Archive

Archive for June, 2016

Weekly QuEST Discussion Topics and News, 1 July

QuEST 1 July 2016:

We are extremely happy this week to have our colleague Sandy V – who SUCCESSFULLY defended her dissertation research last week to provide a discussion on her research.  As usual interactions are expected and appreciated.  Sandy has an accepted paper coming out in Biologically inspired computing architectures and also a presentation at the upcoming CogPsy meeting.

A Novel Machine Learning Classifier Based on a Qualia Modeling Agent (QMA)

This dissertation addresses a problem found in standard machine learning (ML) supervised classifiers, that the target variable, i.e., the variable a classifier predicts, has to be identified before training begins and cannot change during training and testing. This research develops a computational agent, which overcomes this problem.

The Qualia Modeling Agent (QMA) is modeled after two cognitive theories:

Stanovich’s tripartite framework, which proposes learning results from interactions between conscious and unconscious processes; and, the Integrated Information Theory (IIT) of Consciousness, which proposes that the fundamental structural elements of consciousness are qualia.

By modeling the informational relationships of qualia, the QMA allows for retaining and reasoning-over data sets in a non-ontological, non-hierarchical qualia space (QS). This novel computational approach supports concept drift, by allowing the target variable to change ad infinitum without re-training, resulting in a novel Transfer Learning (TL) methodology, while achieving classification accuracy comparable to or greater than benchmark classifiers. Additionally, the research produced a functioning model of Stanovich’s framework, and a computationally tractable working solution for a representation of qualia, which when exposed to new examples, is able to match the causal structure and generate new inferences.

news summary (17)

Advertisements
Categories: Uncategorized

Weekly QuEST Discussion Topics and news, 24 June

QuEST 24 June 2016

There will be no phone availability for this meeting as that is not appropriate for a dissertation defense – we will use this week’s meeting for the oral part of the defense of the QuEST related dissertation work of our colleague Sandy V.  We hope to have Sandy present her work / publications at subsequent QuEST meetings.

A Novel Machine Learning Classifier Based on a Qualia Modeling Agent (QMA)

This dissertation addresses a problem found in standard machine learning (ML) supervised classifiers, that the target variable, i.e., the variable a classifier predicts, has to be identified before training begins and cannot change during training and testing. This research develops a computational agent, which overcomes this problem.

The Qualia Modeling Agent (QMA) is modeled after two cognitive theories:

Stanovich’s tripartite framework, which proposes learning results from interactions between conscious and unconscious processes; and, the Integrated Information Theory (IIT) of Consciousness, which proposes that the fundamental structural elements of consciousness are qualia.

By modeling the informational relationships of qualia, the QMA allows for retaining and reasoning-over data sets in a non-ontological, non-hierarchical qualia space (QS). This novel computational approach supports concept drift, by allowing the target variable to change ad infinitum without re-training, resulting in a novel Transfer Learning (TL) methodology, while achieving classification accuracy comparable to or greater than benchmark classifiers. Additionally, the research produced a functioning model of Stanovich’s framework, and a computationally tractable working solution for a representation of qualia, which when exposed to new examples, is able to match the causal structure and generate new inferences.

news summary (16)

Categories: Uncategorized

Weekly QuEST Discussion Topics, 17 June

QuEST 17 June 2016

We want to discuss an article:

 

The Human Side of Automation

Chapter · January 2015

DOI: 10.1007/978-3-319-19078-5_7

Don Norman – UCSD

Abstract

Full automation is good. But the path toward full automation must traverse

the dangerous stages of partial automation, where the very successes

of automation may make human drivers less able to respond to the unexpected, unavoidable imperfections of the automation. The appropriate

design path to deal with this requires us to reconsider the driver and automation as a team, as collaborative partners.

The first topic this week will be a re-visit / re-statement of our position on autonomy, meaning and understanding.

Next we want to revisit our framework for discussing autonomy and our approach to FAQs – maybe some discussions of questions and answers that should be included:

  • An autonomous system is a participant in the manipulation of any/all three Principles that define autonomy. 

–     Organization / Population: The system can change its C2 relationships with other agents (participates in the negotiation that results in the accepted change – the autonomous system has to ‘understand’ the meaning of the new C2 relationship to respond acceptably), an example is the proxy relationship where a human grants to the autonomy its proxy to accomplish some task and the autonomy decides how to accomplish that task and then executes the task.  The recent fielding of the ground collision avoidance solution fits into this category – it has saved multiple people / aircraft since its fielding – it takes over when the aircraft is appearing to be heading for a controlled flight ground collision – it then turns control back to the human pilot who executes a recovery.

–     Situated Environment:  Its tasks (example changing what it measures to accomplish a task like changing the modes in a modern sensor or even changing the task based on changing conditions…), example is that autonomy is respect to accomplishing something (the task) in association with the granting of the proxy to go off and do that task but how the autonomy accomplishes the task can be changed when appropriate by the autonomy

–     Agent Characteristics:  Lastly changing its (the autonomous agent’s) cognitive approach to accomplishing its task (example – in a machine learning situation changing it decision boundaries or its rules or even the approach to machine learning for a given task).  Three are many efforts associated with transfer learning for example where a system attempts to use its prior knowledge to improve the adaptation to new challenges.

Now with respect to meaning:  The meaning of a stimulus is the agent specific representational changes evoked by that stimulus in that agent.  The update to the representation, evoked by the data, is the meaning of the stimulus to this agent.  Meaning is NOT just the posting into the representation of the data it is all the resulting changes to the representation.  For example, the evoking of tacit knowledge or a modification of the ongoing simulation (consciousness) is included in the meaning of a stimulus to an agent. [26, 30] Meaning is not static and changes over time.  The meaning of a stimulus is different for a given agent depending on when it is presented to the agent.

  • So if we take a word’s meaning to be the means of picking out its referent, then meanings are in our brains.  ** or in the representation of a given agent **
  • That is meaning in the narrow sense.
  • If we use “meaning” in a wider sense, then we may want to say that meanings include both the referents themselves and the means of picking them out.

Wider sense is consistent with what Bob E has been suggesting – meaning making is meaning – it is not just the result of the processes it is the processes themselves also – in addition this suggest the meaning includes the referent itself also – that is problematic with our view of ‘situations’ / qualia – so a point we want to discuss.

Meaning is the changes in an agent’s representation resulting from a query – butunderstanding is the impact of that meaning on accomplishing a particular ISR task 

n  The real key is situation understanding is associated with a task – where meaning is universal (the changes to the representation of an agent resulting from any query is the meaning of that query to that agent) and independent of task (obviously an agent could change its processing of a query dependent on its current task so a task can impact meaning – but whether it does or not the meaning is the changes) –

meaning associated with the accomplishment of the task will be considered understanding from the perspective of the evaluating agent if it raises the evaluating agent’s estimation of the probability of the performing agent responding acceptably to the query – a higher probability for being successful at a task from perspective of an evaluating agent.

 

Along this same line we are attempting to generate a frequently asked questions for those not in the autonomy / AI / human-machine teaming area.  We will briefly present some examples and seek a discussion on how to provide ‘cocktail’ party answers.  We will also take suggestions for additional questions that maybe should be included.

Another topic this week is our pursuit of approaches to making decisions in the tails of the distributions – we have a subfolder in QuEST that is accumulating articles on this topic.

One article we reviewed this week is by a talented ‘big-data’ machine learning person at JHU – Suchi Saria on prediction of Sepsis.

SEPSI S

A targeted real-time early warning score (TREWScore)

for septic shock

Katharine E. Henry,1 David N. Hager,2 Peter J. Pronovost,3,4,5 Suchi Saria1,3,5,6

http://www.ScienceTranslationalMedicine.org 5 August 2015 Vol 7 Issue 299 299ra122

Sepsis is a leading cause of death in the United States, with mortality highest among patients who develop septic shock. Early aggressive treatment decreasesmorbidity andmortality. Although automated screening tools can detect patients currently experiencing severe sepsis and septic shock, none predict those at greatest risk of developing shock. We analyzed routinely available physiological and laboratory data from intensive care unit patients and developed “TREWScore,” a targeted real-time early warning score that predicts which patients will develop septic shock.

TREWScore identified patients before the onset of septic shock with an area under the ROC (receiver operating characteristic) curve (AUC) of 0.83 [95% confidence interval (CI), 0.81 to 0.85]. At a specificity of 0.67, TREWScore achieved a sensitivity of 0.85 and identified patients a median of 28.2 [interquartile range (IQR), 10.6 to 94.2] hours before onset. Of those identified, two-thirds were identified before any sepsis-related organ dysfunction. In comparison, the Modified Early Warning Score, which has been used clinically for septic shock prediction, achieved a lower AUC of 0.73 (95%CI, 0.71 to 0.76). A routine screening protocol based on the presence of two of the systemic inflammatory

response syndrome criteria, suspicion of infection, and either hypotension or hyperlactatemia achieved a lower sensitivity of 0.74 at a comparable specificity of 0.64. Continuous sampling of data from the electronic health records and calculation of TREWScore may allow clinicians to identify patients at risk for septic shock and provide earlier interventions that would prevent or mitigate the associated morbidity and mortality.

 

A second article of interest:

 

A Framework for Individualizing Predictions of Disease

Trajectories by Exploiting Multi-Resolution Structure

 

Peter Schulam

Dept. of Computer Science

Johns Hopkins University

Baltimore, MD 21218

pschulam@jhu.edu

Suchi Saria

Dept. of Computer Science

Johns Hopkins University

Baltimore, MD 21218

ssaria@cs.jhu.edu

For many complex diseases, there is a wide variety of ways in which an individual can manifest the disease. The challenge of personalized medicine is to develop tools that can accurately predict the trajectory of an individual’s disease, which can in turn enable clinicians to optimize treatments. We represent an individual’s disease trajectory as a continuous-valued continuous-time function describing the severity of the disease over time. We propose a hierarchical latent variable model that individualizes predictions of disease trajectories. This model shares statistical strength across observations at different resolutions–the population, subpopulation and the individual level. We describe an algorithm for learning population and subpopulation parameters offline, and an online procedure for dynamically learning individual-specific parameters. Finally, we validate our model on the task of predicting the course of interstitial lung disease, a leading cause of death among patients with the autoimmune disease scleroderma. We compare our approach against state-of-the-art and demonstrate significant improvements in predictive accuracy.

 

Lastly we have another recent article that is along the lines of generating a confabulated representation to complement sensory data.

 

Generative Adversarial Text to Image Synthesis

Scott Reed, Zeynep Akata, Xinchen Yan, Lajanugen Logeswaran REEDSCOT1, AKATA2, XCYAN1, LLAJAN1

Honglak Lee, Bernt Schiele HONGLAK1, SCHIELE2

1 University of Michigan, Ann Arbor, MI, USA (UMICH.EDU)

2 Max Planck Institute for Informatics, Saarbr¨ucken, Germany (MPI-INF.MPG.DE)

 

arXiv:1605.05396v1 [cs.NE] 17 May 2016

 

Automatic synthesis of realistic images from text would be interesting and useful, but current AI systems are still far from this goal. However, in recent years generic and powerful recurrent neural network architectures have been developed to learn discriminative text feature representations.

Meanwhile, deep convolutional generative adversarial networks (GANs) have begun to generate highly compelling images of specific categories such as faces, album covers, room interiors etc. In this work, we develop a novel deep architecture and GAN formulation to effectively bridge these advances in text and image modeling, translating visual concepts from characters to pixels. We demonstrate the capability of our model to generate plausible images of birds and flowers from detailed text descriptions

Categories: Uncategorized

Weekly QuEST Discussion Topics and News, 10 June

QuEST 10 June 2016

Cap is participating in a DOD summit on human machine collaboration, autonomy and AI and also is presenting in an upcoming JASONs study on artificial general intelligence (AGI) – both of these interactions will be opportunities to openly discuss QuEST positions on what is the missing link in current approaches to AGI and autonomy.

The first topic this week will be a re-visit / re-statement of our position on autonomy, meaning and understanding.

First on autonomy we would like to discuss the following – from an agent perspective:

  • Autonomy is defined by three elements: 

–     Organization / Population: The system can change its C2 relationships with other agents (participates in the negotiation that results in the accepted change – the autonomous system has to ‘understand’ the meaning of the new C2 relationship to respond acceptably), an example is the proxy relationship where a human grants to the autonomy its proxy to accomplish some task and the autonomy decides how to accomplish that task and then executes the task.

–     Situated Environment:  Its tasks (example changing what it measures to accomplish a task like changing the modes in a modern sensor or even changing the task based on changing conditions…), example is that autonomy is respect to accomplishing something (the task) in association with the granting of the proxy to go off and do that task but how the autonomy accomplishes the task can be changed when appropriate by the autonomy

–     Agent Characteristics:  Lastly changing its (the autonomous agent’s) cognitive approach to accomplishing its task (example – in a machine learning situation changing it decision boundaries or its rules or even the approach to machine learning for a given task).

Now with respect to meaning:  The meaning of a stimulus is the agent specific representational changes evoked by that stimulus in that agent.  The update to the representation, evoked by the data, is the meaning of the stimulus to this agent.  Meaning is NOT just the posting into the representation of the data it is all the resulting changes to the representation.  For example, the evoking of tacit knowledge or a modification of the ongoing simulation (consciousness) is included in the meaning of a stimulus to an agent. [26, 30] Meaning is not static and changes over time.  The meaning of a stimulus is different for a given agent depending on when it is presented to the agent.

  • So if we take a word’s meaning to be the means of picking out its referent, then meanings are in our brains.  ** or in the representation of a given agent **
  • That is meaning in the narrow sense.
  • If we use “meaning” in a wider sense, then we may want to say that meanings include both the referents themselves and the means of picking them out.

Wider sense is consistent with what Bob E has been suggesting – meaning making is meaning – it is not just the result of the processes it is the processes themselves also – in addition this suggest the meaning includes the referent itself also – that is problematic with our view of ‘situations’ / qualia – so a point we want to discuss.

Meaning is the changes in an agent’s representation resulting from a query – butunderstanding is the impact of that meaning on accomplishing a particular ISR task 

n  The real key is situation understanding is associated with a task – where meaning is universal (the changes to the representation of an agent resulting from any query is the meaning of that query to that agent) and independent of task (obviously an agent could change its processing of a query dependent on its current task so a task can impact meaning – but whether it does or not the meaning is the changes) –

meaning associated with the accomplishment of the task will be considered understanding from the perspective of the evaluating agent if it raises the evaluating agent’s estimation of the probability of the performing agent responding acceptably to the query – a higher probability for being successful at a task from perspective of an evaluating agent.

 

A second topic this week is our pursuit of approaches to making decisions in the tails of the distributions – we have a subfolder in QuEST that is accumulating articles on this topic.

One article we reviewed this week is by a talented ‘big-data’ machine learning person at JHU – Suchi Saria on prediction of Sepsis.

SEPSI S

A targeted real-time early warning score (TREWScore)

for septic shock

Katharine E. Henry,1 David N. Hager,2 Peter J. Pronovost,3,4,5 Suchi Saria1,3,5,6

http://www.ScienceTranslationalMedicine.org 5 August 2015 Vol 7 Issue 299 299ra122

Sepsis is a leading cause of death in the United States, with mortality highest among patients who develop septic shock. Early aggressive treatment decreasesmorbidity andmortality. Although automated screening tools can detect patients currently experiencing severe sepsis and septic shock, none predict those at greatest risk of developing shock. We analyzed routinely available physiological and laboratory data from intensive care unit patients and developed “TREWScore,” a targeted real-time early warning score that predicts which patients will develop septic shock.

TREWScore identified patients before the onset of septic shock with an area under the ROC (receiver operating characteristic) curve (AUC) of 0.83 [95% confidence interval (CI), 0.81 to 0.85]. At a specificity of 0.67, TREWScore achieved a sensitivity of 0.85 and identified patients a median of 28.2 [interquartile range (IQR), 10.6 to 94.2] hours before onset. Of those identified, two-thirds were identified before any sepsis-related organ dysfunction. In comparison,

the Modified Early Warning Score, which has been used clinically for septic shock prediction, achieved a lower

AUC of 0.73 (95%CI, 0.71 to 0.76). A routine screening protocol based on the presence of two of the systemic inflammatory

response syndrome criteria, suspicion of infection, and either hypotension or hyperlactatemia achieved a lower

sensitivity of 0.74 at a comparable specificity of 0.64. Continuous sampling of data from the electronic health

records and calculation of TREWScore may allow clinicians to identify patients at risk for septic shock and provide

earlier interventions that would prevent or mitigate the associated morbidity and mortality.

A second article of interest:

 

A Framework for Individualizing Predictions of Disease

Trajectories by Exploiting Multi-Resolution Structure

 

Peter Schulam

Dept. of Computer Science

Johns Hopkins University

Baltimore, MD 21218

pschulam@jhu.edu

Suchi Saria

Dept. of Computer Science

Johns Hopkins University

Baltimore, MD 21218

ssaria@cs.jhu.edu

For many complex diseases, there is a wide variety of ways in which an individual can manifest the disease. The challenge of personalized medicine is to develop tools that can accurately predict the trajectory of an individual’s disease, which can in turn enable clinicians to optimize treatments. We represent an individual’s disease trajectory as a continuous-valued continuous-time function describing the severity of the disease over time. We propose a hierarchical latent variable model that individualizes predictions of disease trajectories. This model shares statistical strength across observations at different resolutions–the population, subpopulation and the individual level. We describe an algorithm for learning population and subpopulation parameters offline, and an online procedure for dynamically learning individual-specific parameters. Finally, we validate our model on the task of predicting the course of interstitial lung disease, a leading cause of death among patients with the autoimmune disease scleroderma. We compare our approach against state-of-the-art and demonstrate significant improvements in predictive accuracy.

news summary (15)

Categories: Uncategorized

Weekly QuEST Discussion Topics and News, 3 June

QuEST 3 June 2016

The topic this week is first to cover any discussion items from the group since we haven’t been together for a couple of weeks.

One topic I want to remind people – I’m extremely interested in applying QuEST ideas to social and medical issues – specifically what to do about inner city violence and how to do predictive intelligence (for predicting shock onset) – one article I will post for potential discussion is:

Am J Community Psychol (2009) 44:273–286

DOI 10.1007/s10464-009-9268-2

Researching a Local Heroin Market as a Complex Adaptive

System

Lee D. Hoffer • Georgiy Bobashev •

Robert J. Morris

Abstract This project applies agent-based modeling (ABM) techniques to better understand the operation, organization, and structure of a local heroin market. The simulation detailed was developed using data from an 18- month ethnographic case study. The original research, collected in Denver, CO during the 1990s, represents the historic account of users and dealers who operated in the Larimer area heroin market. Working together, the authors studied the behaviors of customers, private dealers, streetsellers, brokers, and the police, reflecting the core elements pertaining to how the market operated. After evaluating the logical consistency between the data and agent behaviors, simulations scaled-up interactions to observe their aggregated outcomes. While the concept and findings from this study remain experimental, these methods represent a novel way in which to understand illicit drug markets and the dynamic adaptations and outcomes they generate. Extensions of this research perspective, as well as its strengths and limitations, are discussed.

 

What I’m in need of here is an ‘action officer’ – someone to take this on and drive it forward by finding related material and pointing me to things to think about and act upon – I can connect that person with an AFOSR contact and a DARPA contact who might be interested in helping us.

We also want to hit the article: Deep residual learning for Image Recognition – from Microsoft research.  For those remote you can find the article at

arXiv:1512.03385v1 [cs.CV] 10 Dec 2015

  • Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously.
  • We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions.
  • We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth.
  • On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8 deeper than VGG nets [41] but still having lower complexity.
  • An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers.
  • The depth of representations is of central importance for many visual recognition tasks.Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset.
  • Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the
  • 1st places on the tasks of ImageNet detection,
  • ImageNet localization,
  • COCO detection, and
  • COCO segmentation.

Heroin market complex adaptive system

microsoft residual learning paper

news summary (14)

Categories: Uncategorized