Category Archives: ToRead

Phil 9.14.17

7:00 – 4:00 ASRC MKT

  • Reducing Dimensionality from Dimensionality Reduction Techniques
    • In this post I will do my best to demystify three dimensionality reduction techniques; PCA, t-SNE and Auto Encoders. My main motivation for doing so is that mostly these methods are treated as black boxes and therefore sometime are misused. Understanding them will give the reader the tools to decide which one to use, when and how.
      I’ll do so by going over the internals of each methods and code from scratch each method (excluding t-SNE) using TensorFlow. Why TensorFlow? Because it’s mostly used for deep learning, lets give it some other challenges 🙂
      Code for this post can be found in this notebook.
    • This seems important to read in preparation for the Normative Mapping effort.
  • Stanford  deep learning tutorial. This is where I got the links to PCA and Auto Encoders, above.
  • Ok, back to writing:
    • The Exploration-Exploitation Dilemma: A Multidisciplinary Framework
    • Got hung up explaining the relationship of the social horizon radius, so I’m going to change it to the exploit radius. Also changed the agent flocks to red and green: GPM
    • There is a bug, too – when I upped the CellAccumulator hypercube size from 10-20. The max row is not getting set
Advertisements

Phil 9.12.17

7:00 – 5:00 ASRC MKT

  • Meeting with Wayne yesterday after Fika. Get him a draft by the end of the week to discuss Monday?
  • More writing
  • Herding in humans (Ramsey M. Raafat, Nick Chater, and Chris Frith)
    • Herding is a form of convergent social behaviour that can be broadly defined as the alignment of the thoughts or behaviours of individuals in a group (herd) through local interaction and without centralized coordination. We suggest that herding has a broad application, from intellectual fashion to mob violence; and that understanding herding is particularly pertinent in an increasingly interconnected world. An integrated approach to herding is proposed, describing two key issues: mechanisms of transmission of thoughts or behaviour between agents, and patterns of connections between agents. We show how bringing together the diverse, often disconnected, theoretical and methodological approaches illuminates the applicability of herding to many domains of cognition and suggest that cognitive neuroscience offers a novel approach to its study.
  • Alignment in social interactions (M.Gallotti, M.T.Fairhurst, C.D.Frith)
    • According to the prevailing paradigm in social-cognitive neuroscience, the mental states of individuals become shared when they adapt to each other in the pursuit of a shared goal. We challenge this view by proposing an alternative approach to the cognitive foundations of social interactions. The central claim of this paper is that social cognition concerns the graded and dynamic process of alignment of individual minds, even in the absence of a shared goal. When individuals reciprocally exchange information about each other’s minds processes of alignment unfold over time and across space, creating a social interaction. Not all cases of joint action involve such reciprocal exchange of information. To understand the nature of social interactions, then, we propose that attention should be focused on the manner in which people align words and thoughts, bodily postures and movements, in order to take one another into account and to make full use of socially relevant information.
  • Herding and escaping responses of juvenile roundfish to square mesh window in a trawl cod end (This is the only case I can find of 3-D stampeding. Note the [required?] dimension reduction)
    • The movements of juvenile roundfish, mainly haddock Melanogrammus aeglefinus and whiting Merlangius merlangus, reacting to a square mesh window in the cod end of a bottom trawl were observed during fishing experiments in the North Sea. Two typical behavioral responses of roundfish are described as the herding response and the escaping response, which were analyzed from video recordings by time sequences of the movement parameters. It was found that most of the actively escaping fish approached the square mesh window at right angles by swimming straight ahead with very little change in direction, while most of the herded fish approached the net at obtuse angles and retreated by sharp turning. The herding and escaping responses showed significant difference when characterized by frequency distributions of swimming speed and angular velocity, and both responses showed large and irregular variations in swimming movement parameters like the panic erratic responses. It is concluded that an escaping or herding response to the square mesh window could be decided by an interaction between the predictable parameters that describe the stimuli of net and angular changes of fish response, such as approaching angle, turning angle and angular velocity.
  • Assessing the Effect of “Disputed” Warnings and Source Salience on Perceptions of Fake News Accuracy
    • What are effective techniques for combating belief in fake news? Tagging fake articles with “Disputed by 3rd party fact-checkers” warnings and making articles’ sources more salient by adding publisher logos are two approaches that have received large-scale rollouts on social media in recent months. Here we assess the effect of these interventions on perceptions of accuracy across seven experiments (total N=7,534). With respect to disputed warnings, we find that tagging articles as disputed did significantly reduce their perceived accuracy relative to a control without tags, but only modestly (d=.20, 3.7 percentage point decrease in headlines judged as accurate). Furthermore, we find a backfire effect – particularly among Trump supporters and those under 26 years of age – whereby untagged fake news stories are seen as more accurate than in the control. We also find a similar spillover effect for real news, whose perceived accuracy is increased by the presence of disputed tags on other headlines. With respect to source salience, we find no evidence that adding a banner with the logo of the headline’s publisher had any impact on accuracy judgments whatsoever. Together, these results suggest that the currently deployed approaches are not nearly enough to effectively undermine belief in fake news, and new (empirically supported) strategies are needed.
  • Some meetings on marketing. Looks like we’re trying to get on this panel. Wrote bioblurbs!
  • More writing. Reasonable progress.

Phil 8.29.16

7:00 – 6:00 ASRC

  • Selective Use of News Cues: A Multiple-Motive Perspective on Information Selection in Social Media Environments – Quite close to the Explorer/Confirmer/Avoider study but using a custom(?) browsing interface that tracked the marking of news stories to read later. Subjects were primed for a task with motivations – accuracy, defense and impression. Added this to paragraph 2.9, where explorers are introduced.
  • Looked through Visual Complexity – Mapping Patterns of Information, and it doesn’t even mention navigation. Most information mapping efforts are actually graphing efforts. Added a paragraph in section 2.7
  • Added a TODO for groupthink/confirmation bias, etc.
  • Chat with Heath about AI.He’s looking to build a MUD agent and will probably wind up learning WEKA, etc. so a win, I think.
  • Working on getting the configurator to add string values.
  • Added to DocumentStatistics. Need to switch over to getSourceInfo() from getAddressStrings in the Configurator.
  • Meeting with Wayne about the proposal. One of the branches of conversation went into some research he did on library architecture. That’s been rattling around in my head.
    We tend to talk about interface design where the scale is implicitly for the individual. The environment where these systems function is often thought of as an ecosystem, with the Darwinian perspective that goes along with that. But I think that such a perspective leads to ‘Survival of the Frictionlesss’, where the easiest thing to use wins and damn the larger consequences.
    Reflecting on how the architecture and layout of libraries affected the information interactions of the patrons, I wonder whether we should be thinking about Information Space Architecture. Such a perspective means that the relationships between design at differing scales needs to be considered. In the real world, architecture can encompass everything from the chairs in a room to the landscaping around the building and how that building fits into the skyline.
    I think that regarding information spaces as a designed continuum from the very small to very large is what my dissertation is about at its core. I want a park designed for people, not a wilderness, red in tooth and claw.

Phil 8.26.16

7:00 – 4:00 ASRC

    • Adding more model feedback
    • Something more to think about WRT Group Polarization models? Collective Memory and Spatial Sorting in Animal Groups
    • Need to be able to associate an @attribute  key/value map with Labeled2Dmatrix rows so that we can compare different nominal values across a shared set of numeric columns. This may wind up being a derived class?
      • Working on adding an array of key/value maps;
      • Forgot to add the name to the @data section – oops!
      • text is added to ARFF out. Should I add it to the xlsx outputs as well?
    • Here’s the initial run against the random test data within the class (L2D.arff).
=== Run information ===

Scheme: weka.classifiers.bayes.NaiveBayes
Relation: testdata
Instances: 8
Attributes: 12
name
sv1
sv2
sv3
p1
p2
p3
p4
s1
s2
s3
s4
Test mode: split 66.0% train, remainder test

=== Classifier model (full training set) ===

Naive Bayes Classifier

Class
Attribute p1 p2 p3 p4 s1 s2 s3 s4
(0.13) (0.13) (0.13) (0.13) (0.13) (0.13) (0.13) (0.13)
=======================================================================
sv1
p4-sv1 1.0 1.0 1.0 2.0 1.0 1.0 1.0 1.0
s2-sv1 1.0 1.0 1.0 1.0 1.0 2.0 1.0 1.0
p2-sv1 1.0 2.0 1.0 1.0 1.0 1.0 1.0 1.0
s1-sv1 1.0 1.0 1.0 1.0 2.0 1.0 1.0 1.0
[total] 4.0 5.0 4.0 5.0 5.0 5.0 4.0 4.0

sv2
p2-sv2 1.0 2.0 1.0 1.0 1.0 1.0 1.0 1.0
s4-sv2 1.0 1.0 1.0 1.0 1.0 1.0 1.0 2.0
p1-sv2 2.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
s1-sv2 1.0 1.0 1.0 1.0 2.0 1.0 1.0 1.0
[total] 5.0 5.0 4.0 4.0 5.0 4.0 4.0 5.0

sv3
p2-sv3 1.0 2.0 1.0 1.0 1.0 1.0 1.0 1.0
p1-sv3 2.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0
s4-sv3 1.0 1.0 1.0 1.0 1.0 1.0 1.0 2.0
p3-sv3 1.0 1.0 2.0 1.0 1.0 1.0 1.0 1.0
p4-sv3 1.0 1.0 1.0 2.0 1.0 1.0 1.0 1.0
s2-sv3 1.0 1.0 1.0 1.0 1.0 2.0 1.0 1.0
s1-sv3 1.0 1.0 1.0 1.0 2.0 1.0 1.0 1.0
[total] 8.0 8.0 8.0 8.0 8.0 8.0 7.0 8.0

p1
mean 1 0 0 0 1 1 0 0
std. dev. 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667
weight sum 1 1 1 1 1 1 1 1
precision 1 1 1 1 1 1 1 1

p2
mean 0 1 0 0 1 0 1 0
std. dev. 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667
weight sum 1 1 1 1 1 1 1 1
precision 1 1 1 1 1 1 1 1

p3
mean 0 0 1 0 1 0 0 1
std. dev. 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667
weight sum 1 1 1 1 1 1 1 1
precision 1 1 1 1 1 1 1 1

p4
mean 0 0 0 1 1 0 0 1
std. dev. 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667
weight sum 1 1 1 1 1 1 1 1
precision 1 1 1 1 1 1 1 1

s1
mean 1 1 1 1 1 0 0 0
std. dev. 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667
weight sum 1 1 1 1 1 1 1 1
precision 1 1 1 1 1 1 1 1

s2
mean 1 0 0 0 0 1 0 0
std. dev. 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667
weight sum 1 1 1 1 1 1 1 1
precision 1 1 1 1 1 1 1 1

s3
mean 0 1 0 0 0 0 1 0
std. dev. 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667
weight sum 1 1 1 1 1 1 1 1
precision 1 1 1 1 1 1 1 1

s4
mean 0 0 1 1 0 0 0 1
std. dev. 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667 0.1667
weight sum 1 1 1 1 1 1 1 1
precision 1 1 1 1 1 1 1 1



Time taken to build model: 0 seconds

=== Evaluation on test split ===

Time taken to test model on training split: 0 seconds

=== Summary ===

Correctly Classified Instances 0 0 %
Incorrectly Classified Instances 3 100 %
Kappa statistic 0
Mean absolute error 0.2499
Root mean squared error 0.4675
Relative absolute error 108.2972 %
Root relative squared error 133.419 %
Total Number of Instances 3

=== Detailed Accuracy By Class ===

TP Rate FP Rate Precision Recall F-Measure MCC ROC Area PRC Area Class
0.000 0.333 0.000 0.000 0.000 0.000 ? ? p1
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.333 p2
0.000 0.333 0.000 0.000 0.000 0.000 ? ? p3
0.000 0.000 0.000 0.000 0.000 0.000 ? ? p4
0.000 0.000 0.000 0.000 0.000 0.000 0.500 0.500 s1
0.000 0.000 0.000 0.000 0.000 0.000 1.000 1.000 s2
0.000 0.333 0.000 0.000 0.000 0.000 ? ? s3
0.000 0.000 0.000 0.000 0.000 0.000 ? ? s4
Weighted Avg. 0.000 0.000 0.000 0.000 0.000 0.000 0.500 0.611

=== Confusion Matrix ===

a b c d e f g h <-- classified as
0 0 0 0 0 0 0 0 | a = p1
0 0 0 0 0 0 1 0 | b = p2
0 0 0 0 0 0 0 0 | c = p3
0 0 0 0 0 0 0 0 | d = p4
0 0 1 0 0 0 0 0 | e = s1
1 0 0 0 0 0 0 0 | f = s2
0 0 0 0 0 0 0 0 | g = s3
0 0 0 0 0 0 0 0 | h = s4
  • Need to add text data from xml or from other(wrapper info? structured data? UI selections?) sources