charles's professional blog

I work on market simulations at HF scale

2009-11-01T09:45:00.004+01:00

Conference on Quantitative Risk Management today

2009-09-18T21:36:00.002+02:00

Huyên Pham organized a conference today, following his prize sponsored by the NATIXIS Foundation for Quantitative Research in association with the SMAI and awarded by the French Academy of Sciences.

Very interesting talks:

Michel Crouhy gave us a clear view on the past crisis

Benjamin Bruder proposed a nice way to include risk limits in option hedging strategies

Jean-Paul Laurent exposed how to include several effects in hedging interest margin

I made a review of the quantitative trading

Nicole El-Karoui presented an elegant way to implicitly define an utility function from a market observable (like a benchmark portfolio)

Rama Cont explained how systemic risk is embedded into the financial network, and increases with speculative use of CDS

Nizar Touzi described how defining a relative benchmark to asset manager (scoring them relatively to the performances of their pairs) encourages them to take more risk

Denis Talay benchmarked technical analysis against more objectively optimized strategies

My CQF masterclass

2009-03-28T19:49:00.003+01:00

"Intraday High-frequency Trading: From empirical evidence to quantitative optimization" is available only for CQF people at:

Cite U Like

2009-03-28T13:38:00.003+01:00

The wonderful bibliography web site (citeulike.org) has now a FAQ.
So I can give you a sorted (by increase on publciation years) view of my main library on microstructure:

Scheme programming: what else?

2009-02-07T17:49:00.005+01:00

Do you need a complete programming language that can be carried on an USB key?

Noting is better than scheme for that.Here is a list of useful links for a schemer:

refcard

TSPL

[ascii]

An Introduction to Computer Science Using Scheme

How to Use Scheme

Schematics Scheme Cookbook

An Introduction to Scheme and its Implementation

R5RS

R6RS

2D real Brownian Motion

2009-01-11T20:10:00.003+01:00

Reality is far better than simulations and models (nothing to add about CDO models?): here is a movie I have made this morning, a real 2D Bornwian motion.

Usual intraday curves for students

2008-12-23T15:11:00.005+01:00

The best way to illustrate rigorous intraday trading is to use numerical applications.
Here are projections of the usual volume and volatility curves of a stock of the main french index (CAC40), in percentage.
Just multiply the result by its sum and multiply it by the ADV for volume and usual intraday volatility you wish for the volatility curve and you will obtain curves good enough for numerical illustrations.

x is the elapsed part of the day (0 at 9:00, 1 at 17:30), Pv(x) is the usual intraday volume and Ps(x) the usual intraday volatility:

Pv(x) = 11 x^3 - 4 x^2 -4 x +3, Ps(x) = -12 x^3 + 25 x^2 -15 x + 5

.

A paper of mine in Wilmott magazine

2008-12-20T14:27:00.005+01:00

The issue of November of the Wilmott mag contains a paper of mine:

Its title is Rigorous Optimisation of Intraday trading.

.

Good statistical references for a Quant

2008-12-14T09:29:00.001+01:00

A lot of quant padawans are asking me good books to read to improve their statistical skills.

For the statistical part of the work, I can recommand two books.

The first one is Méthodes statistiques by Philippe Tassi.
It offers simple elements of mathematical statistic and their applications to simple problems (regression, estimation, hypothesis testing).

And the second one is Mathmatical Statistics by Borovkov.
It is a very complete book on almost all aspects of statistics.

.

How to guard an hotel in St tropez

2008-10-09T23:04:00.008+02:00

Last week end it was the Voiles of Saint Tropez.
We stayed in an amazing hotel at Sainte Maxime, in front of St Tropez, where we could see the golf of St tropez and all the ships.

This hotel is like "plugged" on a small hill, and you consequently have to cross a lot of lifts and corridors to join your bedroom.
It reminded me a chapter of Proofs from the BOOK, by Martin Aigner and Gunter M. Ziegler.

This chapter is called "how to guard a museum". Adapted to the Sainte Maxime's hotel, it gives:

The manager of the hotel wants to make sure that at all times every point of the hotel is watched by a guard. The guards are stationed at fixed posts, but they are able to turn around. How many guards are needed?

It is for instance easy to guess that if the walls of the hotel form a convex polygon, then one guard is enough.

There is a useful theorem for managers of such hotels:

For any hotel with n walls, n/3 guards suffice.

High freq finance with MATLAB

2008-08-02T00:09:00.004+02:00

I just began to write a literateprograms.org page about high freq finance with examples in MATLAB.
Feel free to comment and contribute...

How many fishes in your Dark Pool?

2008-07-11T15:13:00.008+02:00

With the progressive increase of transaction in European Dark Pools (Euromillenium and Equiduct), a lot of quants are thinking about ways to tune their Smart Order Routers.

Some funny mathematical questions arise when you try to estimate the available number of balls in a jar without being able to have any look in it.
Zoologists had to solve one of them: they need to estimate the number of fishes in a lake only being able to capture some of them. A simple version of this is: "having a fishing rod, how can you estimate the number of fishes present in a lake?" (without fishing them all).

I put the answer in the comments of the post...

References

The Estimation of Total Fish Population of a Lake search in Google.scholar

This post is part of the Carnival of mathematics.

Oil price explained

2008-07-05T13:10:00.005+02:00

I do not know if cross links over blogger are a good think or not, but here is a fine one:

FUTURES PRICES DETERMINE PHYSICAL OIL PRICES

Evolution, algorithms and mathematics

2008-06-14T17:11:00.011+02:00

I was today at the Grande gallerie de l'évolution with my son.

It reminds me the first research I conducted (now long ago, around 1994) about the anaysis of natural genetic algorithms.
I started form the work of Raphel Cerf, and the results (about simulated annealing) obtained by Robert Azencott.
I never published my results (may be I should) but only presented them twice at Ecole Normale Supérieure Seminars.

It was the first time I realised that studying the real world (DNA evolution, explained to me by Philippe Tarroux) could help me to solve mathematical problems, and that computer simulations could give me hints about how to solve theoretical problems.

.

Top ten list of SSRN

2008-06-08T18:07:00.005+02:00

Here is a link to the Top ten downloads of the SSRN web site.

The paper I co authored with Yann Braouezec is in the list:

How the Corporate Liquidity Process Affects the Value of the Firm.

By the way, we are currently working on its improvement...

The econometrics of high-frequency financial data

2008-06-01T07:26:00.007+02:00

I was in London the last three days.

I had the chance to attend to a workshop called "continuous-time financial econometrics" and a seminar about "the econometrics of high-frequency financial data".

Very interesting points were presented there, about estimating misc elements of an Ito semi martingale like:

in presence of microstructure noise.

Paper presentation

2008-05-21T21:55:00.006+02:00

Yann Braouezec presented today at the Polytechnique/Columbia "New Directions in Quantitative Finance" conference a paper we co authored about corporate policy.

The paper is called How the Corporate Liquidity Process Affects the Value of the Firm. It is quite theoretic at this stage, but we hope that it can be the right extention of existing theories (Merton, Modigliani Miller, Leland, etc).

It is mainly a stochastic control formulation of the choice of the optimal corporate policy (dividend rate, coupon and nominal) of a reallistic discrete time model of the evolution of a firm. Such a model is able to render at the same time liquidity and solvency critical situations for a firm.
An interesting point is that the model can be reformulated as a more usual model but including an endogeneous stochastic barrier, which is quite innovative (usual stochastic barriers are defined a determinsitic way).

Water and Casino

2008-04-27T10:21:00.007+02:00

A week of holiday at Vittel, a very nice city:

Vittel et son histoire

Connaissez-vous l'histoire de Vittel

Vittel jamais é-li-mi-née depuis 150 ans

The building were you can find Vittel's Casino has been design by Charles Garnier around 1880; it has a wondelful Art Deco style:

A good occasion to optimize gaming strategies... I will put a pdf doc of my (not really useful) results in this field soon...

London, CQF

2008-04-27T09:01:00.006+02:00

I gave my one day long lecture (for professional quants) titled "Intraday High-frequency Trading: From empirical evidences to quantitative optimization" (15 April - London).
It was very interesting and complementary to the lecture I give with J-G Attali to students at Paris VI master in finance.

The main goals of the course were :
- to give a clear idea of factors that affect intra day trading
- to explain how to capture some intra day statistical invariances (volume, volatility curves, etc)
- to be able to understand and implement papers on quantitative formalization of intra day trading

Content

1. practical aspects of auction markets
1.1 bid-ask spread, limit order books
- fixing auctions and continuous auctions
mechanisms and some funny computations
- from price impact to market impact
(from a "trade by trade" description to longest effects)
1.2 intra day trends
- volumes: do not use "average" on log normal distributions !
- volatility : do only add squared volatilities !
(a point on volatility high freq estimation)
- bid-ask spread : when microstructure limits the use of diffusion models
1.3 the two main enemies of intraday trading
- market impact (volume driven) demands to trade slowly
- market risk (volatility driven) demands to trade fast
- those two effects are mixed (market impact is correlated to volatility)
1.4 alternate sources of liquidity
- MTF (Chi-X, turquoise, etc)
- Dark pools
- what is a MiFID-compliant Smart Order Router ?

2. quantitative optimization of high freq trading
2.1 high freq trading in equations
- properly writing down point (1.3)
- cost minimization
- risk minimization
2.2 in the heart of darkness: market impact models
- Review of main models
- standard pitfalls
2.3 from simple to sophisticated optimization
(with simple tools: lagrangian optimization)
- simplest case: constant volatility and volumes
- simple case: random volumes
- useful case: stochastic volatility
- reallistic case: taking into account correlations
2.4 two main evolutions of stock trading
- market making
- portfolio trading
2.5 non parametric approaches
- implicit market impact models
- direct use of such models
2.6 misc
- interpretation of risk avertion

Pyramidal numbers

2008-04-05T19:25:00.004+02:00

The "square" of a number N is the number of ball to be used to draw a square which edge is made of N balls.
The "pyramid" of a number N is the number of balls to be used to build a pyramid (like on the picture).
The "triangle" of a number N is easy to compute : its N + (N-1) + (N-2) + ... 1, i.e.

and consequently the pyramid of N is :

(cf number 354 of the Revue du Palais de la Découverte).

Algorithmic trading, again

2008-04-05T19:19:00.005+02:00

Always the same equations to explain algorithmic trading to trainees: always the same example of a buy order to be cut in n bins of volume. The price is assumed to follow an arithmetic Bronwian motion. An immediate market impact function is added...
Expectation plus variance of the overall price is to be minimised (a risk avertion scalar has to be used).
Some changes of variables and the simple assumption of a linear market impact: an difference equation of order two is found and easily solved...

Variations on the Regularized Incomplete Gamma Function

2008-04-04T06:52:00.000+02:00

You cross the path of the Gamma function very often when you do some maths.
Today it was the Regularized Incimplete Gamma: Q(a,z)=Gamma(a,z)/Gamma(a), i.e. :

Worlfam's related notebook.

Desperately Seeking a Poisson Bridge

2008-03-28T20:27:00.004+01:00

I had a lunch in a fish (in french "a fish" is called "un poisson") restaurant close to l'Arc de Triomphe (a useless bridge wanted by Napoleon to celebrate his own glory), when I remembered that I recently wanted somethink like a "Poisson bridge".
Like Brownian bridge is a Brownian process given its arrival point, I worked with a Poisson process given its arrival time (a sum of K waiting times following an exponential law of parameter lambda such that the Kth jump arrived at T).

Algo Trading on CA Cheuvreux's web site

2008-03-27T21:49:00.003+01:00

Promising, isn't it?

Strongly mixed whiteboard

2008-03-26T21:29:00.002+01:00

Today was a mix of PCAs, density estimations, markovian order books, and bootstrap...

Wait more, pay less

2008-03-25T21:30:00.005+01:00

Did you know that if you arrive randomly in a station, and if the time between two trains is exactly 15 minutes, you will wait two times less than if the time between two trains is a random variable having an expectation of 15 minutes ?
(15 can be replaced by any other number. Question for students: what are the needed assumption?).

What about the waiting time of passive orders in a limit order book?

Sudden hail in March

2008-03-22T14:55:00.009+01:00

I was in the forest thinking about necessary conditions on the orders flow to ensure a stationnary LOB, and suddently: hail!

When you are under such a weather for only 10 minutes, you only think about how you can get out if it goes worst ; you damn yourself to not have think about taking a better coat, you swear to yourself that next time you will be more careful before going out of your swet home ...
What about the actual financial context (Bear Stearn, SocGen, etc) ?

White board daily

2008-03-20T20:43:00.003+01:00

I will try to post every day a picture of my whiteboard or of my drafts (I will only post some non sensitive information).

Today it was about mutual information and conditionnal entropy...

Algo trading recruitment

2008-03-02T07:42:00.003+01:00

This link is a good start to try to build some statistics about the interest of financial institutions about algo trading : Jobs for PhD: Algo Trading.

More generally, here are web site about quant jobs:

Who am I?

2008-02-17T19:40:00.002+01:00

I have been included in the Who's Who in Science and Engineering 2008.

CQF course

2008-02-16T16:00:00.001+01:00

I will give a ourse at Wilmott's CQF this year :

* New for 2008. Only available to Past, Present and Future CQFs. Intraday High-frequency Trading: From empirical evidences to quantitative optimization with Charles-Albert Lehalle PhD, CQF, Head of Quantitative Research at Crédit Agricole Cheuvreux, CALYON Group. Factors that affect intraday trading; how to capture intraday statistical invariants; understand and implement quantitative formalization of intraday trading.

Little Nicolas as PhD student

2007-10-07T10:05:00.000+02:00

The PhD director
At first, one must have a PhD director. The director is a very smart gentleman who ask me a problem and who will help me to solve it. Here, you can see my director ; Christian Mercatier-Bresson took the picture.

The Research

At first, it's so complex that you do not understand a word.

One can search for hours and hours without finding anything. During those hours, my parents have a lot of concerns and when mom ask to dad if it was a good idea to have a thesis made by the little one (it's me), my daddy open the mouth without a word, he moves his arms, and go into the living room to read a newpaper.

The Finding

Somethimes it's amazing because I found thinks my director asked me about. It obviously can append at any time of the day and my parents are not always pleased of that.

She wonders if I'm going crazy. But my mom knows that daddy does not like when she asks about it.

But me, I think it's quite natural to be happy. Moreover, when my director find a theorem, he is very proud and his friends (which are very smart too) are very proud of him too. But my parents do not know anything about that.

Sometimes I'm in trouble because I was wrong. And when I'm wrong my director does not laught at all. "look at me in the eyes, Nicolas", he says "you call that real work?" he asks me then. And if it seems to be a question, you do not have to answer, because he can be very angry.

Seminars
Sometimes, a gentleman very famous and very very smart (nevertheless not as smart as my director) comes to speak about very complex things. It's called a seminar, and a seminar is not funny either. When the gentleman ends his speak, my director ask very complex questions to him, and the gentleman does not always know the answers. But it's not fair because my director is not angry at him in this case.

Viva voce
At the end, there will be a ceremony with a lot of very very smart people (there will be even others directors) and there will be a very very old gentleman who will say that's very good little boy, that the glorious way of Research is now open to me and that my parents and my country can be proud of me and so on. And then there will be a snak with my friends. Super!

And when my dadday will read that in the newpaperhe will be very proud and mom will be so happy that she will give me two pieces of my favorite cake. It's so cool a thesis, at the end.

Celebrity
Besides, grils are very impressed when you say that you've got a PhD in mathematics and that you have found a lot of complex theorems and so on. Edwige's mother smiles at me now, even if she thought that I was an unruly boy before.

Source: G. Taviot.

Attilio's book resources online

2007-06-01T09:47:00.001+02:00

I just recieved this message from Attilio Meucci:


Dear all,

the MATLAB routines that support the book “Risk and Asset Allocation” – Springer have been 
updated and are now available at www.symmys.com . Also find on the same page the “Slides”, 
presentations that walk the reader through the whole book.

The MATLAB routines include many new features:
- more uni-, multi- and matrix-variate distributions
- more copulas
- more graphical representations
- more analyses in terms of the location-dispersion ellipsoid.
- best replication / best factor selection
- FFT-based projection of a distribution to the investment horizon
- caveats about delta/gamma pricing
- step-by-step evaluation of a generic estimator
- non-parametric estimators
- multivariate elliptical maximum-likelihood estimators
- shrinkage estimators: Stein and Ledoit-Wolf, Bayesian classical equivalent
- robust estimators: Hubert M, high-breakdown minimum volume ellipsoid
- missing-data techniques: EM algorithm, uneven-series conditional estimation
- stochastic dominance
- extreme value theory for VaR
- Cornish-Fisher approximation for VaR
- kernel-based contribution to VaR and expected shortfall from different risk-factors
- mean-variance analysis and pitfalls (different horizons, compounded vs. linear returns, etc…)
- Bayesian estimation (multivariate analytical, Monte Carlo Markov Chains, priors for correlation 
  matrices)
- estimation risk evaluation: opportunity cost of estimation-based allocations
- Black Litterman allocation
- robust optimization (calls SeDuMi to perform cone programming)
- robust Bayesian allocation
- more…

A very good book indeed, and useful resources on his web site...

Three very usefull online softwares

2007-06-01T09:29:00.000+02:00

Here are 3 usefull software for people who (like me) computes always and anywhere...

CitULike to maintain a list of references. It is possible to access it from anywhere via the internet and to store pdf version of the papers

SAGEmath is an opensource symbolic computation engine written in Python. An online version is available.

LaTeX online is a full version of LaTeX online (produce png, pdf, and a lot of other formats as output).

7 steps to share an arXiv.org paper on your blog

2007-05-16T13:48:00.000+02:00

Here is a way to share papers published on arXiv on a newsfeed (RSS format) or on a blog (like this one).

First you neew a google account (who has no google account?), and use the Google news reader. If you do not have a e-print mail alert, you probably can go directmy to the 2nd step.

Then here are the steps:

1. Select the paper according to its abstract in your mail reader, click to go to the arXiv page.

2. Once you are on the arXiv page of the paper, click to go the its citebase associated page.

3. Once you are on the citebase page of the paper, click on any of the authors' names.

4. Select the copy link to clipboard on the RSS link.

5. Paste the RSS link of the author's search in your google read 'add' textarea.

6. Now you can star, share, and tag the paper, given that the feed name is always 'citebase search'.

7. Unsubscribe to the feed, the paper will anyway stay in your starred or shared items.

It's done, now you can publish your public google RSS feed, see mine on the right of this blog.

Used links:

arXiv.org

e-Print network

Citebase.org

Audio from an (old) conference (in french)

2007-04-30T22:28:00.000+02:00

Here is the audio file of an conference at the Collège de France given in french by J.M. Lasry the 13th of Oct. 2006:

Audio (m4a)

MATLAB World Tour 2007

2007-04-28T20:28:00.000+02:00

MATLAB world tour 2007 was in Paris last week.
I went there and had the chance to meet Loren Shure.

Master course

2007-04-28T20:24:00.000+02:00

I gave a course at the Master of LPMA this year with Jean-Gabriel Attali.

Our main goal was to give a clear view of quant methods that can be used from portfolio allocation to trading. No lecture notes this year, I did not had enough time to typeset them...

A new book on my shelves

2007-04-28T18:07:00.000+02:00

I just recieved a "new" book I bought at Betterworld.com:

Limit Theorems for Stochastic Processes by Jean Jacod and Albert N. Shiryaev. It's not the latest edition but interesting enought.

I wanted for long a generic book about stochastic processes. Of course I have:

the well-known and always cited Karatzas and Shreve's Brownian Motion and Stochastic Calculus

I have to more deeper Doob's Classical Potential Theory and Its Probabilistic Counterpart too.

But I wanted a book in line with the advanced publications and communications of Pr Jacod that can be used in a micro structure context (see Jacod & Delattte, Zhang, Mykland & Ait-Sahalia, or Barndorff-Nielsen & Shephard), this one is perfect.

MiFID informations

2007-01-30T15:18:00.000+01:00

Here are some usefull links about MiFID (Markets in Financial Instruments Directive):

An overview paper

The European commision web site

Here is a free book from the Asian devlopment bank entitled:

Demutualization of Stock Exchanges: Problems, Solutions and Case Studies:

A new paper of mine

2007-01-30T14:30:00.000+01:00

Appreard (in french) in "Le courrier des statistique".

The contribution of statistics to experimental research in industry and the financial markets - Charles-Albert Lehalle

In both the financial markets and industrial sectors, the research engineer and the physicist are today plunged into an environment that gives them ever more data that is ever more precise. When they have to study manufacturing processes, products, or stock market fluctuations generated by increasingly sophisticated mechanisms, they have to turn to statistics. Implementing an experimental approach, and thus validating or invalidating working hypotheses with large volumes of data, cannot be done without recourse to hypothesis testing. Moreover, detecting a critical change for a process requires being able to continuously monitor the probability distributions for the variables that characterise it. The research centres of industries and financial organisations are therefore now exposed to what can be called a “model risk”: we need to make sure that the model for a macroscopic phenomenon, made up of a very large number of small phenomena studied and understood in isolation by the experts, always works when faced with data or the use of statistical methods. These methods must be integrated into the business expertise, which gives rise to stimulating and fruitful collaborations between physicists and statisticians.

This blog now uses Blogger 2.0

2007-01-30T14:00:00.000+01:00

I just convert it to Blogger 2.0, may be it will be easier to use...

Wilmott's wiki opened

2007-01-28T15:22:00.000+01:00

The wilmottosphere has a new elements, a wikimedia!
Come and contribute, and leave messages on my user discussion page

writely for blogger

2006-10-07T13:38:00.000+02:00

It's now possible to post a writely document in a blogger blog like this one.

Google calendar for "le Cours Bachelier"

2006-10-02T13:31:00.000+02:00

The home page of the cours can be found at bachelier-paris.com, and I've made a public google calendar with it:

Selected RSS feeds (data)

2006-09-30T08:48:00.000+02:00

You can now access to the RSS news I selected using the Google Reader through a dedicated web page, or the frame at the right of this blog.

Triple Withcing Google Calendar (data,finance)

2006-09-15T09:17:00.000+02:00

I made a public google calendar for the "triple witching hours" (The "triple" in this event involves the simultaneous expiration of stock options, index options and index futures):

Pattern Recognition and Machine Learning (maths)

2006-08-22T18:45:00.000+02:00

A new textbook by Christopher M. Bishop is available:

It should be interesting.

stochastic dominance (maths)

2006-08-22T14:52:00.000+02:00

http://www.ucalgary.ca/~sick/Research/SimulateReal.pdf#search=%22hanoch%20levy%20cdf%22 p7

http://cemmap.ifs.org.uk/forms/bodurtha_paper.pdf#search=%22hanoch%20levy%20dominance%22 finance

Economists and financial blogs

2006-08-15T23:34:00.000+02:00

Here is a short list of economists and fincial blogs (a list of other economists blogs):

economistsview@typepad

neweconomist (with a list of a lot of other economists blogs)

the Becker-Posner blog (Nobel Price), and its counterpart

Nassim Taleb's blog

Paul Wilmott's blog

institutional-economics

Brad DeLong's Semi-Daily Journal

Nouriel Roubini's Blog

And some more funny blogs (found in a Bloomberg Market article):

bankersball

equityprivate@typepad

leveragedsellout

dealbreaker

News from Springer

2006-07-26T00:43:00.000+02:00

Springer reprints the famous "Eléments de mathématiques":

Here is the last version of their quantitative finance catalogue.

CQF extra seminar (finance)

2006-07-25T21:06:00.000+02:00

Patrick Hagan.

Patrick Hagan - Managing Smile Risk

Abstract: Market smiles and skews are usually managed by using local volatility models a la Dupire. We discover that
the dynamics of the market smile predicted by local vol models is opposite of observed market behavior: when the price of
the underlying decreases, local vol models predict that the smile shifts to higher prices; when the price increases, these models predict that the smile shifts to lower prices. Due to this contradiction between model and market, delta and vega hedges derived from the model can be unstable and may perform worse than naive Black-Scholes’ hedges.

Quite interesting.

The quote of the lecture: "let's use the Level as numeraire"

What it inspired to me:

read something new about SINGULAR PERTURBATIONS IN OPTION PRICING, like this?
read more books of Albert Shiryaev, as Statistical Experiments and Decisions: Asymptotic Theory, or Statistics of Random Processes, I and II, and Random Iterative Models, by Marie Dulfo.

My CQF results

2006-07-19T18:19:00.000+02:00

I am now CQF graduated!

A Refcard for Borodin and Salminen's book (maths,finance)

2006-06-29T08:38:00.000+02:00

I made a small refcard for readers of the handbook.

Dukascopy free datasets

2006-06-21T13:48:00.000+02:00

Some free datasets are available from Dukascopy, like this dynamic calendar:

You can also go to StyleAdvisor to have a description of common datasets.

Order books at yahoo.finance

2006-06-02T16:54:00.000+02:00

For people who did not know, 'real time' order books are available on yahoo.finance.

Literateprograms.org

2006-05-26T07:53:00.000+02:00

I'm now sysopt of the wiki Literateprograms.org.

Benzecri, Tukey and Hayashi (maths)

2006-04-11T23:30:00.000+02:00

In the 24th letter of the International Federation of Classification Societies Newsletter (IFCS), Pr Noboru Ohsumi, wrote this in memory of Pr Chikio Hayashi.

He set up an interesting perspective between the works of J.P. Benzecri (Analyse de Données), J. Tukey (Exploratory Data Analysis, EDA) and C. Hayashi (Data Sciences).

An example of Tukey's work

Professor Chikio Hayashi and Multidimensional Data Analysis –
Quantification Methods and Data Analysis, from Classification to Data Science

Professor Chikio Hayashi is no longer with us. We have lost a great man. To me, his death marks the loss of a valued mentor who acted as a guide not only in the field of statistics but also through the confusion that exists today in the fields of social research and data analysis.
In this tribute to Professor Hayashi, I have drawn upon thirty years of memories. I treasure the simple ones most – sharing a room on a tour of the Loire River in France, going on an excursion with the professor and his wife in the intense heat of Morocco, trying to decide how much to pay for a carpet, walking together in midday at Rome, and accompanying the professor on countless occasions to different restaurants overseas. I can still see him, wearing his trademark beret, encouraging me to do my best because he is relying on me. At such times I am overcome with emotion.
I cannot describe the full extent of Professor Hayashi's achievements. If I started, there would be no end. Instead I shall write about some of the memories I have of our long relationship. I hope this tribute to him will show the Professor Hayashi that I knew and some things that are not commonly known about him.

The Teachings of a Highly Inquisitive Educator

Professor Hayashi had extraordinary knowledge in many areas. His diverse interests included music, sports, books, calligraphy and the tea ceremony. When a new topic came up, he would often telephone me at the laboratory and say: "Look, there's apparently some new research going on in such and such, do you know about it?" or "If you've got any information or materials on such and such, let me have them." Once he had read the latest information, he quickly formed an opinion, and thus followed the modern work in fractals, genetic algorithms, neural networks, rough set theory, complex systems, support vector machines, data mining, and knowledge discovery.
Professor Hayashi knew how to manage people efficiently. He had an extensive personal network through which he constantly gathered the latest information. One thing he taught me was that this kind of attention to detail is an essential for researchers, particularly in statistics and data analysis. Whenever I approach a problem from the professor's perspective I see it in a different light, and I am grateful for this gift of understanding. Looking back, I believe this approach was part of the professor's unique teaching method.

Multidimensional Data Analysis and Multivariable Analysis – These Are Two Different Things

Professor Hayashi strongly believed that reality is understood through the careful use of words. This is obviously true for a research paper but the professor went further in his approach to technical terms. For example, he would call "statistics" by that name alone and disliked the phrase "statistical science." Similarly, he insisted on the term "multidimensional data analysis" as opposed to "multivariate analysis". People with less insight often said that either term was acceptable but the professor would not allow this.
I once asked Professor Hayashi to read through an article I wrote on multidimensional data analysis. When he returned it I found that he had crossed out the phrase “multivariate analysis” everywhere and had even made a marginal note that "use of that term degrades your paper." Yet he had also written "you're quite right," and "this is good; it's consistent with my view" next to parts he liked. Nonetheless, he was never a nit-picker. He did not concern himself much over trivia such as misspellings, and he was happy as long as the essence of the analysis and the main points were conveyed. This was also true of his teaching style.
In any event, multivariate analysis for him was just an unnecessarily complex arithmetical operation based on Gaussian distributions that had too many restrictions for practical use and could only be applied in research papers. He lamented that there were too many cases where "you can't see the wood for the trees."

Quantification for Exploring Phenomena and Acquiring Information

Most researchers refer to Hayashi's quantification theory but in his books the professor simply called it "quantification." And although Professor Hiroshi Akuto’s names for the quantification methods (Types I-VI) are now common, Professor Hayashi himself called them by different names, reflecting their respective ways of thinking. But Professor Hayashi frequently commented that Professor Akuto's names helped spread the quantification methods. I can only admire this kind of consideration for other people's feelings.
The origin of the quantification method derives from the idea that "numbers (numerical values) do not exist ahead of time but are allocated as appropriate

in accordance with our objectives." Thus we should make a clear distinction between (possibly non-numeric) raw measurements and numeric values used for analysis. While there are common ways to give numerical value to qualitative data, a different approach states that since many things in the world are nonlinear, we should approximate these in an easily manageable linear form, devise creative experiments, and solve problems through data collection and analysis methods appropriate to the phenomena. This approach requires detailed modeling, and is the antithesis of the approach (used by many statisticians) that fits data to a simple conventional model. I believe that the essence of Professor Hayashi’s thinking on this is still not properly understood within mainstream statistics, and his efforts to explain this led him to emphasize "data science" in his later years.

First Acquaintance with Professor Benzécri and Correspondence Analysis

In the early 1960s, Professor Benzécri developed the AFC method (analyse des factorielle correspondances, or correspondence analysis). His work formed the basis for data analysis (analyse des données) in France. Professor Benzécri strongly criticized the existing mathematical statistical approach and his unique lectures at University of Paris VI became widely known.
Some people pointed out that AFC appeared to be the same as quantification method Type III. As expected, Professor Hayashi conducted a detective-like investigation, even going so far as to question the existence of Professor Benzécri and to wonder whether it wasn't a research organization like Bourbaki. His doubts were resolved, however, in 1979 when Professor Hayashi and Professor Benzécri met for the first time. Professor Benzécri subsequently wrote: "I was surprised that, of all places, it was the Far East where there was a researcher who had come up with the same ideas as me, and quite clearly earlier than I had." In this way it became known throughout Europe that quantitative method Type III was the same as correspondence analysis. However, Professor Hayashi insisted that although there were similarities in equations and formulation, the basic ideas and philosophies behind were different. Professor Hayashi's foresight in the analysis of qualitative information stood out and this approach became widely known overseas.

Research Exchange between Japan and France – International Exchange in Data Analysis

Data analysis, particularly EDA (Exploratory Data Analysis), is an approach advocated by Dr. J. Tukey in the early 1960s, at about the same time as the appearance of AFC. Professor Hayashi noted this early on but pointed out that although there were common aspects with his concept of data analysis, Tukey’s EDA assumed only limited data. The analysis envisaged by the Professor Hayashi was linked more closely to practical application, and took a more integrated approach.
As Professor Hayashi’s quantification methods (particularly quantification method Type III) spread internationally, a growing body of researchers in France wished to invite him to speak. In 1979, a group from INRIA (Institut National de la Recherche en Informatique et en Automatique), including Professor Diday, asked Professor Hayashi to be the guest speaker at the international research meeting of "Data Analysis and Informatics" held biannually in Versailles. I still vividly remember the professor's lecture at that time. To put it bluntly, the professor's delivery of his lecture in English was not particularly impressive but it was full of his trademark philosophical significance. The audience was restless when he began but by the time he had finished there was resounding applause as the audience recognized the charisma of this figure from the Far East. Professor Benzécri's disciples flocked to shake Professor Hayashi's hand and this was clearly the moment that his ideas on quantification were seen as being in common with the data analysis of France.
Professor Hayashi tried to invite Professor Benzécri to work in Japan on several occasions, but Benzécri refused to travel by airplane. Instead, Benzécri sent his disciple Professor M. Roux. This led to exchanges with many French researchers, all of whom are today at the forefront of data analysis in France, including Lebart, Jambu, Diday, Nakache and Escoufier. Professor Hayashi established a particularly strong bond with Professor Lebart that included joint work in international comparative research. Professor Hayashi invited Professor Lebart and many other researchers to Japan. My colleagues and I were among those who benefited from the extensive research contacts between our countries. These exchanges and the joint research that was undertaken advanced both the field of data analysis and the area of international comparative studies.
Professor Hayashi regarded these researchers as important human resources. He sought to promote contacts such as the Japanese-French Scientific Seminar, which was held on two occasions: in 1987 in Tokyo and in 1992 in Montpellier, under the auspices of JSPS (Japan Society for the Promotion of Science) and CNRS (Centre National de la Recherche Scientifique). The professor played an active role in academic activities such as these and I will never forget how he constantly encouraged and inspired us.

From Classification Research to the Data Science

The international research exchange initiated by ProfessorHayashi expanded from France to the U.K., Italy, Germany, Portugal and Eastern Europe. He built fruitful personal connections with Professors Gower, Lauro, Rizzi, Bock, and many others. Part of the reason behind this expansion was classification research. For a long time, the professor believed that "classification" was a "fundamental thought process in all science" and he spoke of the importance of classification even when considering quantification. Groups of researchers working in this field began to emerge in many different countries, resulting in the establishment of classification societies in North America, Great Britain, France, Germany, Italy, and other countries. When Japan decided to establish a classification society in 1983, the professor, perhaps because he was a naturally curious and competitive man, first set it up in the form of a research group. In 1991, the research group became a society and has continued to this day. The society's activities blossomed through its participation in the International Federation of Classification Societies (IFCS) and the hosting of the 5th IFCS International Conference in Kobe in 1996. It was here that the professor gave the keynote address, where he spoke about the principles and importance of data science. This talk laid out his view of exploratory data analysis, and was seminal in prompting new research in this area.
Professor Hayashi's achievements were recognized internationally. He was appointed to several positions, including an honorary membership of the Royal Statistical Society, a regular membership in the ISI (International Statistical Institute), and the presidency of the IFCS.
In the last ten years of his life, Professor Hayashi asserted the importance of "data science as a theory of scientific methods." The starting point was the meaning of the words "data science;" this arose in 1992, when the professor was discussing the titles and introductions of a collection of papers to be published at the 2nd Japanese-French Scientific Seminar. The professor used a Japanese phrase that translates as "data science" and for the rest of his life explored the concept behind this phrase. When the papers at the seminar were published, it was proposed for the first time that the term referring to quantification be standardized as "quantification method" and the subsequent English terms were standardized as such. I can recall the professor reprimanding us with the words "you are talking about different concepts" when we referred to "deta kagaku" while another group called it "deta no kagaku" (data science).
The basis of data science is an extremely straightforward concept; the fact that it needed to be expressed is an indication of the chaos that exists today in statistical data analysis and the deteriorating research environment. I am certain that the professor had intended to lead the way in achieving a breakthrough on these problems.
Professor Hayashi's great achievements over a period of many years were recognized when he was awarded the Purple Ribbon Medal in 1981, followed by the Order of the Sacred Treasure, Gold and Silver Star in 1989. Following his death, he was posthumously conferred with "Shou-shi-i" in Japan's official court rank.
The professor was 84 years old when he passed away. It was too soon – we still need his wisdom. But his friends must continue without him, working as evangelists to spread the concepts of data science. I pray for the professor's happiness in the next world.

May he rest in peace.

Noboru Ohsumi (Dec 2002)

Black-Scholes Differential Equation on a TI-calculator

2006-04-05T20:29:00.000+02:00

I just put the proof of the Black-Scholes Differential Equation using a TI calculator on wikipedia:

My mediawiki contributions

2006-03-25T12:57:00.000+01:00

I contribute to a lot of mediawikis:

my fr.wikibooks home page.
my en.wikibooks home page.
my fr.wikipedia home page.
my en.wikipedia home page.

And to the matlab newsgroup.
I believe that the Web3.0 will be a large mediawiki; wait and see...

Abramowith & Stegun's Handbook of Mathematical Functions

2006-03-24T18:54:00.000+01:00

It's online at ConvertIt.com:

Numerical Methods in Finance: A MATLAB-Based Introduction

2006-03-23T08:54:00.000+01:00

Here is an interesting book writen by Paolo Brantimarte:

The sources are downloadable.

Les paradoxes de la croissance française

2006-03-22T20:32:00.000+01:00

An interesting article by Daniel Cohen (not Elie Cohen): Les paradoxes de la croissance française.

He cite some reports by Bart van Ark

Google finance

2006-03-22T09:38:00.000+01:00

Another Google new beta service:

which is a very interesting free financial website, two weak points:

only the US market is available
no data download at this stage

YubNub.org for finance

2006-03-13T11:10:00.000+01:00

YubNub is a wonderfull tool to create web shortcuts.
I use some of them for finance, for instance:

ybchart FP.PA to obtain a 5 days candlestick yahoo finance chart.

ybc FP.PA to have only the image,

they can be combined with split: split {url ybc GAZ.PA} {url ybc SZE.PA}, to give this wonderfull comparison of charts

Matlab problem for database access

2006-02-26T08:36:00.000+01:00

Matlab SP3 includes a java memory leak in the database access main function : "fetch".

Matlab bug report:

Summary		fetch Function Fails with Java Exception
Report ID		284377
Date Last Modified		09 Jan 2006
Product		Database Toolbox
Exists In Version		3.1
Exists In Release		R14SP3, R14SP3+
Fixed In Version		OPEN
Fixed In Release		OPEN

The effect of the problem is very simple: the (bad use of) java code accessing a database cannot free memory. Monitoring java-available memory during database access give such behavior (for a simple "select" sql instruction):

??? Java exception occurred: java.lang.OutOfMemoryError: Java heap space. Error in ==> cursor.fetch at 109 dataFetched = dataFetch(fet,resultSetMetaData,p.NullStringRead,p.NullNumberRead); or Exception in thread "AWT-EventQueue-0" java.lang.OutOfMemoryError:

Statistics and sociology

2006-02-08T08:31:00.000+01:00

Stratégie non paramétrique pour évaluer une modélisation statistique: la comparaison des distributions de résidus entre populations à partir du test de Kolmogorov-Smirnov

Transparent de l'exposé du 31 janvier 2006 par M. Parodi et C.A. Lehalle aux Ateliers M2S3 (Les Ateliers de Modélisation et de Méthodes Statistiques en Sciences Sociales).

De nombreuses questions de sociologie quantitative peuvent se ramener au cas d’une comparaison entre deux populations pour savoir si elles se distinguent significativement sur tel ou tel aspect mesurable. Par exemple, l’inégalité d’accès aux grandes écoles selon l’origine sociale est-elle ou non différente d’une génération donnée à une autre ? Est-elle différente entre les populations masculine et féminine ? En outre, il est fréquent de recourir à une modélisation (unidiff dans l’exemple) pour estimer un indicateur doté de propriétés intéressantes (par exemple, l’invariance quant à certaines variables structurelles). Habituellement, la méthode mise en œuvre repose sur des tests paramétriques. Nous proposons ici une méthode générale et non paramétrique pour mener à bien ce type d’investigations. Elle consiste à estimer le modèle sur une population, de conserver les résidus, puis de l’appliquer à l’autre population pour en extraire également les résidus. Le test de Kolmogorov-Smirnov est alors tout indiqué pour repérer un écart significatif entre les deux distributions de résidus et, du même coup, pour répondre de manière fine à la question initiale. Les avantages de cette procédure sont nombreux, et les possibilités multiples. Ainsi même avec un mauvais fit du modèle, la comparaison des résidus entre populations conserve tout son intérêt et son sens. Cette méthode peut permettre également de juger à partir de quand on modélise du bruit. Nous appliquerons cette méthode sur plusieurs exemples: un modèle unidiff et une régression sur une variable à expliquer continue.

The main goal of our speach was to explain that the product of a modelling task was of course the parameters of the obtained model, but also the residuals (differences between the model output and the orginial dataset).
We think that usually, people only look at residual once (to quantify the likelyhood of the model), and then use the model as if it was a perfect abstraction of the studied phenomemon. On the contrary we try to continuously use the information contend into the residuals.
This allow to

compare two unbalanced groups (in term of number of individuals) ;
compare the influence of being a member of one group without any hypothesis about the nature of this influence (non parametric approach).

My lecture during a finance course at HEC

2006-02-06T08:41:00.000+01:00

I gave a lecture with M. Denize during the course of finance of Thierry Foucault at HEC.
We spoke about liquidity of the markets, expertise of brokers, risk packaging on markets, controlling some aspects of the market impact.

Du café dans les banlieues ?

2005-12-09T13:24:00.000+01:00

décompression mathématique 2bis - (french)

Les industriels et les mathématiciens se parlent

2005-11-16T14:02:00.000+01:00

A l’initiative de la SMAI et du CNRS, un cycle de demi-journées a organisé pour présenter les applications des mathématiques dans l’industrie et les services de façon à renforcer les liens entre mathématiciens et industriels et à les étendre à de nouveaux domaines. Ceci concerne les mathématiques "déjà" appliquées (aux premiers rangs desquelles le calcul scientifique, les statistiques, la modélisation, l’optimisation...), mais aussi des mathématiques plus traditionnelles (géométrie, algèbre, analyse, logique, systèmes dynamiques...).

Un séminaire tout à fait intéressant qui a été un de mes premiers points de reprise de contact avec le monde de la finance des produits dérivés...

Manipulations de données financières avec MATLAB

2005-11-07T12:12:00.000+01:00

Une présentation que j'ai faite à un séminaire MATLAB, l'intro:
"MATLAB est un outil logiciel reconnu et exploité avec succès dans l'industrie (automobile, aérospatiale, défense), alors que son utilisation reste plus récente en finance. Le groupe de recherche en Trading et Arbitrage d'Exane-BNP-Paribas a mis en place une architecture de développement sous MATLAB qui va de la lecture de données financières au fil de l'eau jusqu'à la restitution de résultats à forte valeur ajoutée pour les opérationnels. Cette architecture permet de profiter des atouts reconnus de MATLAB comme le prototypage rapide, la disposition de larges librairies statistiques et de traitement du signal, et le déploiement. Nous présentons l'approche générique qui soutient cette architecture, ainsi que quelques exemples de réalisations métier."

My presentation (in french, powerpoint format) at a Finance seminar organized my MATHWORKS.

PhD thesis online

2005-07-19T10:13:00.000+02:00

The web site thèses-EN-ligne allows to read a lot of french PhD thesis.

Get Cited

2005-07-13T08:35:00.000+02:00

I found this web site where I put my own profile.

It's official goal is :
"getCITED is an online, member-controlled academic database, directory and discussion forum. Its contents are entered and edited by members of the academic community. By putting its content in the hands of its members, getCITED makes it possible to enter in and search for publications of all types. This means that, in addition to the books and articles accessible with other databases, book chapters, conference papers, working papers, reports, papers in conference proceedings, and other such research outlets can all be entered and then searched for within getCITED. In addition, getCITED makes it possible to link publications with all the publications in their bibliographies, thereby making possible a wide variety of publication and citation reports."

I will use it to update my bibliography.

Unfortunately, there are not a lot of french people and institutions in it...

Et s'il suffisait de tout mémoriser ?

2005-06-09T13:23:00.000+02:00

décompression mathématique 2 - (french)

Some papers of mine

2005-04-28T18:05:00.000+02:00

Here are 3 papers of mine, co authored with Robert Azencott (and Jérome Besnard for one of them).

Techniques auto adaptatives sur signaux acoustiques et vibratoires pour le diagnostic en ligne et le monitoring de moteurs	ON THE FLY HEALTH MONITORING OF MECHANICAL HAZARDS FROM UNDER SAMPLED SIGNALS IN FORMULA ONE
C.A. Lehalle, R.Azencott, J. Besnard	C.A. Lehalle, R.Azencott

Monitoring auto adaptatif de Procédés Industriels Complexes
C.A. Lehalle, R.Azencott

Software framework for mathematical signal processing

2005-04-10T20:18:00.000+02:00

I currently work on an open-source framework dedicated to rigorous mathematical applications to signal processing. It is used as base for a wiki-book in french on this topic, available at WikiBooks.

This framework (in the R programming language) is focussed on two main points :

the ability to demand data from different sources: at least a learning set and some test sets (the learning set could be automatically generated from a bufferization of observed data),

the management of data, algorithms, external parameters, internal parameters (estimations), and connection between algorithms on different levels.

Abnormality detection from industry to finance: a data flow point of view

2005-03-13T20:38:00.000+01:00

Actually a lot of abnormality / defects detectors can be seen that are based on modelling the differences between a normal behavior and a defectuous one. See for instance the GS-Watch [Goldman Sachs, 1998].
The main difficulties to set up such approaches are:

(1) the need of an explicit point of view of the mecanisms occuring during normal phases of the behavior

(2) a lot of recordings of defectuous cases.

moreover, such an approach delivers a model that (3) can only identify abnormalities / defects of the same nature than the defects that have been recorded for (2).

Those 3 points can be rigorously proven.

Another approach has been successfully tested [Lehalle & Azencott, 2004]: a detection based on empirical modelling of structural dependencies between all available data. Once a rough model has been built, the distance between observed situations and predicted ones (according to the empirical model) is renormalized to produce an alarm if it is too high.

The advantage of this approach is that it mainly relies on recording of normal cases, and is capable to detecet any deviation from a normal behavior. One other important point is that a sharp modelling of underlying behaviors is often not needed.

References.

[Goldman Sachs, 1998] GS-WATCH - A New Framework for Predicting Financial Crises in Emerging Markets, Alberto Ades, Rumi Masih, Daniel Tenengauzer, Goldman Sachs Economic Research, 1998 <google link>.

[Lehalle & Azencott, 2004] On the Fly Health Monitoring of Mechanical Hazards from Under Sampled Signals in Formula One, C.A. Lehalle and R. Azencott, FISITA 2004 <fisita archive>.

Applied statistics to financial risk analysis

2005-02-17T18:22:00.000+01:00

I took some time last week-end to tune the course I give at ENSAE. The main points actually are:

include selected prooves of some results (truncated standard deviation, the effect of a trend on a shock detector)

improve the multi dimensional garch model part

improve the wavelet analysis part

blue: std of truncated singal, green: real std, red: corrected std

Technical Analysis on financial data

2005-01-12T13:04:00.000+01:00

Financial institutions will be able to use the TraderForce to access technical analysis of financial data provided by a soft MIRIAD produced. I was deeply involved into the development of its algorithmic core.

One of the important point is that the algorithmic core is able to recognize 22 types of figures on 40.000 signals over 3 years in only 8 hours of CPU.

Its features are based on multiscales analysis using dedicated mathematical results. We worked a lot 3 years ago to obtain those results, this peace of software has been used during 1.5 years by diagonline, and know it's traderforce who decided to use it.

This end of year is quite busy

2004-12-18T09:58:00.000+01:00

I went to the SAFEE meeting at BAM (Berlin) this week.
It was interesting for the topics we spoke about, and for the way coordinators manage european projects of such a size.
A lot of confidential topics were discussed. One of them was simply about the status of "european-project classified": what does it really implies ? (an open discussion)

Applied statistics for financial risk

2004-11-29T13:14:00.000+01:00

I proposed a subject for the "applied statistic" course at the ENSAE (Ecole Nationale de la Statistique et de l'Administration Economique - National Scool for Statistics and Economic Administration) about "empirical correlation estimates in finance". I will follow Litterman approach in "Estimating covariance matrices, Robert Littermann Kurt Winkelmann, Goldmann Sachs Risk Management Series".

SIA "fiability and availability of production lines" day

2004-11-24T21:10:00.000+01:00

Yesterday, I went to the "fiability and availability of production lines" day of SIA (french society of automotive engineers) to preent a paper entitled "auto adaptive monitoring of complex industrial processes", co authored by Pr. Robert Azencott. You can get a copy of my slides (english) here.

The first half of the day was dedicated to global methodologies (like SPC ; five presentations) and the afternoon to technologies (like the one I presented, 4 presentations).

Here are some points on with speakers put a real emphasis :

From a process point of view: "if something worked fine once, copy it ; only invent a new methodology when it's really needed".

The "design to stability/efficiency" paradigm recommand the confrontation of a project goals and existing standards (with two closed loops: one dealing with continuous improvement of existing and applied standards, the other with validating new standards).

The ackownledgment letter I recieved :

Bourbaki seminar of nov 2004

2004-11-21T19:13:00.000+01:00

I only went to two lectures at the Bourbaki seminar of this week-end :

Xavier Buff : La mesure d'équilibre d'un endomorphisme de Pk (equilibrium measure for an endomorphism of Pk).

The lecture succeeded in giving a very intuitive view of the Briend and Duval results on such endomorphisms, using a lot of figures of fractals drawings.

The available paper is very complete and comprehensive, giving a lot of interesting proofs and results.

Olivier Biquard : Métriques kahlériennes à courbure scalaire constante : unicité, stabilité.

Unfortunately too technical for me ; I'm not familiar enough with this theory to have been able to really understand the main points of the lecture.

SIA (Automotive Engineer French Society) seminar

2004-11-18T21:01:00.000+01:00

I went to a SIA seminar today, about the use of statistics in automotive industry. Three groups of eight experts were formed : one to think about the use of statistics during the conception phase, another focussed on the industrial faisability phase, and the third one to deal with the production phase. I was part of the latest.

We studied different occuring situations and different potential solutions, and scored them in terms of usability, efficiency, and commonality. Our small think tank concluded and ordered main statistical methodologies this way :

Database structuration

Statistical process control

Design of experiments

Classical data analysis

Repetability and reproductibility measurements
(in french : Répétabilité et Reproductibilité)

Quality control via sampling

Capability measurements (CPK)

Next time we will work on the good practices needed to efficiently use the three first of them.

Online access to my patents

2004-11-14T16:33:00.000+01:00

Here is an online access to a fulltext version of my patents at the European Patent Office.

Informations about the automotive industry

2004-11-09T09:53:00.000+01:00

Here are some downloadable reports about the automotive industry:

difficult to download...

First real post

2004-11-06T15:43:00.000+01:00

This will be my professional blog.
I will try to post here facts and events related to my professional interests.

It is mainly dedicated to be a complement to my linekdin profile.

I will write posts and date them before today to reflect some events in the past.

OCAMAWEB release 6.01

2004-10-21T18:24:00.000+02:00

The new release of OCAMAWEB is available at sourceforge.
I am it's main developer at MIRIAD Technologies, it's our tool for literate programming, now used by a lot of people all over the world.

SIA congress conclusion

2004-10-19T18:03:00.000+02:00

As conclusion to the SIA congress named Confort Automobile et Ferroviaire: Acoustique, Vibratoire, Thermique (Automobile and Railway comfort: Acoustics, Vibratory, Thermal), where I presented a paper, I recieved this letter:

SIA congress

2004-10-13T18:29:00.000+02:00

I went to the congress Confort Automobile et Ferroviaire: Acoustique, Vibratoire, Thermique (Automobile and Railway comfort: Acoustics, Vibratory, Thermal) to present a paper co-authored with R. Azencott and J. Besnard:
auto adaptive techniques on NVH (noise vibration harness) signals for online diagnosis and monitoring of engines.

FISITA'04

2004-05-06T18:03:00.000+02:00

I presented a paper co-authored with R. Azencott at FISITA'04:
On the Fly Health Monitoring of Mechanical Hazards from Under Sampled Signals in Formula One

I encountered there a lot of interesting people in the fields of:
- biomass to fuel
- accidentology
- software design for automotive
- probe sensors

Un nombre premier à 6,320,430 chiffre a été identifié le 2 décembre 2003

2004-01-09T13:20:00.000+01:00

décompression mathématique 1 - (french)

Publications quoting my work

2001-12-01T13:32:00.000+01:00

A book citing some of my articles:

R. Johansson and A. Rantzer (Eds.), Nonlinear and Hybrid Systems in Automotive Control, Springer-Verlag, London, 2003, ISBN1-85233-652-8, (440pp.).

And those two papers (BibTex format):


@inproceedings{BemporadRL:01ba,
    author = "A. Bemporad and Jacob Roll and Lennart Ljung",
    title = "Identification of hybrid systems via mixed-integer programming",
    year = "2001",
    month = "Dec",
    address = "Orlando, USA",
    booktitle = "Proc. {IEEE} Conf. Decision and Control",
    pages = "786-792",
}
@inproceedings{Roll:01ab,
    author = "Jacob Roll",
    title = "Identification of Piecewise Affine Wiener Models via Mixed-Integer Programming",
    year = "2001",
    month = "Mar",
    address = "Norrköping, Sweden",
    booktitle = "Third Conference on Computer Science and Systems Engineering in Linköping",
    pages = "297-304",
}

Old publications

1999-06-06T17:38:00.000+02:00

Some old publications:

Piecewise Affine Neural Networks and Nonlinear Control, C-A. Lehalle, R. Azencott, ICANN'98

Initialization of Piecewise Affine Neural Networks for nonlinear control, C-A. Lehalle, R. Azencott, NOLTA'98

Piecewise affine neural networks and nonlinear control: stability results, C-A Lehalle and R Azencott, ICANN'99

How Piecewise Affine Perceptrons can emulate a stable nonlinear control, C-A. Lehalle, R. Azencott, ISIC/ISA