I retired from personal blogging in July 2008.
But you can find me over at http://blog.xero.com.
I was delighted to speak at the Spark $40k Challenge at Auckland University last night.
http://www.spark.auckland.ac.nz
Sometimes it’s easy to think that we’re not doing enough from New Zealand but there were 80 entrants with 240+ people involved. That is a great sign.
In my speech I covered 3 things.
- Why it’s important for New Zealand that we foster entrepreneurship. This where a large percentage of new jobs will come from.
- That entrepreneurship should be thought of as a series of steps. Each one building experience and networks etc. There are a lot of internal entrepreneurship opportunities and that it is also important to gain experience inside larger organisations.
- That the journey is rewarding and fun
The team running the Spark program was impressive and have been very generous with their time. Congrats Darsel and Co. A great program.
The twelve finalists were:
Stroke Education : this venture promises a comprehensive education strategy for stroke survivors to quicken recovery and improve quality of life. It has been designed by an esteemed University researcher.
Nerian : a venture by researchers from the School of Biological Sciences to isolate the healthy antioxidants in red wine and put them in a range of value-adding products.
Xchange: a website to facilitate person-to-person private loans, bringing together borrowers and lenders in a secure and social environment, with the ability to choose your own interest rates.
Test Tube TV: an online ‘jamming’ platform for musicians. This venture uses ‘podscope’ to allow users to make music with their favourite artists, as well as other users.
Bacteria Inc : this team has designed a non-chemical system to vastly improve the settling of sludge in the wastewater treatment industry, using innovative software and patent protected bacteria.
595 Lab: an online social network for researchers, allowing interaction and collaborations with like-minded colleagues from around the world.
DoshMob: New Zealand ’s first online marketplace for peer-to-peer lending Borrowers and lenders interact through auctions, with interest rates decided by the marketplace.
Fixen Ltd: Fixen sets out to create healthy chocolate products, giving consumers a rich, satisfying chocolate experience without the adverse effect of over indulgence.
Metal & Steel: this team hopes to establish the gateway for all metal and steel industry sales by setting up a live, online trading environment, allowing businesses to reach global import and export traders.
Meter & Bill: a tool for service providers, such as IT Departments, to map the complexities of their resources and services in order to streamline and improve their value to customers.
Steritech: this team uses patented Pressure Assisted Thermal Sterilisation (PATS) technology to sterilise food and beverages at lower temperatures, with significant nutritional, taste and cost benefits.
I noted two main types of companies. Those that were taking technical intellectual propery and commercialising it, and those that had good marketing and business skills and were looking at importing overseas ideas and models into New Zealand. The second group of people were probably the better presenters - as you would expect.
I’d really like to see those very capable business students reaching out to the internal technology groups around the university and applying their excellent skills to comercialising home grown technology for export. Linking these two groups together would be a powerful thing.
The event was held in the new Auckland University Business School. The $110m project is almost complete and is world class. It did make you want to get involved.
All in all a great event and it made me very excited about the future. Congratulations to the finalists.
blog
RKI: PhD students have developed patented technology to send high quality images as small files to any mobile phone or computer, irrespective of the software on those devices and without sacrificing image quality.
I am pretty confident that RKI’s patented technology is a type or a variant of wavelet. The JPEG 2000 previously used Discrete Cosine Transform (DCT), but they have dropped it and is now adopted a wavelet-based image compression standard which is far superior than DCT-based compression. Wavelet is a popular topic in Signal Processing for academic research and industrial commercial application today, where it has its own journals (currently about 5, best of my knowledge) dedicated for publishing researches in wavelet.
The first prominent industrial application of wavelet that caught the public’s attention was in 1993, when Los Alamos National Lab developed a fingerprint image retrieval system for the FBI(briefly described here by one of the scientist in that project). The FBI’s new wavelet-based image retrieval system improved dramatically of how images are sent and stored.
Its uses is now popular in Hollywood for production of animated films ( first used in Bug’s Life and Toy Story in the mid-1990s). I wouldn’t be surprised in Peter Jackson & Weta are using wavelet or not, but if they aren’t yet, I am sure they will adopt it at some stage. Nothing can beat wavelet in producing quality image & graphics today. There is no doubt that all the animated films that are coming out of Holywood , such as Shrek, etc, involved wavelet in their making.
I’ve only used wavelet for time-series burst detection (automated detection of discontinuity or jump pattern in financial time-series, where it could be forewarned in advanced before it happened) but I haven’t used wavelet for any else, since its application have been applied in many areas. The only time that I came close to developing wavelet for anything else apart from time-series application, was a project in an Auckland company that specializes in spam detection , internet security for children, network security, etc. The company canned the project before development even started, but initially they wanted to a system to scan websites that contain Objectionable materials and automatically blocked those urls from young users (children) in getting access to them. The system was to be developed based on the architecture described in the following paper. Warning:- the following paper contains some nude pictures that might be offensive to some readers.
System for Screening Objectionable Images
So, I would like to bet that RKI’s technology must be a type of wavelet. There are many types of wavelets, and new ones keep appearing in literatures from time to time.
I have come across a discussion paper from the NZ Reserve Bank, that researchers there used wavelet for econometric modeling.
Spectral analysis of New Zealand output gaps using Fourier and wavelet techniques
Perhaps, RKI team might want to poach staffs from the Reserve Bank (just joking), assuming that their (RKI) technology is wavelet-based.
The Spark Committee would like to thank Rod for coming and helping us celebrate our qualifiers for 2007.
Spark has done many incredible things for entrepreneurship since it was started in 2003 and this year has been no exception. It is a student led organisation run to promote an entrepreneurial culture and a spirit of enterprise and innovation at The University of Auckland. The Qualifiers Ceremony on Thursday night was a great showcase of this.
We received 80 entries by around 240 individuals who all are keen to become entrepreneurs.
If you go to http://www.spark.auckland.ac.nz you can find out more information or contact Darsel at spark-ceo@auckland.ac.nz.
The support of people like Rod help make all of this possible.
We would love to have more people like him involved so if you are keen me in touch.
A big thanks to Rod from Darsel and the 2007 Spark Organising Committee
It is interesting that there are two finalists based around the debt marketplace model as pursued by prosper.com
prosper.com
Umm, interesting concept. From prosper.com ’s website it says this:
The way Prosper works is intuitive to people who have used eBay. Instead of listing and bidding on items, people list and bid on loans using Prosper’s online auction platform.
I would encourage these finalists to get in touch with Dr. Mike Barley (via UniServices) to seek his expertise in designing Software Agent systems for your online auction application. However if you feel that such outside help is not needed, then I would highly recommend this very popular open source Java project for Autonomous Software Agent Systems, called JADE. I had only used JADE once in a project that I was involved about 6 years ago. So, I haven’t looked at it again since then, but I will use it for something that I am currently developing (financial market analytics). I have no experience in online auction development but I have come across numerous papers that describe the use of software-agent in online auction application. Ref number #1 (see link) briefly describe using JADE on such system. JADE has a very active community of users. Lots of written tutorials from the community. I believe that if you develop software agent to your system, it is a world beater.
#1) A Trustworthy Agent Based Online Auction System
#2) Real-Time Trust Management for Agent Based Online Auction Systems
#3) An Agent-based Platform for Online Auctions
If you want more online info on the topic, then try searching for papers at citeseer. There are tons of freely donwloadable papers at citeseer on software agents or any computing topics really. Citeseer is a repository of peer review papers that have been submitted by researchers there to be publicly made available.
You might also want to look at WEKA Java open source machine learning project, because the trend these days in software agent development based on titles that I have come across, has been shifted into making the software agents learn its environment automatically and adapt itself to new inputs (which were not foreseen by the developer or programmer). These new trends where some called learning agents, and there are numerous publications at citeseer.
Ref number #3, describes the use of Markov Chain (MC) in agent-based Platform, there is no direct MC algorithm in WEKA itself but look at the WEKA website for algorithm extension, because other researchers have extended WEKA capabilities by developing new algorithms which are not core WEKA. However there is numerous open sources in MC algorithm available from the internet.
I use Markov Chain algorithm for financial risk analysis. In fact Standard & Poor’s credit rating is developed using Markov Chain. There are different variants of Markov Chain in use today and their robustness (accuracy) do vary, so I don’t know which version that Standard & Poor’s is using. Interestingly, a form of Markov Chain called Hidden Markov Model (HMM) is popular in speech recognition software development of today and Bio-informatics DNA sequence analysis . Google PageRank algorithm was formulated using Markov Chain. If you’re interested to know more about Markov Chain and its use (perhaps in regards to your online auction system), then I am happy to point out more info Markov Chain.
Hi,
I am trying to develop models for the auction and bidders’ strategy using Markov chain.
Would you please point me out more info on MArkov chain for these applications.
Jagannath
Jagannath,
Here are 2 papers (freely downloadable - PS or PDF format) that I found which cover the application of Markov Chain to online auction:
- An Adaptive Agent Bidding Strategy based on Stochastic Modeling
- Optimal Design of Internet-based Auctions
I am sure that there are more out there, perhaps you can try to search in established academic & research-based publishers web sites such as Springer, Elsevier, IEEE Transactions series or ACM . If you do find some, then you can email the authors directly (which their email are usually published with the abstracts of their publication) requesting free copies of their papers. That’s what I do, if I can’t find any freely downloadable version of that specific publication online. If the author refuses to send you a free copy because he/she insists that you must buy the article online, then try looking it up at your local University library to see if the particular journal (and its volume number) is available, where you can photocopy it. Most of the journals (in all disciplines) from those publishers I have mentioned above are available at any University library.
If you need some clarifications on those publications, just track down the author(s), via Google mostly or from their email addresses which are available from their papers and ask them directly. In my experience, authors are very generous to answer any question from anyone who is trying to implement their published algorithm.
There are some Markov Chain based free codes available on the internet (in different languages), that you can grab and modify for your application, instead of trying to reinventing it. I use the Matlab MCMC (Markov Chain Monte Carlo).
Jagannath,
The following might also be useful . It is freely downloadable.
Profiling Price Dynamics in Online Auctions Using Curve Clustering
Authors from the document shown above used eBay data but they also quoted other research works where those authors collected their data for their work by using a crawler. Unless, you’re from the online auction vendor itself where you get access to the vast amount of historical data, collecting data via a crawler is daunting since you will have to process the collected data prior to your modeling. If you’re from an online auction vendor such as TradeMe, then you have no problem. But if you’re an independent online trader or a potential competitor to TradeMe (perhaps you & your group are working secretly somewhere), then you must use a crawler to crawl your competitor’s site (I assume it would be TradeMe). There are free crawlers already available on the net.
You might want to download this free Matlab Curve Clustering Toolbox if you intend to implement the models described in the paper above. The toolbox is not from the authors of the online auction paper shown above, but it is the same algorithms described in the Price Dynamics… Online Auctions… paper.
You might also look at the Pspline fortran API, which the authors used in their paper.
I don’t know what platform you’re doing your development in, but I guess, if you do in .NET, then I think that the Pspline API is no problem in calling fortran code directly in .NET (I am not a .NET developer to know about this, I am just assuming). As far as I know, Matlab codes can’t run in a .NET environment, unless you have the Matlab runtime environment in your machine , which you need to buy a Matlab license and its .NET compiler. If you don’t want to do that, then your only alternative is to port the Matlab codes into C#, using this .NET matrix algebra package. You need a matrix algebra package since linear algebra algorithms are used in the Curve Clustering Toolbox, that’s why you need that .NET matrix package. The .NET matrix package is a direct porting of freely available JaMA (Java Matrix Algebra). By the way, JAMA is Matlab in Java (not quite), so JAMA (java) codes are almost similar to that of Matlab. JaMA was developed my MathWorks (Matlab owner vendor), but they have relinguished the rights to JaMA where anyone can use it , either commercial or non-commercial.
If you do it in Java, then I guess that you have to use JaMA to port the Curve Clustering Toolbox into Java, and write native Java codes interface to call the Pspline fortran API directly.
I use JAMA in all my Java development, where I have extended it with more functions and classes that were missing from JAMA itself. My extended JAMA (which I call JAMAX) is much closer to Matlab (function for function) than JAMA. If you’re interested in my package, then I can sent you a copy. Don’t worry, its free.