Author Archives: pgfeldman

Phil 2.12.21

Nice little number sequence of a day. For the US at least.

Cool thing! This is pageviews.toolforge.org, that lets you look at and compare Wikipedia page views. These are subjects of recent Google Doodles, which shows the power of diversity injection using ‘random’ prompts.

https://pageviews.toolforge.org/?project=en.wikipedia.org&platform=all-access&agent=user&redirects=0&start=2021-01-01&end=2021-02-11&pages=Justicia_Acu%C3%B1a|Fredy_Hirsch|Shadia|Jim_Wong-Chu|Petrona_Eyle

Book

  • 2:00 Meeting with Michelle

GOES

  • Add code to AngleController for RW efficiency – done. Not sure if the scripted reaction wheel efficiency failures is being applied correctly. Right now it’s in AngleController as a commanded value to the Rwheels. I think it should probably be handled in the simulation proper
  • Add padding to DictionaryEntries for variables that are added later – done

GPT Agents

  • Still Ingesting
  • Intrinsic vs. Extrinsic Structure
  • Graphs and networks
  • A Comprehensive Survey of Graph Embedding: Problems, Techniques and Applications
  • A graph is hypothetical structure that does not exist in the real world. Any real world data representation may be reduced to a graph, but it is NOT a graph. A network topology can be first reduced to a graph and then algorithms like Dijkstra’s and Kruskal’s can be applied to it for various purposes like routing. All real world situations are subsets of graph problems (of that domain). Most problems of networking, especially those concerning routing reduce to a graph based problem. You will never find a graph in real world scenario. But you will find its subsets or derivatives in practical applications.

Jury Room

  • Read Jarod Grovers’ COMPX593 Thesis proposal

ML Group

  • 3:30 Meeting. Added some topics to the paper list

Phil 2.11.21

Diversity injection works!

https://twitter.com/seanjtaylor/status/1359760058603565057

Named Tensor Notation

https://namedtensor.github.io/

Cleora is a general-purpose model for efficient, scalable learning of stable and inductive entity embeddings for heterogeneous relational data.

Book – shoveled snow instead

GPT Agents

  • Couldn’t connect with Jay via email? Sent links by Twitter DM
  • Got started on ingesting data

GOES

  • Modify Maneuver Creator, Writer, and Reader to handle DataDictionary entries
  • Had to update ValueAxis to handle reaction wheel efficiency permutations
  • Added writing and reading DictionaryEntrys to the DataDictionary class. Pretty happy with it, though I need to work with the history some?
  • 2:00 Meeting

Phil 2.10.21

Helping out another older PhD student with Sage Advice. Hopefully it’s helpful!

Book

  • More sentences and paragraphs

GPT Agents

  • Download and ingest latest data (and keep track of row ids!)
  • Get TF, HF, Ecco, and drivers working right

GOES

  • 3:00 Meeting
  • Start adding rw efficiency and other options to the script. I should be able to make dictionary entries directly and set them
  • Add rw efficiency multiplier to AngleController. May have to add a method to see if an entry exists?

MDS

  • 10:00 Meeting

JuryRoom

  • 5:00 Meeting
  • Put a list of journals for publications in 2021
  • Journal paper is submitted to Elsevier Journal of Cultural Heritage
  • Conference paper is submitted to ACM Computer Human Interaction
  • Maori SW dev article

Phil 2.9.21

Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention

  • Transformers have emerged as a powerful tool for a broad range of natural language processing tasks. A key component that drives the impressive performance of Transformers is the self-attention mechanism that encodes the influence or dependence of other tokens on each specific token. While beneficial, the quadratic complexity of self-attention on the input sequence length has limited its application to longer sequences — a topic being actively studied in the community. To address this limitation, we propose Nyströmformer — a model that exhibits favorable scalability as a function of sequence length. Our idea is based on adapting the Nyström method to approximate standard self-attention with O(n) complexity. The scalability of Nyströmformer enables application to longer sequences with thousands of tokens. We perform evaluations on multiple downstream tasks on the GLUE benchmark and IMDB reviews with standard sequence length, and find that our Nyströmformer performs comparably, or in a few cases, even slightly better, than standard Transformer. Our code is at this https URL.

Book – Working on snippets

GPT Agents

  • 10:00 Meeting with Jay Alammar – that went *really* well!
  • 3:00 Meeting that I’m going to be a bit late for

GOES

  • 11:00 Meeting with Vadim
  • Nothing obvious about the roll problems, but that shouldn’t stop the work on getting the pitch maneuver running and generating data
    • Add rwheel efficiencies to the script in some kind of loop
    • Adjust AngleController to use rwheel efficiency from the ddict
  • Got my intersection code running well
  • 10:00 Meeting for first period report. Hopefully we’ll be able to get a meeting with the TPOC next week to see what he wants?

Phil 2.8.21

Winter has arrived

The desire for dominance displays from The Republican Party Is Radicalizing Against Democracy

  • As for the party’s base, what policy issues are MAGA rally-goers wound up about? Not the deficit or taxes, and not the ACA. In the past, those issues gave expression to their underlying grievances, but no longer. After the election, one GOP polling firm asked Republicans about their biggest concerns for a post-Trump Republican Party. Forty-four percent wanted a party that would “fight like Donald Trump,” while only 19 percent worried that a post-Trump GOP would “abandon Donald Trump’s policies.”

Book

  • Work on short descriptions

GPT Agents

  • More Ecco. Set up meeting for tomorrow morning? Yes! 10:00

GOES

  • 11:00 Meeting with Vadim. Rescheduled for tomorrow
  • Verified that the rotations are working right in the testbed:
  • Aiming a projectile at a moving target
    • Without gravity, bullets travel in a straight line at constant velocity. To determine where to shoot, imagine shooting an infinite number of bullets simultaneously in all directions, forming an expanding sphere of death. While the sphere is expanding, the target is moving along a straight line (or standing still). If we can determine the instant at which this expanding sphere touches the moving target, we know where to aim: at the target’s position when it gets hit by the expanding sphere of death.
    • Hooray!
# based on this blog post: https://playtechs.blogspot.com/2007/04/aiming-at-moving-target.html
def calc_intersect_time(p:np.array, v:np.array, s:float) -> float:
print("calc_intersect_time()")
a = s * s - np.sum(np.square(v))
bv = p * v
b = np.sum(bv)
cv = np.square(p)
c = np.sum(cv)

disc = b*b + a*c

t = 0
if disc >= 0:
t = (b + np.sqrt(disc)) / a
return max(0, t)

return t

Phil 2.5.21

They Stormed the Capitol. Their Apps Tracked Them

  • While there were no names or phone numbers in the data, we were once again able to connect dozens of devices to their owners, tying anonymous locations back to names, home addresses, social networks and phone numbers of people in attendance. In one instance, three members of a single family were tracked in the data.

Book

  • 2:00 Meeting with Michelle

GOES

  • Run pure pitch/roll/yaw maneuvers
  • Yaw – really nice
https://viztales.files.wordpress.com/2021/02/image-4.png
  • Roll – WTaF?
https://viztales.files.wordpress.com/2021/02/image-5.png
  • Pitch – ok?
https://viztales.files.wordpress.com/2021/02/image-6.png
  • 11:00 Meeting with Vadim

GPT Agents

  • Install Ecco and try the intro (here in the video, also ranked choices, and token probability) locally, then message him to see if he’d like to discuss maps.
  • Have the code running locally (the NVIDIA update worked too!):
  • Messaged Jay Alamar on twitter. Working on setting up a video chat
  • Woohoo!
The IJCAI-21 summary reject phase has now ended and we are pleased to inform you that your submission #1558 "Navigating Human Language Models with Synthetic Agents" will enter the full-paper review phase. 
  • 3:30 Meeting

Phil 2.4.21

Book

  • Tweak the conclusions to add flocking and stampeding – closer to really done?

GOES

  • Finish adding script management – done!
  • There are issues though:
  • Adding a ManeuverCreator class that will produce pure pitch, roll, and yaw scripts
  • Status report – done!

JuryRoom

  • 5:00 Meeting with Priyank

Phil 2.2.21

Can Small and Synthetic Benchmarks Drive Modeling Innovation? A Retrospective Study of Question Answering Modeling Approaches

  • Datasets are not only resources for training accurate, deployable systems, but are also benchmarks for developing new modeling approaches. While large, natural datasets are necessary for training accurate systems, are they necessary for driving modeling innovation? For example, while the popular SQuAD question answering benchmark has driven the development of new modeling approaches, could synthetic or smaller benchmarks have led to similar innovations?
  • This counterfactual question is impossible to answer, but we can study a necessary condition: the ability for a benchmark to recapitulate findings made on SQuAD. We conduct a retrospective study of 20 SQuAD modeling approaches, investigating how well 32 existing and synthesized benchmarks concur with SQuAD — i.e., do they rank the approaches similarly? We carefully construct small, targeted synthetic benchmarks that do not resemble natural language, yet have high concurrence with SQuAD, demonstrating that naturalness and size are not necessary for reflecting historical modeling improvements on SQuAD. Our results raise the intriguing possibility that small and carefully designed synthetic benchmarks may be useful for driving the development of new modeling approaches.

Book

  • Finished the Charles Dawson and Alfred Wegoner bits. Working on play

GOES

  • Finally got the sim working again! Now back to getting the scriptreader in

NESDIS

  • Got my account working!

GPT Agents

  • Meeting and working on probes

Phil 2.1.21

Book

  • Working on conclusions

GPT-2

  • Sent Antonio a note on research directions

GOES

  • Verify yaw flip works, and then add script reading

Phil 1.29.21

Looks like snow is coming!

Book:

  • More conclusions.
  • 2:00 Meeting with Michelle

GOES

  • 11:00 meeting with Vadim
  • I got to pull out my Game Programming Gems #2 for something that wasn’t really on the internet!
  • Status report

MDS

  • 10:00 meeting with Clay

GPT Agents

  • 3:30 Meeting

Phil 1.28.21

Called Jim Donnies. Best to drop off on Saturday between 9:00 – 12:00

Chocolate! – done!

Shopping

Checked by reality, some QAnon supporters seek a way out

  • More than a week after Donald Trump departed the White House, shattering their hopes that he would expose the worldwide cabal, some QAnon adherents have concocted ever more elaborate stories to keep their faith alive. But others like Smith are turning to therapy and online support groups to talk about the damage done when beliefs collide with reality.

How to Signal Trust in a Google Search

  • This study examined the problem of news distrust in a new way. We tested whether trust or distrust in a news outlet can be cued at the “domain level” — alongside search information about the news outlet — rather than beside or within a specific story. To do this, we used information about the news outlet — signals — that could be shown in a sidebar called a Knowledge Panel7 when people do a Google search for that news outlet. We conducted one experiment in the U.S. and Germany, and another in the U.S., Germany, and Brazil. We used multiple countries in order to determine how trust signals work across different cultures with varying levels of news trust.8 The Google News Initiative funded this research.

Impeachment and Deplatforming Aren’t Enough to Move Forward

  • In our forthcoming book, You Are Here, Ryan Milner and I explain the limitations of focusing on obvious sources of harm within the information ecosystem. To do this, we use biomass pyramids—diagrams that visually represent the cumulative weight and population size of various organisms within the same ecosystem. Apex predators—the lions, tigers, and bears—are at the very top of the pyramid. There are fewer of them, and as a result their cumulative weight is often relatively low, but they pose outsized threats to the rest of the ecosystem. At the same time, they are utterly dependent on all the lower strata. Take away their prey, take away their prey’s prey, take away the ecological conditions that sustain the biomass, and the lions and tigers and bears don’t have a chance.

Book:

  • More conclusions. I am not going in the direction that I thought when I started writing.

GOES

  • Verifying that everything still works after checking out Vadim’s changes. Nope, things are very wrong.
  • I think I fixed the physics step by making it the calculated time, but the ReactionWheelController needs to be updated to take the scaled_dclock values
  • Update TopController – not yet, I guess
  • 2:00 customer meeting

MDS

  • Ok, back to figuring out intersections while I wait for Vadim

Phil 1.27.21

Ok, so we’re still in the middle of a pandemic, it looks like around 400 people might be charged with sedition for the Jan 6 attack on the Capitol, and the GOP doesn’t want to have anything to do with impeachment, but the crazy seems to be much lower.

Vaccines are rolling out, and we’ll probably be at closer to 200M vaccinations by April. The government appears to be doing things again. It feels like we are past the turning point in whatever metaphor you’d like to use here, and now entering the mopping up phase.

The Fairness Doctrine sounds a lot better than it actually was

  • What America needs instead is a creative, comprehensive effort by both the private sector and the government to disincentivize conspiracies and misinformation on the many platforms on which they flourish. Some social media companies have begun this work, clearing out QAnon sites and banning some far-right and White power users and communities who pose a threat. That work needs to continue, with careful attention to the biggest offenders who game algorithms and media structures to spread misinformation. But sources of misinformation also need to be demonetized, whether they are YouTube channels or national cable networks, and algorithms tweaked to slow down the spread of extreme content.

Book

  • More conclusions. I’m tempted to look at antisemitism through the dominance hierarchy lens. To show that if you truly feel that you are at the top of the hierarchy (just under your God), and use the Bible as the scaffolding for that understanding, then the fact that the same bible says that the Jews are chosen by God must really require some mental gymnastics. And for some, those gymnastics lead to genocide, since once the Jews are gone, then there is no one between you and God, and the rightful order is restored.
  • I think this is related to slavery and racism in the US. To be able to treat people as property requires that they be less than people, otherwise dominance makes no sense. The hierarchy of God -> slave holder -> slave justifies any dominance behavior. And in the end the mental gymnastics that are required for that are the same as those required by anti-Semites. These are both co-created fictional social realities that exist in the communities that believe themselves righteously superior to those they oppress. And as we now understand, that process of creating hierarchical social realities can often stampede to the worst possible places.

NOAA

  • SEVEN trainings by Feb 1! But I can’t get to my email, so that’s going nowhere

GOES

  • 11:00 Meeting with Vadim. He has most of the changes in. Waiting for some final work and then I’ll start rolling in the new TopController

JuryRoom

  • 5:00 Meeting

Phil 1.26.21

This is looking a lot like maps!

Paper at https://journals.aps.org/prx/abstract/10.1103/PhysRevX.11.011012

Book

  • Starting conclusions. Going at it from explore/exploit, and why it’s important to have primordial movement, which is the domain of youth

IRAD

  • Talked to Aaron about using the time for a state switch as part of the input vector, or alternatively doing something like a waterfall CNN-style input to account for time. Didn’t seem to get much traction.

GOES

  • 10:00 Meeting with Vadim. I think we worked through the issues. 11:00 meeting tomorrow to discuss
  • Something’s wrong with ScriptReaderScratch – fixing

GPT Agents

JuryRoom

  • 5:30 HW review. Suggested that we needed a usability study for the experimenter’s as well