For more theme orientation and a better rhythm, Andreas Binder and Michael Aichinger will post Wednesday and Friday. They will write with the eye of a mathematician / physicist.
Their posts will not be limited to advanced methods and schemes, but about various topics around quant finance. And not surprisingly, if models and methods are involved, they will be most probably be implemented and tested in Mathematica.
The first 2 posts can be found at UnRisk Insight Mathematics.
What I Have Done Wrong - Five Lessons I Have Learned.
The only constant is change, Heraclitus.
Exciting things happened in the recent days. They motivated me to change my business compass. marketing strategies and product positioning again.
Time to look back a little - analyzing how I failed. Before I started my own business (1989), I worked for Austria's largest industry. My final responsibility: Factory Automation Software. I have described my technological background and skills here.
When you start your own business, it's just you and your ideas. My basic idea was: Work at the intersection point of mathematics and computer science, build a network of distinguished technology partners and transform the joint knowledge into margins.
You need to add customers. If you are lucky you combine the right product with the right market segments. You need to define the right marketing and promotion mix. What most of the entrepreneurs underestimate: you need to reinvent yourself quite often.
Work on things that matter and for those who care
Unfortunately, it is not that simple. We decided early for Open Innovation (in-licensing and out-licensing), do lab work, not industry work, ... but you need a brand promise, and don't forget to develop and maintain a company culture.
Did I get anything right?
Yes, I think the technology selection, the partner plans and the principle product-market fits. And I knew from my work at industry that you need to learn from turbulences and how to manage innovative teams.
Here is what I did wrong in the past and how we changed that
1. Don't diversify in a crisis
First, we decided to focus, from the beginning. If you want, we are a Mathematica and UnRisk company.
You have ups and downs and when the .com bubble bursted and at the beginning of the financial crisis, we suffered from revenue reductions. At the first I thought diversification will help. But if a crisis is caused by a drastic change of market correlations, diversification does not work.
After the financial crises, we focussed. And it worked so well.
2. Don't fall into the eager sellers and stony buyers trap
If you are bursting of ideas and have clever procedures and tools to turn them into products swiftly, you want to release new things often.
But the eager sellers and stony buyers principle tells us that the mismatch is nine to one: sellers overrate their innovations threefold and buyers overrate threefold what they have.
We overcame this by making the improvements really big and delivering know-how packages instead of software only. The UnRisk Academy was established to extend product use training with courses giving full explanation on what is behind the curtain.
3. Don't work to get picked - choose yourself
I just refer to one of Steve Jobs' strategic success factors: don't make profits make products.
In search of perfection, we ask ourself "is this the best we can do?". And there is the trap of striving for the perfect competence, leading to industrial work - you forget to disrupt yourself and you miss possibilities.
Work to get picked, is reactive - if you choose yourself, you set the pace. I have really underestimated the power of this view change.
4. Don't forget the numbers
I underestimated the financial side of the business. Even, if you are an estimated innovator, build great products, have selected your target markets right, a well received brand, ... you are still forced to understand the financial impact of your decisions - don't fly blind.
We are a comparatively small outfit, but I have built a system that tells me daily where we are. And "where" means business units across the cost structure (P&L) with bottom line at risk estimates (a simple color table).
And it is vital to integrate the teams into the control flow.
5. Don't take care of that all
I have managed an innovative business in an innovative industrial environment before. I reported directly to the board level. And, starting the new company, I thought I need to do it all myself. My first instinct was to not to hire a team ...
But management is to achieve your goals through your teams. And an indispensable driving force for innovation is co-creation and co-operation.
I now put energy in communication with the teams. My fear of shifting the status quo is much bigger than the fear to lose time in communication.
We want to continue creating more value than we can capture and I know I will fail again, but we have a real stable foundation now.
We will continue helping our clients to conserve capital, build quantitative skills and leverage technology.
Exciting things happened in the recent days. They motivated me to change my business compass. marketing strategies and product positioning again.
Time to look back a little - analyzing how I failed. Before I started my own business (1989), I worked for Austria's largest industry. My final responsibility: Factory Automation Software. I have described my technological background and skills here.
When you start your own business, it's just you and your ideas. My basic idea was: Work at the intersection point of mathematics and computer science, build a network of distinguished technology partners and transform the joint knowledge into margins.
You need to add customers. If you are lucky you combine the right product with the right market segments. You need to define the right marketing and promotion mix. What most of the entrepreneurs underestimate: you need to reinvent yourself quite often.
Work on things that matter and for those who care
Unfortunately, it is not that simple. We decided early for Open Innovation (in-licensing and out-licensing), do lab work, not industry work, ... but you need a brand promise, and don't forget to develop and maintain a company culture.
Did I get anything right?
Yes, I think the technology selection, the partner plans and the principle product-market fits. And I knew from my work at industry that you need to learn from turbulences and how to manage innovative teams.
Here is what I did wrong in the past and how we changed that
1. Don't diversify in a crisis
First, we decided to focus, from the beginning. If you want, we are a Mathematica and UnRisk company.
You have ups and downs and when the .com bubble bursted and at the beginning of the financial crisis, we suffered from revenue reductions. At the first I thought diversification will help. But if a crisis is caused by a drastic change of market correlations, diversification does not work.
After the financial crises, we focussed. And it worked so well.
2. Don't fall into the eager sellers and stony buyers trap
If you are bursting of ideas and have clever procedures and tools to turn them into products swiftly, you want to release new things often.
But the eager sellers and stony buyers principle tells us that the mismatch is nine to one: sellers overrate their innovations threefold and buyers overrate threefold what they have.
We overcame this by making the improvements really big and delivering know-how packages instead of software only. The UnRisk Academy was established to extend product use training with courses giving full explanation on what is behind the curtain.
3. Don't work to get picked - choose yourself
I just refer to one of Steve Jobs' strategic success factors: don't make profits make products.
In search of perfection, we ask ourself "is this the best we can do?". And there is the trap of striving for the perfect competence, leading to industrial work - you forget to disrupt yourself and you miss possibilities.
Work to get picked, is reactive - if you choose yourself, you set the pace. I have really underestimated the power of this view change.
4. Don't forget the numbers
I underestimated the financial side of the business. Even, if you are an estimated innovator, build great products, have selected your target markets right, a well received brand, ... you are still forced to understand the financial impact of your decisions - don't fly blind.
We are a comparatively small outfit, but I have built a system that tells me daily where we are. And "where" means business units across the cost structure (P&L) with bottom line at risk estimates (a simple color table).
And it is vital to integrate the teams into the control flow.
5. Don't take care of that all
I have managed an innovative business in an innovative industrial environment before. I reported directly to the board level. And, starting the new company, I thought I need to do it all myself. My first instinct was to not to hire a team ...
But management is to achieve your goals through your teams. And an indispensable driving force for innovation is co-creation and co-operation.
I now put energy in communication with the teams. My fear of shifting the status quo is much bigger than the fear to lose time in communication.
We want to continue creating more value than we can capture and I know I will fail again, but we have a real stable foundation now.
We will continue helping our clients to conserve capital, build quantitative skills and leverage technology.
The Debt-to-GDP Cliff of 90% Does Not Exist
Mid Apr-13 I read about this here. And many quality media have discussed it ...
Not necessary to repeat that the results of the famous academic paper were often used and influenced important economic and political decisions ... Interesting that the mistakes were discovered by a student.
A few days ago, Michael Kelly, Wolfram Technology Group, published Is High Debt Bad for Economic Growth? in Wolfram Blog.
Michael showed that such errors are unlikely when using Mathematica's explicit functional programming style - stability and robustness tests are straight forward.
I totally agree and add that this is one of the reasons why we have decided to integrate our UnRisk derivatives analytics engines (numerically optimized in C++) into Mathematica.
The programming power behind UnRisk solutions is UnRisk-Q - its financial objects, functions and utilities are declarative, functional Mathematica language constructs. Cross-model tests and high level across-scenario tasks are so easy ...
Why is Mathematica not commonly used in economical circles? IMO, because it is not so obvious to see that economic data, models and methods need to be organized orthogonally.
Not necessary to repeat that the results of the famous academic paper were often used and influenced important economic and political decisions ... Interesting that the mistakes were discovered by a student.
A few days ago, Michael Kelly, Wolfram Technology Group, published Is High Debt Bad for Economic Growth? in Wolfram Blog.
Michael showed that such errors are unlikely when using Mathematica's explicit functional programming style - stability and robustness tests are straight forward.
I totally agree and add that this is one of the reasons why we have decided to integrate our UnRisk derivatives analytics engines (numerically optimized in C++) into Mathematica.
The programming power behind UnRisk solutions is UnRisk-Q - its financial objects, functions and utilities are declarative, functional Mathematica language constructs. Cross-model tests and high level across-scenario tasks are so easy ...
Why is Mathematica not commonly used in economical circles? IMO, because it is not so obvious to see that economic data, models and methods need to be organized orthogonally.
The New Is For Now
Browsing through the popular technology publications, I find data science, big data, context technologies, ... Even in the Mar-13 issue of the Wilmott magazine - after all serving the quantitative finance community - there is a cover story about news analytics (serving the big data quants?)
My view to this: big data quants - news analytics a new form of riding the price waves?
Mathematics is not outdated - but not so in at the moment
News analytics is one factor in a multi-factor strategy with white box models based on more and better mathematics. We need data to calibrate and recalibrate them.
But at the moment "nobody" seems to be interested in mathematics?!
Maybe we need to mine psychology?
Unlike used, established, tested, verified, ... the new is always temporary. New is for now. Tomorrow we get another opportunity to make something new - if we choose it. Maybe it has a market?
But we do not pander: A Workout In Computational Finance - it is about mathematical models that are understandable and computational, about solvers that do not destroy their information and about the inverse problems of calibrating and recalibrating them to reality data. Good news: the book is available at Wiley.
My view to this: big data quants - news analytics a new form of riding the price waves?
Mathematics is not outdated - but not so in at the moment
News analytics is one factor in a multi-factor strategy with white box models based on more and better mathematics. We need data to calibrate and recalibrate them.
But at the moment "nobody" seems to be interested in mathematics?!
Maybe we need to mine psychology?
Unlike used, established, tested, verified, ... the new is always temporary. New is for now. Tomorrow we get another opportunity to make something new - if we choose it. Maybe it has a market?
But we do not pander: A Workout In Computational Finance - it is about mathematical models that are understandable and computational, about solvers that do not destroy their information and about the inverse problems of calibrating and recalibrating them to reality data. Good news: the book is available at Wiley.
Why Software Developers Should Know More About Machine Learning
Machine Learning is about decision making. And like any learning it is about understanding things better. Simplified, in a machine learning project, we try to extract models from data. We hope that those models are understandable and computational.
But the purpose varies from analytics, over predictive modeling to control. And there is no way to find one method that fits all. Rule bases, trees, .. are understandable but to make them computational, you need to apply fuzzy variants and to make them more accurate and robust you should apply regularization techniques (numerically optimize fuzzy sets, membership functions, ...).
Neural nets are kind of black boxes, but computational.
To calculate control functions for given goal parameters you need to apply an inverse problem view - as you do it f you calibrate and recalibrate models.
What is the difference between doing a machine learning project and a development project?
In both yo usually deal with data. And data analysis as an exploratory approach should be the begin of anything. But machine learning experts have a different relationship with code than developers; think of the programming paradigms, derived from the best problem decomposition: data-, function-, or object-oriented. To create models automatically, you do not think so much about such things - they are inherent in the structure of the models ....
But more and more monstrous amounts of data are produced by machines and the best way to deal with all that information is by machines.
So, people who do large scale machine learning (like at Google) think similar to software engineers that need to design large scale industry type systems.
What about software analytics?
Yes, there is another aspect. Modern analytic and model-based approaches in software engineering aim on the automatic generation of software from domain models, ... For this task program code is data.
We made some experiments with defect prediction in large industrial software systems (static core analysis) and found out that fuzzy decision tree methods fit quite well to this purpose.
The experiments have been made with our mlf.
So, software developers may enrich their development methodology with machine learning and apply machine learning in the software quality assurance process - wit a two-sided positive effect.
But the purpose varies from analytics, over predictive modeling to control. And there is no way to find one method that fits all. Rule bases, trees, .. are understandable but to make them computational, you need to apply fuzzy variants and to make them more accurate and robust you should apply regularization techniques (numerically optimize fuzzy sets, membership functions, ...).
Neural nets are kind of black boxes, but computational.
To calculate control functions for given goal parameters you need to apply an inverse problem view - as you do it f you calibrate and recalibrate models.
What is the difference between doing a machine learning project and a development project?
In both yo usually deal with data. And data analysis as an exploratory approach should be the begin of anything. But machine learning experts have a different relationship with code than developers; think of the programming paradigms, derived from the best problem decomposition: data-, function-, or object-oriented. To create models automatically, you do not think so much about such things - they are inherent in the structure of the models ....
But more and more monstrous amounts of data are produced by machines and the best way to deal with all that information is by machines.
So, people who do large scale machine learning (like at Google) think similar to software engineers that need to design large scale industry type systems.
What about software analytics?
Yes, there is another aspect. Modern analytic and model-based approaches in software engineering aim on the automatic generation of software from domain models, ... For this task program code is data.
We made some experiments with defect prediction in large industrial software systems (static core analysis) and found out that fuzzy decision tree methods fit quite well to this purpose.
The experiments have been made with our mlf.
So, software developers may enrich their development methodology with machine learning and apply machine learning in the software quality assurance process - wit a two-sided positive effect.
FEM in Quantitative Finance Matters - But Do Not Apply It Naively
The FEM is a numerical method for solving partial differential equations (PDEs), and has become particularly popular in engineering and physics.
More recently also the quantitative finance community has gained interest to apply this method to the various PDEs arising in the modelling of financial instruments - it is one method of choice in the pricing and calibration engine of UnRisk from the beginning (at a time when it was not common at all in quantitative finance circles).
It is not applicable to every financial model and if, you need to use it with care - in this article it is outlined that mean reverting models for interest rates tend to become numerically difficult in regions sufficiently far away from the mean-reverting level (in the convection-reaction-diffusion PDE, convection dominates the diffusion). We apply streamline diffusion techniques to obtain stable numerical schemes.
In Mathematics of our UnRisk Insight blog you will find a series of posts about FEM (and other mathematical approaches and problems) in quant finance.
They are compilations of chapters of the book A Workout In Computational Finance that has been officially released by Wiley recently.
The workout is built of inspiring sessions about how to enjoy the freestyle of quant work, but staying strong, agile and balanced. It is about the power of mathematica schemes computing financial behavior, traps of wrong method traps and the fitness killer of sloppy implementations.
The owners of then book will be entitled to use a web page, inviting them to play around with methods discussed in the book.
More recently also the quantitative finance community has gained interest to apply this method to the various PDEs arising in the modelling of financial instruments - it is one method of choice in the pricing and calibration engine of UnRisk from the beginning (at a time when it was not common at all in quantitative finance circles).
It is not applicable to every financial model and if, you need to use it with care - in this article it is outlined that mean reverting models for interest rates tend to become numerically difficult in regions sufficiently far away from the mean-reverting level (in the convection-reaction-diffusion PDE, convection dominates the diffusion). We apply streamline diffusion techniques to obtain stable numerical schemes.
In Mathematics of our UnRisk Insight blog you will find a series of posts about FEM (and other mathematical approaches and problems) in quant finance.
They are compilations of chapters of the book A Workout In Computational Finance that has been officially released by Wiley recently.
The workout is built of inspiring sessions about how to enjoy the freestyle of quant work, but staying strong, agile and balanced. It is about the power of mathematica schemes computing financial behavior, traps of wrong method traps and the fitness killer of sloppy implementations.
The owners of then book will be entitled to use a web page, inviting them to play around with methods discussed in the book.
Wolfram Community Has Been Launched
Wolfram announced the launch of Wolfram Community, a new site designed to unite Wolfram users. It is the go-to-place to discuss questions and issues relating to Mathematica and other Wolfram products in a crowdsourced community setting.
Groups are organized across areas, products and functionality.
Groups are organized across areas, products and functionality.
Subscribe to:
Posts (Atom)