Category Archives: Machine Learning

Phil 1.23.2020

7:00 – 6:00 ASRC GOES

Check Copenhagen Wheel serial number for this recall

  • Get comments back to Antonio tonight! – done
  •  Dissertation
    • Writing chapter summaries
      • Research design – done
      • Simulation – done
      • Adversarial herding – done
      • Belief space cartography – done
      • Human study – done
      • Discussion – done
      • Conclusions – done!
      • Sent pdfs to common vision. Need to set the pdfs out with a note tomorrow morning
    • Caught up with Wayne a bit and commented out the “Why this is HCC section” Keep it for a backup slide though
  • Very interesting Invisibilia on AI
  • GSAW prep
    • Registration
  • Told Aaron about OpML 20. It’s only two pages and due on Feb 25. Maybe a quick writeup of Optevolver?

Phil 1.21.20

7:00 – 6:00 ASRC GOES

  • Dissertation
    • Chasing TODOs
    • TODO: Add transition paragraph (ch_background.tex) – done
    • TODO: stiff, moving platform (ch_background.tex) – done
    • TODO: clarify multicellular vs individuals vs dangerous stampedes Connect the lists. (sec_biological_basis.tex) – done
  • GSAW prep
    • Tix and hotel
  • TF Dev Conf
    • Tix and hotel

Phil 1.20.20

Transformers from Scratch

  • Transformers are a very exciting family of machine learning architectures. Many good tutorials exist, but in the last few years transformers have mostly become simpler, so that it is now much more straightforward to explain how modern architectures work. This post is an attempt to explain directly how modern transformers work, and why, without some of the historical baggage.

Dissertation

  • Folding in Wayne’s edits
    • Made the Arendt paragraph of velocity less reflective and more objective.
    • TODO: Defend facts to opinion with examples of language, framing, what is interesting, etc.-done
    • TODO: Heavy thoughts, light and frivolous, etc. We ascribe these, but they are not there – done
    • TODO: We have a MASSIVE physical bias. Computers don’t. Done
    • TODO: COmputers and people must work together
  • Title case all refs (Section, Table, etc) – done
  • \texttt all urls (reddit, etc) – done
  • search for and / or slashes
  • Fix underlines as per here– done! 
    % for better underlining
\usepackage[outline]{contour}
\usepackage{ulem}
\normalem % use classical emph

\newcommand \myul[4]{%
	\begingroup%
	\renewcommand \ULdepth {#1}%
	\renewcommand \ULthickness {#2}%
	\contourlength{#3}%
	\uline{\phantom{#4}}\llap{\contour{white}{#4}}%
	\endgroup%
}

     

Phil 1.15.20

I got invited to the TF Dev conference!

The HKS Misinformation Review is a new format of peer-reviewed, scholarly publication. Content is produced and “fast-reviewed” by misinformation scientists and scholars, released under open access, and geared towards emphasizing real-world implications. All content is targeted towards a specialized audience of researchers, journalists, fact-checkers, educators, policy makers, and other practitioners working in the information, media, and platform landscape.

  • For the essays, a length of 1,500 to 3,000 words (excluding footnotes and methodology appendix) is appropriate, but the HKS Misinformation Review will consider and publish longer articles. Authors of articles with more than 3,000 words should consult the journal’s editors before submission.

7:00 – ASRC GOES

  •  Dissertation
    • It looks like I fixed my LaTeX problems. I went to C:\Users\phil\AppData\Roaming\MiKTeX\2.9\tex\latex, and deleted the ifvtex folder. Re-ran, things installed, and all is better now
    • Slides
  • GOES
    • Pinged Isaac about the idea of creating scenarios that incorporate the NASA simulators
    • Meeting
  • GSAW
    • Slides
    • Speakers presenting in a plenary session are scheduled to speak for 15 minutes, with five additional minutes allowed for questions and answers from the audience
    • Our microphones work best when the antenna unit is clipped to a belt and the microphone is attached near the center of your chest.
    • We are NOT providing network capabilities such as WiFi. If you require WiFi, you are responsible for purchasing it from the hotel and ensuring that it works for the presentation.
    • Charts produced by the PC version of Microsoft PowerPoint 2013, 2016 or 365 are preferred
    • . In creating your slides, note that the presentation room is large and you should consider this in your selection of larger fonts, diagram size, etc. At a minimum, a 20-point font is recommended
  • GPT-2 – Maybe do something with Aaron today?

Phil 1.10.20

7:00 – 4:30 ASRC PhD, BD, GOES

  • Dissertation
    • Stampedes are a form of runaway attention, and precision/recall aid that process
    • Starting on forward. Using the Arab Spring and GamerGate as the framing
  • 11:00 VOLPE Meeting
    • Pursuing the resilience proposal was well received. Next, go up and meet with the folks?
  • Install card – done! Passed the smoke test

Phil 12.31.19

CodeIt

Get tix for ET 2020

7:00 – 4:30 PhD

  • Starting slides as a way to do the chapter overviews and summaries
  • GPT-2 agents
    • Got rid of Huggingface’s transformers library. Too much hidden stuff to understand
    • Aaron found a couple of other projects on GitHub – trying those
    • Downloaded the 715M model and associated files

And I’m a guest editor!IEEE

Phil 1.30.19

7:00 – 7:00 ASRC PhD

ClimateTree

  • Nice visualization, with map-like aspects: The Climate Learning Tree
  •  Dissertation
    • Start JuryRoom section – done!
    • Finished all content!
  • GPT-2 Agents
    • Download big model and try to run it
    • Move models and code out of the transformers project
  • GOES
    • Learning by Cheating (sounds like a mechanism for simulation to work with)
      • Vision-based urban driving is hard. The autonomous system needs to learn to perceive the world and act in it. We show that this challenging learning problem can be simplified by decomposing it into two stages. We first train an agent that has access to privileged information. This privileged agent cheats by observing the ground-truth layout of the environment and the positions of all traffic participants. In the second stage, the privileged agent acts as a teacher that trains a purely vision-based sensorimotor agent. The resulting sensorimotor agent does not have access to any privileged information and does not cheat. This two-stage training procedure is counter-intuitive at first, but has a number of important advantages that we analyze and empirically demonstrate. We use the presented approach to train a vision-based autonomous driving system that substantially outperforms the state of the art on the CARLA benchmark and the recent NoCrash benchmark. Our approach achieves, for the first time, 100% success rate on all tasks in the original CARLA benchmark, sets a new record on the NoCrash benchmark, and reduces the frequency of infractions by an order of magnitude compared to the prior state of the art. For the video that summarizes this work, see this https URL
  • Meeting with Aaron
    • Overview at the beginning of each chapter – look at Aaron’s chapter 5 for
    • example intro and summary.
    • Callouts in text should match the label
    • hfill to right-justify
    • Footnote goes after puntuation
    • Punctuation goes inside quotes
    • for url monospace use \texttt{} (perma.cc)
    • indent blockquotes 1/2 more tab
    • Non breaking spaces on names
    • Increase figure sizes in intro

Phil 12.28.19

Calculating Political Bias and Fighting Partisanship with AI

  • I tried this with the abstract to a paper that Google Scholar has been suggesting I read:
  • The Thorny Challenge of Making Moral Machines: Ethical Dilemmas with Self-Driving Cars
    • The algorithms that control AVs will need to embed moral principles guiding their decisions in situations of unavoidable harm. Manufacturers and regulators are confronted with three potentially incompatible objectives: being consistent, not causing public outrage, and not discouraging buyers. The presented moral machine study is a step towards solving this problem as it tries to learn how people all over the world feel about the alternative decisions the AI of self-driving vehicles might have to make. The global study displayed broad agreement across regions regarding how to handle unavoidable accidents. To master the moral challenges, all stakeholders should embrace the topic of machine ethics: this is a unique opportunity to decide as a community what we believe to be right or wrong, and to make sure that machines, unlike humans, unerringly follow the agreed-upon moral preferences. The integration of autonomous cars will require a new social contract that provides clear guidelines about who is responsible for different kinds of accidents, how monitoring and enforcement will be performed, and how trust among all stakeholders can be engendered.

      The online analyzer (https://www.thebipartisanpress.com/analyze-bias/#). Found this to have some left bias. I think its unable to distinguish between opinion presentation and fact presentation?

Phil 12.26.19

ASRC PhD 7:00 – 4:00

  • Dissertation
    • Limitations
  • GPT-2 agents setup – set up the project, but in the process of getting the huggingface transformers, I wound up setting up that project as well
    • Following directions for
      • pip install transformers
      • git clone https://github.com/huggingface/transformers
        • cd transformers
        • pip install .
      • pip install -e .[testing]
        • make test – oops. My GNU Make wasn’t on the path – fixed it
        • running tests
          • Some passed, some failed. Errors like: tests/test_modeling_tf_t5.py::TFT5ModelTest::test_compile_tf_model Fatal Python error: Aborted
          • Sure is keeping the processor busy… Like bringing the machine to its knees busy….
          • Finished – 14 failed, 10 passed, 196 skipped, 20 warnings in 1925.12s (0:32:05)
  • Fixed the coffee maker
  • Dealt with stupid credit card nonsense

Phil 12.23.19

7:00 – 4:30 ASRC

  • 2020 International Conference on Social Computing, Behavioral-Cultural Modeling, & Prediction and Behavior Representation in Modeling and Simulation
    • SBP-BRiMS is an interdisciplinary computational social science conference focused on both modeling complex socio-technical systems and using computational techniques to reason about and study complex socio-technical systems. The participants in this conference take part in forming the conversation on how computation is shaping the modern world and helping us to better understand and reason about human behavior. Both papers addressing basic research and those addressing applied research are accepted. All methodological approaches are encouraged; however, the vast majority of papers use computer simulation, network analysis or machine learning as the method of choice in addressing human social and behavioral activities. At the conference, these paper presentations are complemented by data science challenge problems, demonstrations of new technologies, and a government funding panel.
    • Regular Paper Submission (10 – page max) : 21-February-2020 (Midnight EST)
    • Tuesday, July 14, 2020 – Friday, July 17, 2020 George Washington University, Washington DC, USA
  • Dissertation
    • More conclusions. Got through H2
  • Evolver
    • Figuring out how to merge changes from develop onto master. Hooray – success! The IntelliJ directions (here) were very helpful.
    • And everything is now visible on GitHub

Phil 12.20.19

ASRC GOES 7:00 – 4:30

Phil 12.19.19

7:00 – 4:30 ASRC GOES

  • Dissertation
    • Conclusions – got through the intro and starting the hypothesis section
  • NASA GitHub
  • Evolver
    • More documentation for sure, maybe more debugging?
    • Had to update my home system
    • Looks like the fix is working. I ran it again, and no problems
    • A little more documentation before heading down to the NSOF
  • Simulations
    • Meeting with Isaac – Lots of discussion. The question is how to handle the simulations. NOAA is used to these and has extremely high fidelity ones, but we need sims that can train on many permutations. Here’s an IEEE article on augmented reality training robocars that should be cited
      • industry must augment road testing with other strategies to bring out as many edge cases as possible. One method now in use is to test self-driving vehicles in closed test facilities where known edge cases can be staged again and again.
      • Computer simulation provides a way around the limitations of physical testing. Algorithms generate virtual vehicles and then move them around on a digital map that corresponds to a real-world road. If the data thus generated is then broadcast to an actual vehicle driving itself on the same road, the vehicle will interpret the data exactly as if it had come from its own sensors. Think of it as augmented reality tuned for use by a robot.
  • NSOF Meeting
    • UI demonstrations
    • Got my card activated!

Phil 12.18.19

7:00 – 5:30 ASRC GOES

  • Recalls V46 and VB2/NHSTA 19V-818
  • Fireplace
  • Dissertation
    • Pull in Rachel’s comments – done
    • Begin conclusions!
  • More documentation.
    • Creating the readme for the TF2_opt_example
    • Created the new file, and verifying that everything works – looking good
    • Whoops! I was still using 
      from tensorflow_core.python.keras import layers
    • instead of 
      from tensorflow.keras import layers
    • which gave me a tensorflow/core/common_runtime/bfc_allocator.cc:905] InUse at error, at least according to this. Going to have to update the library.
    • Nope – that didn’t work. Trying to clear the GPU directly using cuda libaries as described here 
      • That causes the execution to stop. I think you have to do something to re-open the GPU
    • Trying Keras clear_session(). It’s tricky, because it can’t be in the GPU context. Seeing if it works in the loop that creates the TFOptimizerTest object.
      • That worked! Just worried that it might have to do with the complexity of the model. THis time, the evolver came up with a 980 neuron, one layer architecture. Last time, it choked on 800 X 5. Rerunning.
  • More on hyperparameter optimization (HPO). These articles goes into the scikit libraries
  • An alternate take: An Introductory Example of Bayesian Optimization in Python with Hyperopt A hands-on example for learning the foundations of a powerful optimization framework
  • Deploy to PiPy
  • Mission Drive meetings
    • Satellite tool kit? STK’s physics-based, multi-domain modeling, simulation, and analysis environment supports the fast, cost-effective, and responsive approaches needed to realize the full value of digital engineering.
    • What’s new in STK 11.7
    • Set up a one hour meeting tomorrow before the main meeting at the NSOF with Isaac. Something about how to recognize the pattern of switching from one satellite ground station to another.
  • In general, Bing directs users to conspiracy-related content, even if they aren’t explicitly looking for it. For example, if you search Bing for comet ping pong, you get Pizzagate-related content in its top 50 results. If you search for fluoride, you get content accusing the U.S. government of poisoning its population. And if you search for sandy hook shooting, you will find sources claiming that the event was a hoax. Google does not show users conspiracy-related content in its top 50 results for any of these queries. (Stanford Internet Observatory)
  • In 2000, Lucas Introna and Helen Nissenbaum published a paper called “Shaping the Web: Why the Politics of Search Engines Matters.” Examining how the internet had developed to that point and where it was likely to go next, Introna and Nissenbaum identified a specific threat facing the public: search engines, they argued, could conceivably be “colonized by specialized interests at the expense of the public good” and cease to be reliable, more or less transparent sources of information. If the authors’ fears of rampant commercialism affecting the way search engines operate were prophetic, it has also become clear that commercial interests are only part of the problem. If Google became a public utility tomorrow, societies would still have to come up with ethical standards for how to deal with harmful content and the vectors, such as data voids, by which it reaches users. 
    • Add cite to the “diversity is algorithmically crowded out” line in the ethical considerations section?

Phil 12.10.19

7:00 – ASRC GOES

  • Dissertation – got through the stories and games section. Then de-emphasizing lists, etc.
  • LMN prep (done) and demo
  • Evolver
    • Migrate to cookie cutter – done
    • Github – done
    • Try to make a package – done!
    • Start on paper/tutorial for IEEE ICTAI 2020. Need to compare against Bayesian system. Maybe just use the TF optimizer? Same models, same data, and they are very simple