Archive for April, 2015

Weekly QuEST Discussion Topics and News, 24 Apr

April 24, 2015 Leave a comment

QuEST for 24 April 2015

Our colleague Dean W will lead a discussion and is seeking feedback on his research:

What am I trying to do? – Increase the resilience of cyber-physical (specifically Industrial Control/SCADA) systems by applying formal verification techniques in a system of systems approach.

How will I do this? – Using model checking tools and specifically modeling malicious interactions of an external agent with the system under test as  a means of discovering any emergent vulnerabilities in the system of systems that normal function checking would not be looking for.

The Capt Amerika would like to flip through some charts from a recent review of technology trends for discussions from WEBBMEDIA group 2015 trend report – although the topics themselves are of general interest one of the discussion points we would like to emphasize is relationship of these trends to QuEST. For example:


First year on the list


At its essence, an algorithm is simply a set of rules or processes that must be followed in order to solve a problem. For thousands of years (Euklid’s algorithm is 2,500 years old!) algo­rithms have been used to increase speed and efficiencies, and they’ve been applied to assist with our everyday tasks. In the coming year, we’ll see the launch of services using algorithms to create stunning designs, to curate the news and even to target voters for individual mes­saging in close political districts. We’ll see the rise of public algorithm exchanges. We will also begin questioning the ethics of how algorithms can be used, and we’ll scrutinize the tendency of some algorithms to go awry.

Project Dreamcatcher from Autodesk

Algorithmic Design

Project Dreamcatcher from Autodesk is the next wave of computational design systems. While it doesn’t replace a designer herself, it does give her the ability to feed a project’s de­sign requirements, constraints and exemplars into Dreamcatcher, whose algorithm will then return possible design concepts. If you’ve ever been in a meeting when a few people offer up an app they’d like to emulate, while others prefer a different user interface, algorithmic design systems can take the best of both, combine them into one and then help you refine the favored design.

Algorithm Marketplaces

Long ago, developers realized that everyone wins when knowledge is freely exchanged. As a result, communities of developers are offering up their algorithms in emerging algorithm marketplaces. Algorithmia is building a sort of Amazon for algorithms, where developers can upload their work to the cloud and receive payment when others pay to access it.DataXu offers a marketplace for its proprietary algorithms. Meantime Github, the code sharing network started by Linux creator Linus Torvalds, will continue to grow.

Algorithmic Curation

Algorithmic curation is a process that automatically determines what content should be displayed or hidden and how it should be present­ed to your audience. Facebook’s NewsFeed already uses an algo­rithm to curate all the posts created in your network to serve only the content it thinks will engage you most. It has deployed a new service, FB Techwire, across its network to surface embeddable news sto­ries for media organizations. Google and Yahoo news will continue to refine their algorithms, which use our online behaviors to deter­mine which content to show. In 2016 and beyond, we expect to see algorithms curating news content not just based on our interests, but also for our most recent behavior. Rather than delivering a full breaking news story to our mobile phones, algorithms will deliver the “waiting in line at Starbucks” version of that story, a more in-depth longread to our tablets, and a video version of that story once we’re in front of our connected TVs. As a result, news organizations and other content producers have thrilling opportunities in the year ahead to supercharge and personalize content in ways we have never seen before. (See also: Consumer > Device.)

Another example



Second year on the list

2015 Tech Trends | | © 2014 Webbmedia Group

Key Insight

SVPAs made our list last year because they were just beginning to enter the market as stand-alone mobile apps. (Others call this technology “predictive applications” or “predictive intelligence.”) They used semantic and natural language processing, mined data from our calendars, email and contact lists and used the last few minutes of our behavior to anticipate the next 10 seconds of our thinking in order to help consumers manage daily tasks, finances, diet and more. In 2015, we will see SVPA technology become a key part of emerging platforms and devices.

news summary (17)

Categories: Uncategorized

Weekly QuEST Discussion Topics and News, 17 Apr

April 16, 2015 Leave a comment

QuEST 17Apr 2015:

We have this week a guest speaker – from Penn – Prof Dan Guralnik – He has been part of a Multi-University Research Initiative associated with topics of interest to QuEST.

We propose a self-organizing memory architecture for perceptual experience provably capable of supporting autonomous learning and goal-directed problem solving in the absence of any prior information about the agent’s environment. The architecture is simple enough to ensure (1) a quadratic bound (in the number of available sensors) on space requirements, and (2) a quadratic bound on the time-complexity of the update-execute cycle. At the same time, it is sufficiently complex to provide the agent with an internal representation which is (3) minimal among all representations which account for every sensory equivalence class consistent with the agent’s belief state; (4) capable, in principle, of recovering a topological model of the problem space; and (5) learnable with arbitrary precision through a random application of the available actions. These provable properties — both the trainability and the efficacy of an effectively trained memory structure — exploit a duality between {\it weak poc sets}~ —~ a symbolic (discrete) representation of subset nesting relations~ —~ and {\it non-positively curved cubical complexes}, whose rich convexity theory underlies the planning cycle of the proposed architecture.

news summary (16)

Categories: Uncategorized

Weekly QuEST Discussion Topics and News, 10 Apr

This week our colleague ‘Sam’ will present some background material to
continue our discussion of transfer learning specifically related to deep
learning systems.

In his words, “In this talk, I will give an introduction to transfer learning, including common definitions, motivation from a machine learning perspective and descriptions of broad strategies for transfer learning. The talk will conclude with how transfer learning is achieved within deep convolutional neural networks.”news summary (15)

Weekly QuEST Discussion Topics and News, 3 Apr

QuEST 3 April 2015

  • I mentioned last week that I was extending my upcoming plenary talk at the Defense sensing symposia to include not only a discussion of autonomy and specifically autonomy for offensive cyber operations to also include a discussion of autonomous ISR with a focus on persistence and coalition issues.  I would like to present the flow of the presentation for comments / insertion of ideas from the QuEST group on these topics.
  • We also want to return to the topic we briefly mentioned last week as we closed – formalism for defining the unexpected query (UQ) – taken from the transfer learning literature – specifically a 2010 survey article by Pan – A Survey of Transfer Learning, IEEE transactions on knowledge and data engineering, vol 22, no 10, oct 2010.  We want to define the term ‘query’ and then ‘unexpected query’ using their formalism and also address the question from our colleague Andres R. on how does the UQ relate to generalization?  Lastly we need to establish a position on transfer learning and consciousness.  So if one of the purposes of consciousness is to respond to the UQ – AND – transfer learning is an area of research that attempts to respond to the UQ – what is it we think consciousness (QuEST) brings to transfer learning?  In the article section 2.3 provides a means to have this discussion
  • In transfer learning, we have the following three main research issues – so what does QuEST bring to these areas?:
  • 1) What to transfer – What:  asks which part of knowledge can be transferred across domains or tasks. Some knowledge is specific for individual domains or tasks, and some knowledge may be common between different domains such that they may help improve performance for the target domain or task.  In our terms we have experience responding to queries (recall how we defined a query – as an agent capturing a stimuli and responding).  We capture those experiences for later use (knowledge).  Some knowledge is specific for very specific types of queries but some knowledge may be useful for improving performance for other task/queries.  Our previous discussions on Gist should be re-introduced here.
  • 2) How to transfer – After discovering which knowledge can be transferred,learning algorithms need to be developed to transfer the knowledge, which corresponds to the “how to transfer” issue.  Under this topic we need to enforce our ideas of situating / simulating.
  • 3) When to transfer – When:  asks in which situations, transferring skills should be done. Likewise, we are interested in knowing in which situations, knowledge should not be transferred.  In some situations, when the source domain and target domain are not related to each other, brute-force transfer may be unsuccessful. In the worst case, it may even hurt the performance of learning in the target domain, a situation which is often referred to as negative transfer.  Under this topic we need to discuss our ideas of the competing narratives to generate a stable/consistent / useful confabulation.
  • So the question becomes does consciousness provide an advantage along any or all of these transfer learning axes?

news summary (13)

Categories: Uncategorized