Monthly Archives: February 2018

Phil 2.11.18

Introduction to Learning to Trade with Reinforcement Learning

  • In this post, I’m going to argue that training Reinforcement Learning agents to trade in the financial (and cryptocurrency) markets can be an extremely interesting research problem. I believe that it has not received enough attention from the research community but has the potential to push the state-of-the art of many related fields. It is quite similar to training agents for multiplayer games such as DotA, and many of the same research problems carry over. Knowing virtually nothing about trading, I have spent the past few months working on a project in this field.
  • This sounds to me like reinforcement learning figuring out game theory. Might be useful for NOAA as well

Worked on getting the MapBuilder app into a useful standalone app: 2018-02-11 (1)

2.9.18

7:00 – 5:00 ASRC MKT

  • Add something about a population of ants – done
  • Add loaders for the three populations, and then one for trajectories
    • Promoted WeightWidget to JavaUtils
    • Moving 3d and UI building out of start
    • Ugh, new IntelliJ
    • Made the graph pieces selectable
    • Got drawmode (LINE) working
    • Reading in trajectories
    • Need to load each as a child and then draw all of them first, then make that selectable. Done!
  • Go over draft with Aaron. Hand off for rewrite 1? Nope – family emergency
  • 2:00 meeting with Aaron and IC team? Nope
  • Intro to deep learning course from MIT: introtodeeplearning.com
    • An introductory course on deep learning methods with applications to machine translation, image recognition, game playing, image generation and more. A collaborative course incorporating labs in TensorFlow and peer brainstorming along with lectures. Course concludes with project proposals with feedback from staff and panel of industry sponsors.
  • Topics, Events, Stories in Social Media
    • This thesis focuses on developing methods for social media analysis. Specifically, five directions are proposed here: 1) semi-supervised detection for targeted-domain events, 2) topical interaction study among multiple datasets, 3) discriminative learning about the identifications for common and distinctive topics, 4) epidemics modeling for flu forecasting with simulation via signals from social media data, 5) storyline generation for massive unorganized documents.
  • Communication by virus
    • The standard way to think about neurons is somewhat passive. Yes, they can exciteor inhibit the neurons they communicate with but, at the end of the day, they are passively relaying whatever information they contain. This is true not only in biologicalneurons but also in artificial neural networks. 

Phil 2.8.18

7:00 – 5:00 ASRC MKT

  • I need to put together an equation that describes group cohesion. Something like
    • C = Var(L) * Var(O) * Avg(V)/Var(V) * Si, for some population p, over a period of time t, where
      • C is group cohesion
      • Var(x) is the variance
      • Avg(x) is the average
      • L is location
      • O is orientation
      • V is velocity
      • Si is social influence, which is in turn a function of awareness and reach. In network terms, the range is from unconnected through partially connected to fully connected
    • C should have (at least?) three phases: Nomad (low), Flocking(mid), and Stampede(high). The intuition here is that the higher the velocity, the lower the variance has to be in location and orientation to obtain the same level of cohesion. A high velocity, tightly clustered group is a stampede, where social influence overrides environmental awareness.
  • The Shape of Art History in the Eyes of the Machine (Many mapping implications)
  • More writing
    • Citations are all in

Phil 2.7/18

7:30 – 5:30 ASRC MKT

  • Freezing rain and general ick, so I’m working from home. Thus leading to the inevitable updating of IntelliJ
  • Working on the 3D mapping app.
    • Reading in single spreadsheet with nomad graph info
    • Building a NodeInfo inner class to keep the nomad positions for the other populations
    • Working! 2018-02-07
    • Better: 2018-02-07 (2)
    • Resisting the urge to code more and getting back to the extended abstract. I also need to add a legend to the above pix.
  • Back to extended abstract
    • Added results and future work section
    • got all the pictures in
    • Currently at 3 pages plus. Not horrible.
  • Demographics and Dynamics of Mechanical Turk Workers
    • There are about 100K-200K unique workers on Amazon. On average, there are 2K-5K workers active on Amazon at any given time, which is equivalent to having 10K-25K full-time employees. On average, 50% of the worker population changes within 12-18 months. Workers exhibit widely different patterns of activity, with most workers being active only occasionally, and few workers being very active. Combining our results with the results from Hara et al, we see that MTurk has a yearly transaction volume of a few hundreds of millions of dollars.

Phil 2.6.18

7:30 – 5:00 ASRC MKT

  • Took four much needed days off on Sanibel island. Forgot to pack some things? Need to call the hotel at (239) 215-3401
  • Starting CI 2018 abstract. And oddly, the abstract isn’t showing??? Sent a note to the conference chair. IN the meantime, I have a subsection for the abstract. It appears to be acmlarge for the most part, so maybe use that????
  • Was going to get back to Angular, but stuck with 404s on CRUD operations: 404
  • Working on the 3D map application. Decided to go with JavaFX and their 3d implementation. It’s going quickly. MapApp1
  • I’ve also gotten the graph generator creating spreadsheets that the map app can read in. So the next job will be to wire everything together, where the position information is based off the nomad trajectories, with the size and visitor (height) data being overlayed with the different colors.

Phil 2.1.18

7:00 – 3:30 ASRC MKT

  • Communications Handbook for IPCC scientists
  • The Barnes-Hut Approximation
    • Efficient computation of N-body forces
      By: Jeffrey Heer
      Computers can serve as exciting tools for discovery, with which we can model and explore complex phenomena. For example, to test theories about the formation of the universe, we can perform simulations to predict how galaxies evolve. To do this, we could gather the estimated mass and location of stars and then model their gravitational interactions over time.
  • Need to get started on the extended abstract for Collective Intelligence 2018! One month! March 2, 2018!
    • Set up the LaTex template for the conference. Done
    • Think I want to call it Mapping Simon’s Anthill
  • Need to contact the CHIIR 2018 folks to see what is expected for the DC
  • More Angular, feeling my way through the Http code, which has been deprecated. Looked at the similar code in Tour of Heroes. We’ll see if the old stuff works and then try to update? Need to ask Jeremy.
  • Back to BIC. Evolutionary reasons for cooperation as group fitness, where group payoff is maximized. This makes the stag salient in stag hunt.
  • A thorough explanation of synchronization/phase locking. My mental model is this: Imaging a set of coaxial but randomly oscillating identical weights sliding back and forth in their section of lightweight tubing. From the outside, the tube would be stationary, as all the forces would be cancelling. If the weights can synchronize, then the lightweight tube will be doing most of the moving. Since the mass of the tube is lower than the mass of the combined weights,   The force required for the whole system will be lower, and as a result (I think?) the system will run more efficiently and longer. Need to work out the math.