[Update] AUSRT GUI; MNIST Data for Constructive Algorithms; Conferences and Competitions
Summary
There isn’t much to report from last week. I have received a new hard drive for creating local data backups; I’ll be spending some time setting backups this week.
I’m ready to start working on importing some charts to a new GUI for the Adelaide University Solar Race Team. The next steps are: 1) getting the GUI to display plots, 2) getting the plots to animate/update, 3) getting simulated data from ROS to the GUI, and 4) displaying the simulated data in the GUI. If I get this working soon, I’ll have a go at replacing the telemetry communication and logging in ROS with Python libraries.
Testing my constructive algorithm code stalled momentarily while I was modifying scripts for downloading and unzipping the MNIST data set. Now I’m working through some errors in the training and dynamic node creation.
Last week I listed conference paper tasks: 1) revisiting the literature survey, 2) writing an outline of the code development plan, 3) developing the theory and learning algorithms, and 4) implementation and testing. I still have more work to do on the literature survey and algorithm outlines for my proposed conference papers. I’m now aiming to get these near completion this week.
I still have some more work to do to finish setting up Malmo on my desktop. I’ll pencil this in for some more attention this week too.
I might also have some more work on the robot arm art display coming in the next week or two.
See my Wordpress site (https://tobylightheart.wordpress.com) for updates on my work on learning techniques for people and other topics of writing.
Updated Project and Task List
- AI Collaborative Network - Data Collaboration
- NfP meeting (unable to attend). [2019-07-09, Tuesday]
- Adelaide University Solar Race Team
- Create GUI skeleton in Qt Designer. [Complete]
- Add plots to GUI. [Aim: 2019-07-04, Thursday]
- Display simulated telemetry.
- Add plots of projected battery charge and usage.
- Create plan for code development and unit and integration tests.
- Literature and Study
- Review literature on meta-learning and constructive neural networks. [Due weekly: Wednesday]
- Constructive Neural Networks
- Test Dynamic Node Creation algorithm using NumPy. [Aim: 2019-07-04, Thursday]
- Add automatic neuron construction from error detection to PyTorch Blitz code.
- Selective Neural Networks
- Theory development: network structures to perform neuron selection rather than detection.
- PyTorch methods for implementing neuron or module selection.
- Conference paper: Reinforcement learning with neural network construction for visual control
- Initial literature survey [Aim: 2019-07-03, Wednesday]
- Theory development plan
- Code development plan
- Submission [Deadline: 2019-09-23]
- Conference paper: Self-supervised motion control learning with reinforcement-based target selection
- Initial literature survey [Aim: 2019-07-03, Wednesday]
- Theory development plan
- Code development plan
- Submission [Deadline: 2019-09-23]
- NeurIPS competitions (https://nips.cc/Conferences/2019/CallForCompetitions)
- Install and test Malmo (https://www.aicrowd.com/challenges/neurips-2019-minerl-competition) [Aim: 2019-06-03, Wednesday]
- Register for MineRL Competition
- MineRL first round submissions close [2019-09-22]
Tasks This Week
- Survey recent literature on constructive neural networks and reinforcement learning for physics simulations/robotics. [Due: 2019-07-03, Wednesday]
- Conference paper: Reinforcement learning with neural network construction for visual control
- Conference paper: Self-supervised motion control learning with reinforcement-based target selection
- Install and test Malmo [Aim: 2019-07-03, Wednesday]
- Test/Debug Dynamic Node Creation code (NumPy). [Aim: 2019-07-04, Thursday]
- Adelaide University Solar Race Team
- Add plots to new AUSRT GUI. [Aim: 2019-07-04, Thursday]
- Add plots from simulated telemetry [Aim: 2019-07-04, Thursday]