m)!
(2p − 2m)!
(2m − 2b)!
=
(−1)k
2p−k
(2m − 1)b
(2p + 1)!!
(2p − 2m)!
(p − 2m + b)!
m
¯b
(p − m)¯b
×
2
¯b
2¯b
=
(−1)k
2p−k (2p + 1)!!
(2m − 1)b (2m)
¯b
(p − 2m + b)!
(2p − 2m)!
(2p − 2m)¯b
=
(−1)k
2p−k (2p + 1)!!
2m − b
(p − 2m + b)!
(2p − 2m + b − 1)! ×
(2p)!!
2p p(p − 1)!
=
(−1)k
22p−k
(2p)!!
(2p + 1)!!
k
p
(2p − k − 1)!
(p − k)!(p − 1)!
(22)
16. Differential recurrence III
Finally, we have
c(p)
k =
1
(−2)p
(2p)!!
(2p + 1)!!
(p = k)
c(p−1)
k +
(−1)k
22p−k
(2p)!!
(2p + 1)!!
k
p
2p − k − 1
p − k
(p > k)
.
(23)
for which we are going to make the proof. Note that this form does not
include m = (k + 1)/2 .
17. Differential recurrence (old version) I
When k is even (k = 2m),
(2m)!
(2p + 1)!!
p − m − 1
m − 1
(2p − 2m − 1)!!
(2m − 1)!!
=
(2m)!!
(2p + 1)!!
(p − m − 1)!
(p − 2m)!(m − 1)!
(2p − 2m − 1)!
(2p − 2m − 2)!!
=
1
2p−m−1
2mm!
(2p + 1)!!
(2p − 2m − 1)!
(p − 2m)!(m − 1)!
×
m!
m!
=
k
2p−k
1
(2p + 1)!!
(2p − 2m − 1)!
(p − 2m)!
×
(p − 1)!
(p − 1)!
=
k
2p−k
(p − 1)!
(2p + 1)!!
2p − k − 1
p − k
×
p
p
(2p)!!
(2p)!!
=
k
2p−k
(2p)!!
(2p + 1)!!
(p − 1)!
2p p!
2p − k − 1
p − k
=
1
22p−k
(2p)!!
(2p + 1)!!
k
p
2p − k − 1
p − k
. (24)
18. Differential recurrence (old version) II
When k is odd (k = 2m − 1),
−
(2m − 1)!
(2p + 1)!!
p − m
m − 1
(2p − 2m − 1)!!
(2m − 3)!!
= −
(2m − 1)(2m − 2)!!
(2p + 1)!!
(p − m)(p − m − 1)!
(p − 2m + 1)!(m − 1)!
(2p − 2m − 1)!
(2p − 2m − 2)!!
= −
2m−1
2p−m−1
(2m − 1)(p − m)
(2p + 1)!!
(2p − 2m − 1)!
(p − 2m + 1)!
×
2p − 2m
2p − 2m
= −
1
2p−k−1
k
(2p + 1)!!
p − m
2p − 2m
(2p − 2m)!
(p − 2m + 1)!
×
(p − 1)!
(p − 1)!
= −
k
2p−k
(p − 1)!
(2p + 1)!!
2p − k − 1
p − k
×
p
p
(2p)!!
(2p)!!
= −
1
22p−k
(2p)!!
(2p + 1)!!
k
p
2p − k − 1
p − k
. (25)
19. Linear equation to solve
What we want to solve is
Ax = b, (A ∈ N
(p+1)×(p+1)
0 , x, b ∈ Qp+1
) (26)
where
Ai j =
2i
j
and bj =
1
2j + 1
, (0 ≤ i, j ≤ p) (27)
For example when p = 5,
1 0 0 0 0 0
1 2 1 0 0 0
1 4 6 4 1 0
1 6 15 20 15 6
1 8 28 56 70 56
1 10 45 120 210 252
1
−5/11
8/33
−4/33
8/165
−8/693
=
1
1/3
1/5
1/7
1/9
1/11
. (28)
gives the coefficients of the 12th-order integrator
20. LU decomposition
An LU decomposition A = LU is available in,
Ai j =
2i
j
, Li j =
i
j
, Ui j = 22i−j i
j − i
, (29)
irrespective to the matrix size (p + 1). Thus, LUx = b can be solved
as x = U−1L−1b.
Example
1 0 0 0 0 0
1 2 1 0 0 0
1 4 6 4 1 0
1 6 15 20 15 6
1 8 28 56 70 56
1 10 45 120 210 252
=
1 0 0 0 0 0
1 1 0 0 0 0
1 2 1 0 0 0
1 3 3 1 0 0
1 4 6 4 1 0
1 5 10 10 5 1
1 0 0 0 0 0
0 2 1 0 0 0
0 0 4 4 1 0
0 0 0 8 12 6
0 0 0 0 16 32
0 0 0 0 0 32
(30)
For the proof, we prepare the tools in mathematical combinatorics.
21. Formal power series, and coefficient extraction
Formal power series
f (t) =
∞
k=0
tk
fk
f (t) is referred to as a generating function of a sequence
( f0, f1, f2, . . .).
Coefficient extraction
[tn
] f (t) = fn
This is an operator but with a weak associativity.
Shifting
[tn
]tk
f (t) = [tn−k
] f (t)
Newton’s binomial theorem
[tk
](1 + t)n
=
n
k
22. Proof for A = LU
2i
j
=[tj
](1 + t)2i
=[tj
] 1 + t(t + 2)
i
=[tj
]
∞
k=0
i
k
tk
(t + 2)k
=[tj−k
]
∞
k=0
i
k
∞
=0
k
t 2k−
=
∞
k=0
i
k
k
j − k
22k−j
. (31)
This discussion is valid for finite matrices, for k iterates from 0 to
min(i, j), touching only the non-zero elements of the upper and lower
triangular matrices.
23. Inverse triangle matrices, L−1 and U−1
For the lower triangle Li j =
i
j
,
L−1
i j
= (−1)i+j i
j
, (32)
and for the upper triangle Ui j = 22i−j i
j − i
,
[U−1
]i j =
(−1)i+j
22j−i
i
j
2j − i − 1
j − i
(1 ≤ i ≤ j)
1 (i = j = 0)
0 (otherwise)
. (33)
The both do not depend on the matrix size 0 ≤ i, j ≤ p. The proof is a
little bit technical and we put them later.
24. Ux = L−1b
L−1
b
i
=
∞
j=0
(−1)i+j i
j
1
2j + 1
=
∞
j=0
(−1)i+j i
j
1
0
x2j
dx
=(−1)i
1
0
1 − x2 i
dx
=(−1)i (2i)!!
(2i + 1)!!
. (34)
This vector does not depend on the vector length p. A proof of the
fourth identity (an integral to double factorials) is in the next slide.
25. Integral and double factorial
We apply the integration by parts:
In =
1
0
1 + x2 n
dx
=
1
0
x 1 + x2 n
dx
= x 1 − x2 n 1
0
−
1
0
x 1 − x2 n
dx
= − 2n
1
0
−x2
1 + x2 n−1
dx
= − 2n
1
0
1 − x2
1 + x2 n−1
− 1 + x2 n−1
dx
= − 2n (In − In−1) . (35)
For In =
2n
2n + 1
In−1 and I0 = 1, we have In =
(2n)!!
(2n + 1)!!
.
26. The solution, x = U−1L−1b
The last element of unknown vector x(p) is now available in
x(p)
p = L−1
b
p
/ [U]pp =
1
(−2)p
(2p)!!
(2p + 1)!!
. (36)
In a general case,
x(p)
k =
p
j=k
U−1
k j
L−1
b
j
. (37)
From the p-dependency of the solution vector x(p), we have a
recursion
x(p)
k − x(p−1)
k = U−1
k p
L−1
b
p
=
(−1)k
22p−k
k
p
(2p)!!
(2p + 1)!!
2p − k − 1
p − k
.
(38)
This is (23), exactly what we wanted!!
27. Riordan array
For the proof of our matrix form L−1 and U−1, we make a brief review
of Riordan arrays.
A Riordan array is an infinite lower triangular matrix taking a pair of
formal power series d(t) (d0 0) and h(t) (h0 = 0, h1 0) 1, of
which element in the n-th row k-th column (counted from (0, 0)) is
R d(t), h(t)
n,k
= [tn
]d(t) (h(t))k
(39)
They make a group, providing explicit forms of
matrix multiplication, identity, and inverse.
In this study, we exploit the Riordan arrays to find matrix inverses of
which coefficients are expressed in binomials.
1Notice that Handbook.pdf by Sprugnoli takes a slightly different notation
R(d(t), th(t)) with d0, h0 0, which does not make a practical difference.
28. Matrix multiplication
By definition, the product of two Riordan arrays is
R d(t), h(t) R a(t), b(t)
n,k
=
∞
j=0
[tn
]d(t) (h(t))j
· [yj
]a(y) (b(y))k
= [tn
]d(t)
∞
j=0
(h(t))j
· [yj
]a(y) (b(y))k
= [tn
]d(t) · a (h(t)) (b (h(t)))k
= R d(t) · a(h(t)), b(h(t))
n,k
. (40)
In the third equality, we used a relation
∞
j=0
xj
· [yj
] f (y) = f (x), (41)
with x = (h(t))j and f (y) = a(y)(b(y))k .
29. Riordan group
Identity:
R 1, t
n,k
= [tn
]tk
= δnk (42)
Inverse:
R d(t), h(t)
−1
= R
1
d(¯h(t))
, ¯h(t) (43)
where ¯h(t) is an inverse function of h(t) such that
¯h(h(t)) = h(¯h(t)) = t.
Proof.
Just substitute a(t) = 1/d(¯h(t)) and b(t) = ¯h(t) for the matrix
multiplication to obtain an identity R(1, t) as a product.
30. Maxima script
Since I am not a mathematician but rather a programmer as you know,
I need concrete examples and numeric dumps to study the arrays.
Maxima is a nice free software for such purposes.
RAelem(d, h, n, k)
:= coeff(taylor(d*h^k, t, 0, n), t^n);
genRA(d, h, nmax)
:= genmatrix(
lambda([i,j], RAelem(d, h, i-1, j-1)),
nmax,
nmax);
Today, these software are not clever enough to find a general form or a
proof automatically, but they help us greatly.
31. Pascal matrix (the matrix L) I
We consider d(t) = 1/(1 − t) and h(t) = t/(1 − t),
L = R
1
1 − t
,
t
1 − t
(44)
The elements are binomial coefficients:
Ln,k = [tn
]
1
1 − t
t
1 − t
k
=[tn−k
](1 − t)−k−1
=[tn−k
]
∞
=0
−k − 1
1−k−1−
(−t)
=(−1)n−k −k − 1
n − k
=
n
k
. (45)
Here, we used −n
k = (−1)k n+k−1
k .
32. Pascal matrix (the matrix L) II
Now we seek for its inverse.
h =
t
1 − t
⇐⇒ t =
h
1 + h
⇐⇒ ¯h(t) =
t
1 + t
(46)
Thus,
L−1
= R
1
1 + t
,
t
1 + t
(47)
L−1
n,k
[tn
]
1
1 + t
t
1 + t
k
=[tn−k
](1 + t)−k−1
=
−k − 1
n − k
=(−1)n−k n
k
(48)
33. Pascal matrix (the matrix L) III
The inverse relation to the infinite triangle matrices is valid for the
finite ones, thus, for example:
1 0 0 0 0 0
1 1 0 0 0 0
1 2 1 0 0 0
1 3 3 1 0 0
1 4 6 4 1 0
1 5 10 10 5 1
−1
=
1 0 0 0 0 0
−1 1 0 0 0 0
1 −2 1 0 0 0
−1 3 −3 1 0 0
1 −4 6 −4 1 0
−1 5 −10 10 −5 1
This confirms the relation Li j =
i
j
and [L−1]i j = (−1)i+j i
j
in the
main story.
34. Catalan’s triangle (the matrix U) I
Let us consider special case that d(t) = 1, and set h(t) = t(1 − t).
B =R 1, t(1 − t) (49)
Bn,k =[tn
](t(1 − t))k
=[tn−k
](1 − t)k
=(−1)n−k k
n − k
(50)
1 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0
0 −1 1 0 0 0 0 0
0 0 −2 1 0 0 0 0
0 0 1 −3 1 0 0 0
0 0 0 3 −4 1 0 0
0 0 0 −1 6 −5 1 0
0 0 0 0 −4 10 −6 1
35. Catalan’s triangle (the matrix U) II
Now, let us find the inverse.
h = t(1 − t) ⇐⇒ t =
1 ±
√
1 − 4h
2
, but ¯h(t) =
1 −
√
1 − 4t
2
B−1
=R 1,
1 −
√
1 − 4t
2
(51)
B−1
n,k
= [tn
]
1 −
√
1 − 4t
2
k
=
k
n
2n − k − 1
n − k
(52)
for 1 ≤ k ≤ n, otherwise δnk .
1 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0
0 1 1 0 0 0 0 0
0 2 2 1 0 0 0 0
0 5 5 3 1 0 0 0
0 14 14 9 4 1 0 0
0 42 42 28 14 5 1 0
0 132 132 90 48 20 6 1
36. Catalan’s triangle (the matrix U) III
Proof. If we put
w(t) =
1 −
√
1 − 4t
2
(53)
Then, t =
w
1 − w
or w =
t
1 − w
. What we want is
[B−1]n,k = [tn](w(t))k and we can apply the following Lagrange
inverse theorem:
[tn
] F(w) w = tφ(w) =
1
n
[tn−1
]F (t) φ(t)
n
(54)
In our case, F(w) = wk and φ(t) = (1 − t)−1.
37. Catalan’s triangle (the matrix U) IV
[tn
] wk
w =
t
1 − w
=
1
n
[tn−1
]ktk−1
(1 − t)−n
=
k
n
[tn−k
](1 − t)−n
=
k
n
(−1)n−k −n
n − k
=
k
n
2n − k − 1
n − k
(55)
for 1 ≤ k ≤ n. For other cases n = 0, k = 0, or k > n, we have trivial
elements [B−1]n,k = δnk .
38. Catalan’s triangle (the matrix U) V
Just transposing and multiplying diagonal matrices from our result
({ai bj Mi j }−1 = {a−j b−i[M−1]i j })
Bi j =(−1)i+j j
i − j
,
B−1
i j
=
j
i
2i − j − 1
i − j
for 1 ≤ j ≤ i otherwise δi j,
we have the last piece needed in our study:
Ui j =(−1)i+j
22i−j
Bji = 22i−j i
j − i
[U−1
]i j =
(−1)i+j
22j−i
B−1
ji
=
(−1)i+j
22j−i
i
j
2j − i − 1
j − i
for 1 ≤ i ≤ j otherwise δi j,
except that Lagrange inverse theorem is used without proof.
39. Lagrange inverse theorem I
From “Handbook of Mathematical Functions”
If y = f (x), y0 = f (x0), f (x0) 0, then
3.6.6
x = x0 +
∞
k=1
(y − y0)k
k!
dk−1
dxk−1
x − x0
f (x) − y0
k x=x0
3.6.7
g(x) = g(x0) +
∞
k=1
(y − y0)k
k!
dk−1
dxk−1
g (x)
x − x0
f (x) − y0
k x=x0
where g(x) is any function indefinitely differentiable.
In formal power series,
[tn
] w(t) w = tφ(w) =
1
n
[tn−1
] φ(t)
n
, (56)
[tn
] F(w) w = tφ(w) =
1
n
[tn−1
]F (t) φ(t)
n
. (57)
40. Lagrange inverse theorem II
This is the last one piece for which we need to see a proof.
As an inverse of a Riordan array R(1, h(t)), that is R(1, ¯h(t)) for
d(t) = 1, we assume the following form:
dn,k = R 1, ¯h(t)
n,k
=
k
n
[tn−k
]
t
h(t)
n
. (58)
If the following relation
vn,k =
∞
j=0
dn, j[yj
] (h(y))k
= δnk, (59)
if satisfied, it is equivalent to
[tn
] ¯h(t)
k
= R (1, h(t))−1
n,k
= R 1, ¯h(t)
n,k
=
k
n
[tn−k
]
t
h(t)
n
.
(60)
41. Lagrange inverse theorem III
We start the multiplication
vn,k =
∞
j=0
j
n
[tn−j
]
t
h(t)
n
[yj
] (h(y))k
, (61)
and by applying a differentiation rule
j[yj
] (h(y))k
= [yj−1
] (h(y))k
= [yj
]y h (y) k (h(y))k−1
, (62)
we continue the calculation, as in
vn,k =
k
n
∞
j=0
[tn−j
]
t
h(t)
n
[yj
]yh (y) (h(y))k−1
=
k
n
[tn
]
t
h(t)
n
th (t) (h(t))k−1
. (63)
Here, we used a convolution rule:
[tn
] f (t)g(t) =
∞
j=0
[tj
] f (t) · [yn−j
]g(y). (64)
42. Lagrange inverse theorem IV
When k = n,
vn,n =[tn
]tn
t
h (t)
h(t)
=[t0
]
h (t)
h(t)/t
=[t0
]
h1 + 2h2t + 3h3t2 + · · ·
h1 + h2t + h3t2 + · · ·
= 1 (65)
When k n,
vn,k =
k
n
[tn
]tn+1
(h(t))k−n−1
h (t)
=
k
n
[t−1
]
1
k − n
(h(t))k−n
= 0 (66)
Here, (h(t))k−n is a formal Laurent series of which derivative cannot
contain a non-zero coefficient for t−1.
43. Lagrange inverse theorem V
Let us set φ(t) = t/h(t) and w = ¯h(t). Then w(t) is an implicit
function of t such that w = tφ(w).
For a special case k = 1, we have
[tn
] w(t) w = tφ(w) = dn,1 =
1
n
[tn−1
] φ(t)
n
.
In more general cases, for a formal power series F(w), we have
[tn
] F(w) w = tφ(w) =[tn
]
∞
k=0
Fk wk
=
∞
k=0
Fk dn,k
=
∞
k=0
Fk
k
n
[tn−k
] φ(t)
n
=
1
n
[tn−1
]
∞
k=0
kFktk−1
φ(t)
n
=
1
n
[tn−1
]F (t) φ(t)
n
.
44. Summary
Combinatorics, a branch of mathematics, turned out to be a
powerful tool to study numerical integrators for the N-body
problem. Especially the formal power series, Lagrange inverse
theorem and modern Riordan arrays are practical tools for even
non-mathematicians.
LU decomposition is not only for Top500/Green500, but a
general tool for studying integers.
Mathematics is useful.
Satoko Yamamoto did a great work.