Planet Sakai

June 28, 2016

Ian Boston

Referendums are binary so should be advisory

If you ask the for the solution to the multi faceted question with a binary question you will get the wrong answer with a probability of 50%. Like a quantum bit, the general population can be in any state based on the last input or observation, and so a Referendum, like the EU Referendum just held in the UK should only ever be advisory.  In that Referendum there were several themes. Immigration, the economy and UK Sovereignty. The inputs the general population were given, by various politicians on both sides of the argument, were exaggerated or untrue. It was no real surprise to hear some promises retracted once the winning side had to commit to deliver on them. No £350m per week for the NHS. No free trade deal with the EU without the same rights for EU workers as before. Migration unchanged. The Economy was hit, we don’t know how much it will be hit over the coming years and we are all, globally, hoping that in spite of a shock more severe than Lehman Brothers in 2008, the central banks, have quietly taken their own experts advice and put in place sufficient plans to deal with the situation. Had the Bank of England not intervened on Friday morning, the sheer cliff the FTSE100 was falling off, would have continued to near 0.  When it did, the index did an impression of a base jumper, parachute open drifting gently upwards.

Screen Shot 2016-06-28 at 20.08.07

The remaining theme is UK Sovereignty. Geoffrey Robertson QC  makes an interesting argument in the Guardian Newspaper, that in order to exit the EU, the UK must under its unwritten constitution vote in parliament to enact Article 50 of the Lisbon Treaty. He argues that the Referendum was always advisory. It will be interesting, given that many of those who have voted now regret their decision, if they try and abandon the last theme that caused so many to want to leave. The one remaining thing so close to their heart that they were prepared to ignore all the experts, believe the most charismatic individuals willing to tell them what they wanted to hear. UK Sovereignty, enacted by parliament by grant of the Sovereign. I watched with interest not least because the characters involved have many of the characteristics of one of the US presidential candidates.

If you live in the UK, and have time to read the opinion, please make your own mind up how you will ask your MP to vote on your behalf. That is democracy and sovereignty in action. Something we all hold dear.

by Ian at June 28, 2016 07:09 PM

June 27, 2016

Michael Feldstein

LMS Outage: Exclusive view of UC Davis contract with Scriba

When I first heard about the recent LMS outage at UC Davis – which left the school with no LMS access for a full week and without a fully functional LMS through the remainder of the spring term – it was quite clear that this was an unusual situation. There have been plenty of other LMS outages, but with this one not only did UC Davis lose days of system access, during the outage they had no idea if and when the system would be restored. Any useful communication with the vendor ceased, and the school had to scramble and invent their own disaster recovery on the fly.

What we didn’t know at the time was whether UC Davis staff had planned for such a contingency and had reasonable controls in place. Thanks to the media relations and IT teams at UC Davis, we at e-Literate now have the contract with Scriba to help answer these questions. I have also talked to Ian Dolphin, executive director at the Apereo Foundation that runs the Sakai project, and Michael Sanders, CEO of Scriba, to gather and verify the data in this post.

In a nutshell, this was a screwup of colossal proportions by Scriba and Scriba alone. Furthermore, Scriba is no longer a Sakai Commercial Affiliate.

It is worth mentioning that UC Davis never selected Scriba as a vendor. In 2009 they hired rSmart to be their Sakai hosting provider, but rSmart sold their Sakai business to Asahi Net International (ANI) in 2013. In 2015 a private equity firm, Vert Capital, acquired ANI and renamed the company Scriba under unusual circumstances.

rSmart and ANI, 2009 – 2015

From the perspective of a university paying for LMS hosting, the original 2009 contact could best be described as a lousy contract with a good company. rSmart did not position itself primarily as a hosting company, and they were better known for creation of a polished Sakai distribution that customers would manage themselves on-site. Like other Sakai commercial affiliates, they did some customization and some hosting, but not as a core of their business. Perhaps for this reason, the service level definitions were limited to loose language that put more restrictions on UC Davis’s claims and requests than it did on rSmart’s obligations. There was nothing stronger than “commercially reasonable efforts” by rSmart to fix issues, and incidents were only worked during “normal business hours”.

Pasted image at 2016_06_21 09_50 PM

The Support Guidelines listed response times of 2 business hours to 5 business days, depending on the severity of the incident. And the contract excluded Backup & Recovery.

Scriba CEO Sanders described SmartSite as a highly-customized version of Sakai, both in terms of software modifications and deployment architecture. SmartSite uses an Oracle database instead of MySQL, uses its own firewalls, and there is no virtualization. As Sanders described, the system is running on “bare metal servers”. Whether this was the reason for the lack of realistic service levels in the rSmart contract or not, we don’t know. But this customization does play a role in this year’s outage as we’ll see in a bit.

Upon the company purchase by ANI in 2013, UC Davis amended its contract to acknowledge the new company, while keeping the same terms as before. This amended contract ran from 2013 – 2015.

Scriba Corp, 2015 – Present

When Vert Capital acquired ANI and renamed to Scriba, UC Davis amended the contract with all new terms and extended until summer 2017. At the same time, the school decided to run an LMS evaluation and migrate to a new system, so the current contract amendment should be seen as a transition. Owing to the second purchase of the company, UC Davis put in new terms to handle the contingency of Scriba going bankrupt or going out of business – 90 days notice and full provision of systems and data.

But the difference in service levels from rSmart days is like night and day. The amended contract now defines exactly what Scriba (ANI) was agreeing to provide, including new service levels of 30 minutes – 1 day depending on severity, with 24×7 coverage. But here is the smoking gun – UC Davis paid Scriba for 24-hour disaster recovery service that included a live backup application environment with data mirrored at least once a day.


Scriba’s CEO Michael Sanders confirmed by phone that these provisions were in effect, contractually. But there were several massive failures:

1) Scriba gave UC Davis (and other customers) just one day notice of a 2.5 day emergency outage. And this outage was a rip-and-replace movement to a different 3rd-party data center with no overlap of systems. Even if Scriba had met its outage plan, the company would have violated its contract with UC Davis, as they agreed to 6 weeks of notification for changes that could impact system performance, with a complete freeze on changes three weeks at the end of each term.


2) Scriba was not able to recover the UC Davis system within 2.5 days – it took 7 days in total. As noted above, UC Davis had purchased 24-hour disaster recovery with database mirroring at least every 24 hours. Yet Sanders told me that once the outage went longer than expected, Scriba discovered that the mirrored database was out of sync. Scriba then faced a choice – work on the production system or work on the disaster recovery system. They chose to do the former.

3) Scriba did not have adequate staff to handle incidents for a major hosting customer. Sanders described to me how the company did not have an Oracle DBA, as they felt they could not justify one for just a single customer (others on MySQL). So they had a part-time contractor as DBA. The contract shows that UC Davis paid for premium support.


4) Once the outage moved into its fourth day, Scriba ceased communicating with UC Davis despite clear contractual terms requiring updates every 30 minutes.


What happened was that Scriba stopped communicating for the most part. Sanders described that week as a “perfect storm” of problems, as they missed interconnections in bringing the system back online. Further, a key sys admin had resigned and his last day was Wednesday of that week (May 25). Sanders said that he thought his help desk was talking to UC Davis much more often than they were. It got so bad that Scriba’s voice mail system filled up, and the company was mostly unavailable by phone or email.

5) Scriba still has not figured out how to talk to their customers. As of Thursday morning when I interviewed Sanders (June 23), he was unaware that UC Davis had made their new locally-hosted system the new system of record. In a surreal moment, I had to tell Sanders that his customer was no longer using their site and had gone on their own. In an email to faculty and students on June 21, the UC Davis CIO and Vice Provost jointly described the changes:

We have developed a fully functional campus-hosted version of SmartSite that is available to all faculty, students, and staff at This new SmartSite contains all of the content it did before the outage, but does not rely on an external vendor.

Sanders thought this a bad plan, and he said he would have to call UC Davis the next week. I do not believe that he has talked to UC Davis since the May 20 – 27 outage, at least as of my interview with him last week.

Gross Negligence

I’m sorry for being so direct, but this outage that significantly affected a campus of 32,000 students for the last three weeks of the term comes down to a case of gross negligence. Scriba had no business signing the amended contract terms last year, as they clearly had inadquate staff and processes to actually meet the service levels promised. Scriba did not even attempt to meet their contractual obligations in complying with freezes, 6-week notification, 24-hour recovery, and communications during events.

I have tried to figure out whether there is fault with UC Davis. While the school’s insistence on a highly-customized environment was a questionable one, UC Davis seems to have acknowledged this dead-end path and has worked hard to pick a new LMS (they announced Spring 2016 that they’re moving to Canvas by next year). The customization was a poor decision five or more years ago, but UC Davis had a contract with a vendor who promised to support that system, and they paid extra for premium support and 24-hour disaster recovery.

You could argue that they signed a lousy contract in 2009, but they fixed this problem last year. The 2015 contract amendment is a model in clearly-defined service levels and communication plans. Unfortunately, Scriba simply did not meet their obligations and apparently does not have the capability to do so.

Scriba No Longer A Sakai Commercial Affiliate

In a postscript to this story, I went on the Apereo website and noticed that Scriba is no longer listed as a Commercial Affiliate. According to Ian Dolphin:

Scriba have been dropped as a commercial affiliate for non-payment of dues.

The Apereo board sent a notice of intent on June 15th, and on June 21st the board decided to end Scriba’s participation as commercial affiliate.

Sanders claimed that he knew that Apereo was changing it’s dues structure based on a January email, but he never found the actual invoice. He confirmed the non-payment of dues, but he also stated that Scriba is re-applying for membership. This claim is bizarre. If your primary business is acting as a Sakai commercial affiliate, it is hard to believe that you would simply avoid paying dues even if you missed the invoice. Furthermore, they also did not respond to the notice of action on June 15th. There is a pattern here.

Invitation to Publish

I asked Sanders if Scriba would be going out of business, as several of their main customers are leaving and their Apereo affiliation is gone. Sanders claimed that “we’re going to be fine” and had even re-signed a few customers over the past few weeks. I cannot tell if this is a case of ed tech’s own Baghdad Bob or if there truly are Scriba customers happy to work with them. I personally would be quite surprised to see Scriba in business in 2017.

I struggled with this post partially due to Sanders’ willingness to be open with me on the record. The issue is not him, however; the issue is the company and its actions, however. How do the owners of the company, Vert Capital, allow contracts to be signed that the company cannot fulfill? How does the ownership not make sure that they have the staff needed to meet their obligations? I sent a request to Vert for comment or interview with no response.

Nevertheless, I realize this is a harsh post. I have offered to Michael Sanders that he can send me a response that I will post here on e-Literate in case the company feels I have got any information wrong or if there are explanations for the behavior of the company.

The post LMS Outage: Exclusive view of UC Davis contract with Scriba appeared first on e-Literate.

by Phil Hill at June 27, 2016 02:52 PM

June 25, 2016

Michael Feldstein

Personalized Learning at Law Schools

I recently had the honor of speaking at the CALI (Center for Computer-Assisted Legal Instruction) conference. I was invited by one of my early heroes, John Mayer. When I first arrived on the ed tech blogging scene, John was already here, doing stuff. He inspired me.

Anyway, you may or may not know that law schools are currently experiencing an enrollment crisis. As a result, they are accepting students who are below their normal standards. These students are, unsurprisingly, not doing as well (on average) as their predecessors. So I ask the question: Is the problem that the students are “worse,” or is it that nobody is actually teaching law school students, and that the ones being admitted could succeed if only somebody taught them?

Here’s the video of the keynote, if you’re interested:

The post Personalized Learning at Law Schools appeared first on e-Literate.

by Michael Feldstein at June 25, 2016 06:24 PM

June 23, 2016

Michael Feldstein

We’re Giving a Course on Personalized Learning Next Month

Working with our good friends at ELI, we’re going to be offering a three-session synchronous course called Personalized Learning: Finding the Model That Fits Your Institution July 6th through 20th. As you know, we’re still in early days for personalized learning. Most campus communities are still trying to figure out what it is and what it’s good for—if they’re aware of it at all as a group. Plus, one size most emphatically does not fit all. The goal of the course is to help participants sharpen their own ideas for appropriate ways to facilitate exploration of the topic at their own institutions.

The format will be highly participatory, especially in the second and third sessions. Phil and I expect that we will all learn a lot from each other.

Join us!

The post We’re Giving a Course on Personalized Learning Next Month appeared first on e-Literate.

by Michael Feldstein at June 23, 2016 06:53 PM

June 21, 2016

Dr. Chuck

Beyond MOOCs: Open Education at Scale (Abstract)

Here is a draft abstract for an upcoming keynote – comments welcome.

Massively Open Online Course (MOOC) providers like edX and Coursera have revealed an almost unlimited desire for education for people of all ages and all walks of life. While these pioneering efforts have achieved much, these learning opportunities are still in relatively short supply. Each course is costly to produce, deploy, and support. These costs are a rate limiting factor in scaling online education to the point where we begin meeting the much larger demand for high quality, plentiful and relevant education worldwide. We need to build a Next Generation Digital Learning Environment (NGDLE) that makes it so any teacher can build and efficiently deploy their own open courses to a worldwide audience. In this presentation, we will look at how we can build an open source infrastructure that is based on open standards and open content that will make creating an open education experience within the reach of any teacher, anywhere in the world. We will look at how educational technology will need to change to reduce the cost to produce, share, and even remix online educational content.

by Charles Severance at June 21, 2016 11:36 PM

June 17, 2016

Apereo Foundation

Alex Balleste


More than two years ago Juanjo Meroño and I started the project LooWID ( As many of you already know it's an open source videoconference platform based on WebRTC that allow users to join in small rooms and share webcam, screen and audio and share files directly from browser to browser. Eduardo Rey has been also involved designing and implementing the interface and that helped a lot to have a nice platform.

It has been an good experience since we moved to make it open source instead of offering just as a service. Before we opened the project we tried to get feedback from friends and family and there were some interesting results. We asked them to fill a survey to guess what business model would be the best for LooWID. We were concerned that we should offer it as free but we needed to find a way to fund the infrastructure, and perhaps if the project grows contract an small team to work on it.  We were asking about the way the project must be funded. There were some options like, mandatory pay for use, advertisements, or donations. Around  90% of answers were 'donations' but in the following questions asking if they would donate to use a service like that, in small donations regularly or a bigger one (10€ - 15€) once, they all absolutely answer 'NO'.

So our conclusion was that people won't pay for a service that is offered by others for free, so we decided to change the direction and go to a fully opened to try to involve more people in the project in order to keep it alive. We knew that our motivation in the project would end if there was no response, so opening we wanted to get more people involved looking for more feedback to get energy to continue.
In some way we got it because we got a lot of nice words in the project, a lot of translations of platforms like German, Russian, Hungarian, other helped promoting the app with their tweets, and also Jose Rabal helped us with security, so the project got some good stuff, but the same main developers  that were at the beginning.

Now 1.4.1 release is out and included some interesting things like oembed integration in chat, but we realized that we can not afford big challenges like the one we were working on: the WebRTC relay. We were looking a way to grow up the numbers of users that were consuming streams in a LooWID room creating a relay in each  client in order to build multiple tree structures that would guarantee thousands of users getting video and audio running a simple 1GB Ram server. That sounds great, isn't it? But reality is that the time we can invest after work is too small and that makes it impossible to assume bigger goals like that :-(

So we've frozen the project for a while, trying to rest and get energy again. I hope that soon we'll start working again in some new features.

by Alex Ballesté ( at June 17, 2016 01:36 PM

June 07, 2016

Dr. Chuck

Live, Online, Graduation Ceremony for Python for Everybody Specialization Capstone – Wed June 8 at 9AM

We are reaching the end of the first session of the Python Specialization capstone. We have two things planned to celebrate the time and commitment of the students who have made it through five classes, 25 weeks of work and completed the capstone.

First, the University of Michigan School of Information will be sending signed paper certificates, a temporary tattoo, and a waiver of the application fee to the University of Michigan School of Information Masters Program to everyone who completed the Capstone. We will be sending the packet to every student that has completed the first capstone regardless of geography.

Second, we will be the first MOOC to have an online, live, graduation ceremony and we want to invite anyone to watch our celebration. The graduation ceremony will be held:

Wednesday June 8, at 9:00AM Eastern time


The URL will be ready Tuesday night. The ceremony will be streamed live on YouTube using Google Hangouts and then later a recording will be uploaded to Coursera for those without access to YouTube.

The agenda for the graduation ceremony will be to (1) thank those who have worked behind the scenes to make the course successful, (2) hear a short commencement speech from my colleague Colleen van Lent (who also teaches the Web Design for Everybody) specialization, and then (3) read the student’s names as we show each student’s picture.

There are 1165 students who have completed the capstone – but it is optional to participate in the ceremony.

If you have completed the capstone session 1 and want to participate in the graduation ceremony, go back to the course site and read the instructions for joining the ceremony that we sent you in email.

Thanks to everyone who made this possible and congratulations to the graduates on finishing the specialization.

I hope to see you at graduation.

by Charles Severance at June 07, 2016 05:23 PM

June 02, 2016

Adam Marshall

Announcing the ConnextOx Innovation Project

oae-logoFrom Vinton Cheng, Clinical Research Fellow, CRUK/MRC Oxford Institute of Radiation Oncology: “The ConnectOx project aims to bring researchers together and promote collaboration through researcher profiling and smart algorithms. By providing another route for researchers working within disparate subjects to find each other, our vision is that opportunities for novel and interdisciplinary connections will become enhanced, which in turn will break down the barriers holding back the free movement of knowledge. We will be working with key players in this space, *Research and Apereo OAE, to bring this project to fruition.”


by Adam Marshall at June 02, 2016 02:25 PM

May 20, 2016

Adam Marshall

WebLearn unavailable on Tuesday 24 May 2016 from 7-9am

cisco-routerWebLearn will be unavailable on Tuesday 24 May 2016 7-9am. This is necessary because of the need to undertake essential maintenance of the underlying AFS file system. There will be no service during this period.

We apologise for any inconvenience that this essential work may cause.

by Adam Marshall at May 20, 2016 01:41 PM

May 09, 2016

Sakai Project

Sakai 11 timeline

The Sakai core team is planning a Beta release the week of May 9, 2016. A Beta release is feature complete and within striking distance of an RC01, our first release candidate. We are anticipating having an RC01 shortly after Open Apereo 2016 and an expected release date in late June. 

by NealC at May 09, 2016 10:35 PM

JA Sakai Conference 2016 - A few notes from Sakai coordinator

Japanese Sakai schools meet three times a year, at JA Sakai,  at Open Apereo, and a JA Sakai unconference.

by NealC at May 09, 2016 10:32 PM

Sakai 10.7 released!

The Sakai Core Team is happy to announce the Sakai 10.7 maintenance release  for general availability! Congratulations worldwide team! There were 47 fixes  [1] in this release across the core tools. 2 new security fixes. 
if you're running 10.6 that this upgrade is of critical importance. Several of the fixes in 10.7 were regressions from 10.6, especially against Assignment and Samigo.

by NealC at May 09, 2016 10:31 PM

April 26, 2016

Adam Marshall

WebLearn and Turnitin courses Trinity Term 2016

IT Services offers a variety of taught courses to support the use of WebLearn and the plagiarism awareness software Turnitin. Course books for the formal courses (3-hour sessions) can be downloaded for self study. Places are limited and bookings are required.

Click on the links provided to book a place, or for further information. Bookings open 30 days in advance, but you can express an interest in a course and receive a reminder to book when booking opens.

WebLearn courses:

Plagiarism awareness courses (Turnitin):

Byte-sized lunch time sessions:

These focus on particular tools with plenty of time for questions and discussion

User Group meeting:

by Jill Fresen at April 26, 2016 03:37 PM

Dr. Chuck

More Tsugi Refactoring – Removal of the mod folder

I completed the last of many refactoring steps of Tsugi yesterday. when I moved the contents of the “mod” folder into its own repository. The goal of all this refactoring was to get it to the point where checking out the core Tsugi repository did not include any end-user tools – it just would include the administrator, developer, key management, and support capabilities (LTI 2, CASA, ContentItem Store). The key is that this console will also be used for the Java and NodeJS implementations of Tsugi until we build the functionality in the console in each of those languages and so it made no sense to drag in a bunch of PHP tools if you were just going to use the console. I wrote a bunch of new documentation showing how the new “pieces of Tsugi” fit together:

This means that as of this morning if you do a “git pull” in your /tsugi folder – the mod folder will disappear. But have no fear – you can restore it with the following steps:

cd tsugi
git clone mod

And your mod folder will be restored. You will now have to do separate git pulls for both Tsugi and the mod folder.

I have all this in solid production (with the mod restored as above) with my Coursera and on campus Umich courses. So I am pretty sure it holds together well.

This was the last of a multi-step refactor for this code to modularize it in multiple repositories so as to better prepare for Tsugi in multiple languages as well as plugging Tsugi into various production environments.

by Charles Severance at April 26, 2016 02:11 PM

April 03, 2016

Steve Swinsburg

Tool properties in tool registration files

I discovered this feature by accident when setting up a new tool and configuring its registration file. The registration file is what you use to wire up a webapp in Sakai so that it can be added to sites. You can give it a title, description, tell it what site types are supported and a few other settings.

One of the recent features in Sakai is the ability to get a direct URL to any tool within Sakai. This is useful when you want to link to a tool without the portal around it.


Note the links on the right hand side are part of the tool registration. The ones on the left are controlled within the tool code itself and together it makes for a nice navbar when in full screen mode.

However, if you have a tool that doesn’t need any header items, for example a summary tool or widget, and there are multiples of them on screen, you still get the Link and Help items which can clutter the UI. You can disable the Help in the tool registration file via:

<configuration name="help.button" value="false" />

However the Link doesn’t have a corresponding configuration option (oversight maybe… blame me, I wrote the code…). However you can disable it with a tool property – although this is normally something reserved for an admin user to set into the tool placement within the portal, which is a manual step per placement. But what I have discovered is that you can add the tool property to the tool registration file and it is automatically linked up! Magic.

<configuration name="sakai:tool-directurl-enabled" value="true" />

This is coming in very handy as we are creating a series of relatively small widgets to place on the home screen of a site and the header toolbar was cluttering the UI. Now it is nice and clean with the header toolbar completely removed.


by steveswinsburg at April 03, 2016 10:14 PM

February 12, 2016


Known Issue: Incomplete list of students enrolled by section in Roster and Gradebook 2

For course sites where multiple course sections have access, instructors or teaching assistants use the drop-down menu on top of their list of students in Gradebook 2 to filter the students by class section. Be aware that using this feature might only return a partial list of students. Gradebook 2 uses the Roster tool to… Continue reading

by Mathieu Plourde at February 12, 2016 06:29 PM

February 03, 2016

Ian Boston

Ai in FM

Limited experience in either of these fields does not stop thought or research. At the risk of being corrected, from which I will learn, I’ll share those thoughts.

Early AI in FM was broadly expert systems. Used to advise on hedging to minimise overnight risk etc or to identify certain trends based on historical information. Like early symbolic maths programs (1980s) that revolutionised the way in which theoretical problems can be solved (transformed) without error in a fraction of the time, early AI in FM put an expert with a probability of correctness on every desk. This is not the AI I am interested in. It it only artificial in the sense it artificially encapsulates the knowledge of an expert. The intelligence is not artificially generated or acquired.

Machine learning  covers many techniques. Supervised learning takes a set of inputs and allows the system to perform actions based on a set of policies to produce an output. Reinforcement learning favors the more successful policies by reinforcing the action. Good machine, bad machine. The assumption is, that the environment is stochastic. or unpredictable due to the influence of randomness.

Inputs and outputs are simple. They are a replay of the historical prices. There is no guarantee that future prices will behave in the same way as historical, but that is in the nature of a stochastic system.  Reward is simple. Profit or loss. What is not simple is the machine learning policies. AFAICT, machine learning, for a stochastic system with a large amount of randomness, can’t magic the policies out of thin air. Speech has rules, Image processing also and although there is randomness, policies can be defined. At the purests level, excluding wrappers, financial markets are driven by the millions of human brains attempting to make a profit out of buying and selling the same thing without adding any value to that same thing. They are driven by emotion, fear and every aspect of human nature rationalised by economics, risk, a desire to exploit every new opportunity, and a desire to be a part of the crowd. Dominating means trading on infinitesimal margins exploiting perfect arbitrage as it the randomness exposes differences. That doesn’t mean the smaller trader can’t make money, as the smaller trader does not need to dominate, but it does mean the larger the trader becomes, the more extreme the trades have to become maintain the level of expected profits. I said excluding wrappers because they do add value, they adjust the risk for which the buyer pays a premium over the core assets. That premium allows the inventor of the wrapper to make a service profit in the belief that they can mitigate the risk. It is, when carefully chosen, a fair trade.

The key to machine learning is to find a successful set of policies. A model for success, or a model for the game. The game of Go has a simple model, the rules of the game. Therefore it’s possible to have a policy of, do everything. Go is a very large but ultimately bounded Markov Decision Process (MDP).  Try every move. With trying every move every theoretical policy can be tested. With feedback, and iteration, input patterns can be recognised and successful outcomes can be found. Although the number of combinations is large, the problem is very large but finite. So large that classical methods are not feasible, but not infinite so that reinforcement machine learning becomes viable.

The MDP governing financial markets may be near infinite in size. While attempts to formalise will appear to be successful the events of 2007 have shown us that if we believe we have found finite boundaries of a MDP representing trade, +1 means we have not. Just as finite+1 is no longer finite by the original definition, infinite+1 proves what we thought was infinite is not. The nasty surprise just over the horizon.

by Ian at February 03, 2016 01:09 PM