Archive for September, 2015

Weekly QuEST Discussion Topics and News, 25 Sept

September 24, 2015 Leave a comment

QuEST 25 Sept 2015:


We will have our colleague ‘Tom’ M. give us a talk related to our recent focus on deep learning and a specific application we are very interested in ATR.


Title: ATR for Time Critical Targeting



For years, real-time automatic target recognition (ATR) for time critical targeting has been a challenging research problem.  Contested environments cause additional complexity for collecting target signatures.  Hence, next generation ATR research must address: (1) How ATR can be executed in real-time on a size, weight, and power (SWaP) constraint system and (2) How ATR performance can be improved when available data are sparse.  Deep Learning has been very effective for target recognition from optical imagery.  IBM has developed a special hardware called “True North” to execute DL algorithm on a SWaP constrained system.  True North uses a fraction of power compare to GPU, CPU, or FPGA.  In this talk, we will present how DL algorithm combined with True North can address the critical challenges of ATR.

news summary (26)

Categories: Uncategorized

Weekly QuEST Discussion Topics and News, 18 Sept

September 17, 2015 Leave a comment

QuEST for 18 Sept 2015:

We will have a discussion on thinking – specifically starting with a presentation by Geof Hinton on:

Aetherial Symbols

Geoffrey Hinton

University of Toronto


Google Inc.

Kosslyn thought that we process images by operating on

pixels in our head. Pylyshyn thought that we process

sentences by operating on symbols in our head.

  • They were both making the same naive mistake.

– What we have in our heads is not a cleaned up

version of the input. ** this begs the question why we are seeking to populate an internal representation for our computer agents with cleaned up versions of the  sensory data ** There are no pixels or symbol

strings in the head.

  • All we have in our heads is big activity vectors that cause

more big activity vectors.

– So why did people think we have symbol strings or

images in the head?

– Maybe they misunderstood how everyday language

refers to internal states.

How does intuitive reasoning work?

  • How can a neural net infer the answer to a

simple question without doing any explicit,

sequential reasoning?

  • Represent each word by a large vector of

features each of which has causal effects.

– The combined effects of all the features

capture knowledge that symbolic AI would put

into an explicit rule of inference.

Natural reasoning

  • Once we can turn sentences into thought

vectors, we can learn to predict a thought vector

from previous thought vectors.

– This should allow us to model natural

reasoning, but it hasn’t yet been done.

—- this last point is where we want to focus –

The variables used to create/use these thought vectors:

  • Nearly all artificial neural nets make do with just two

kinds of variables:

– Neural activities are used to represent what the net is

currently thinking about.

– Weights are used to store the long-term knowledge.

Will we require a new type of variable to do QuEST thought vectors?

Let’s architect (in our discussions during the QuEST meeting) a dual process theory compliant model using a deep learning CNN / RNN infrastructure.  In the DPT solution the units of cognition in the 2nd system, vocabulary manipulated and used to populate working memory are what we call qualia.  Qualia are the units of cognition used in working memory.  Any agent that implements a DPT solution and generates a DPT sys2 where the units of cognition (vocabulary of working memory) are compliant with our QuEST tenets of Structural coherence, Situated and simulated are called QuEST agents.

As a means to prime the discussion imagine a CNN/RNN system for generating narrative descriptions of images/video snippets.  As currently implemented in the AC3 system.  The thought vectors (weight space representation of the sentences) are generated in a reflexive manner.  If you will using the terminology of Evans / Stanovich the autonomous mind.  Those thought vectors can be used to generate an output (reflexive response) for the agent but can also be used to create a working memory representation.  This is where our discussion will focus.

news summary (25)

Categories: Uncategorized

Weekly QuEST Discussion Topics, 11 Sept

September 10, 2015 Leave a comment

QuEST 11 Sept 2015:

Capt Amerika is on a panel at next week’s Air Force Association meeting in DC with Lt Gen Otto AF/A2, also a person from Microsoft and a person from the commercial space industry – panel information is below – this week’s QuEST meeting will be to have an open discussion by any interested in laying out important aspects of the topics we think should be made – I’ve posted on the VDL and they are also out there on the web several documents as background:  AF ISR 2023 Delivering Decision Advantage, Revolutionizing Intelligence Analysis (ISR whitepaper from the A2 ISR Horizons series)

Panel Title: Empowering Tomorrow’s Analyst: Revolutions in Analytics

Panel Description: Transformative technological concepts such as automation, big data utilization, innovative Airman-machine teaming, and the integration of commercial analytics will drastically alter the environment in which our future Air Force analysts conduct cross-domain operations. This panel will focus on how the military and industry are embracing the onset of this information revolution and its impact upon the Air Force ISR Enterprise. We will also explore  the effects of  open source information integration, partner nation collaboration, and the fusion of industry and academia in enabling military decision advantage.

Panel Information:

  • Panel Date/Time: 14 Sep 15/1510-1555
  • Panel Length: 45 mins

o   Moderator Intro: 5 mins

o   Panel Member Intro: 20 mins (~5 mins per member)

o   Q&A: 20 mins

  • Panel Members:

o   Brig Gen (Ret) Scott Van Cleef, AFA Chairman of the Board – Moderator

o   Lt Gen Robert Otto, Deputy Chief of Staff ISR (AF/A2)

o   Dr. Steven Rogers, AFRL, Senior Scientist for Automatic Target Recognition and Sensor Fusion

o   James Crawford, CEO, Orbital Insight

o   Sam Druker, Microsoft Director of Data Science

  • Audience: Approx. 100 – 150 Government, Military, Academia, Industry, and Press

Given the topics:  automation, big data utilization, innovative Airman-machine teaming, and the integration of commercial analytics – from a QuEST perspective what are the key ideas we want to ensure are in the discussion?

Automation:  ideas versus automatic/autonomy?  Where we’ve had some success – FMV cells?  Task space parsing to find the sweet spot?  Automation / Autonomy and the unexpected query?

Big Data utilization:  what is big data in ISR context?

Innovative Airman-machine Teaming:  Patterson/Petersen effort?  PCPADx CP1 effort with FMV cells

Commercial Analytics:  what is unique about our needs?

Cross domain integration:  25th AF mission:  Provide full-spectrum decision advantage to warfighters and national leaders through globally integrated ISR, electronic attack, information operations, and strategic C2 – convergence of technical vision for cyber/EW, strike, ISR

How military embracing information revolution – impact on ISR enterprise:  ISR Horizons documents

Effects of Open Source Information:  security issues – maybe the future is not in access to the information – old ISR assumed that to be – now the advantage is in decision making –

Partner Nation collaboration in ISR enterprise:  issues that Mike K lists –

Fusion of industry and academia enabling military decision advantage: Facebook and LeCun, Google and Hinton, … clearly commercial work is establishing unique relationships with academia to solve their big issues – a model we need to consider, beyond the afosr approach of funding basic research

Categories: Uncategorized

No QuEST Meeting, 4 Sept

September 3, 2015 Leave a comment

Due to the holiday weekend, we will not be having a QuEST Meeting this week, Friday the 4th.

Have a safe and happy long weekend!

Categories: Uncategorized