ASRC GOES 7:00 – 4:30

• Dissertation
  • Conclusions – starting the hypothesis section
• Evolver
  • Finish documentation
  • Add session_clear() to TF2OptimizationTest
  • Deploy to PiPy
    • Folowed a mix of these directions:
    • How to upload your python package to PyPi
    • How to Publish an Open-Source Python Package to PyPI
    • Packaging and distributing projects
    • The legacy test upload doesn’t work, but

      twine upload dist/*

      did create the project. And optevolver is on pypi!

  • Need to set the develop branch to the main branch. Monday
• AIMS NextGen
  • Isaac made really nice progress on the Sim + ML plan
  • Sent him the schedule

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
  • Part 0 – Introduction
  • Part 1 – Scikit-Optimize
• 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
  • How to upload your python package to PyPi
  • How to Publish an Open-Source Python Package to PyPI
• 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?

7:00 – 8:00 ASRC

• Saw this on Twitter this morning: Training Agents using Upside-Down Reinforcement Learning
  • Traditional Reinforcement Learning (RL) algorithms either predict rewards with value functions or maximize them using policy search. We study an alternative: Upside-Down Reinforcement Learning (Upside-Down RL or UDRL), that solves RL problems primarily using supervised learning techniques. Many of its main principles are outlined in a companion report [34]. Here we present the first concrete implementation of UDRL and demonstrate its feasibility on certain episodic learning problems. Experimental results show that its performance can be surprisingly competitive with, and even exceed that of traditional baseline algorithms developed over decades of research.
• I wonder how it compares with Stuart Russell’s paper Cooperative Inverse Reinforcement Learning
  • For an autonomous system to be helpful to humans and to pose no unwarranted risks, it needs to align its values with those of the humans in its environment in such a way that its actions contribute to the maximization of value for the humans. We propose a formal definition of the value alignment problem as cooperative inverse reinforcement learning (CIRL). A CIRL problem is a cooperative, partial- information game with two agents, human and robot; both are rewarded according to the human’s reward function, but the robot does not initially know what this is. In contrast to classical IRL, where the human is assumed to act optimally in isolation, optimal CIRL solutions produce behaviors such as active teaching, active learning, and communicative actions that are more effective in achieving value alignment. We show that computing optimal joint policies in CIRL games can be reduced to solving a POMDP, prove that optimality in isolation is suboptimal in CIRL, and derive an approximate CIRL algorithm.
• Dissertation
  • In the Ethics section, change ‘civilization’ to ‘culture’, and frame it in terms of the simulation – done
  • Last slide should be ‘Thanks for coming to my TED talk’
  • Ping Don’s composer and choreographer, if I can find them
  • Cool! A T-O style universe map (Unmismoobjetivo , via Wikipedia). The logarithmic distance effect is something that I need to look into:
• Evolver
  • Quickstart
  • User’s guide
  • Finished commenting!
  • Flailing on geting the documentation tools to work.
    • Installing Sphinx
• ML Seminar
  • Double Crab Cake Platter (2) – 2 Vegetables – $34.00
  • Went over the Evolver. The Ensemble charts really make an impression, but overall, the code walkthrough is too difficult – there are two many moving parts. I need to write a paper with screengrabs that walk through the whole process. I’ll need to evaluate against Bayesian tuners, but I also have architecture search
  • The venue could be IEEE ICTAI 2020: The IEEE International Conference on Tools with Artificial Intelligence (ICTAI) is a leading Conference of AI in the Computer Society providing a major international forum where the creation and exchange of ideas related to artificial intelligence are fostered among academia, industry, and government agencies. It will be in Baltimore, I think.
• Meeting with Aaron. He thinks that part of the ethics discussion needs to be an addressing of the status quo

7:00 – 8:00 ASRC GOES

• Dissertation – back to designing for populations
• Timesheet revisions
• Applying for MS Project
• Evolver – more documentation
• GOES Meeting
  • Bought a copy of MS Project for $15
  • Send Erik a note about permission to charge for TF Dev Conf
  • Good chat with Bruce about many things, including CASSIE as a Cloud service
  • Re-send links to common satellite dictionary
  • Vadim got a pendulum working
• Meeting with Roger
  • Got a tour of the new building
  • Lots of VR discussion
  • Some academic future options

7:00 – 4:00 ASRC GOES

• Dissertation – reworked the last paragraph of the Reflection and reflex section
  • Spooky to see this: Adversarial Examples that Fool both Computer Vision and Time-Limited Humans
    • Citing papers including ‘human’
• Evolver – more documentation
• Send this out to the HCC mailing list: The introvert’s academic “alternative networking” guide
• Arpita’s proposal defense
  • Stanford: Open information extraction (open IE) refers to the extraction of relation tuples, typically binary relations, from plain text, such as (Mark Zuckerberg; founded; Facebook). The central difference from other information extraction is that the schema for these relations does not need to be specified in advance; typically the relation name is just the text linking two arguments. For example, Barack Obama was born in Hawaii would create a triple (Barack Obama; was born in; Hawaii), corresponding to the open domain relation was-born-in(Barack-Obama, Hawaii).
  • Open Information Extraction 5
  • UKG Open Information Extraction
  • Supervised Ensemble of Open IE
  • Datasets
    • AW-OIE
    • AW-OIE-C
    • WEB
    • NYT
    • PENN
  • Why the choice of 100 dimensins for your symentic embedding? How does it compare to other dimensions?
  • Contextual embedding for NLP?
  • Input-Output Hidden Markov Model (version on GitHub)