original file link

• Meeting Date: 29.06.2018
• Attendee: Chanoh, Won, Mike, Danial

Paper reviewed

• Paper Title: Hybrid computing using a neural network with dynamic external memory

• Reason: some of the recent deep SLAMs are utilizing differentiable neural computer(DNC). This is the original paper on DNC.

• Useful materials: ​https://greydanus.github.io/2017/02/27/differentiable-memory-and-the-brain/​

• Code(pytorch and tensorflow) and videos are available.
  • Videos
    • https://www.youtube.com/watch?v=6SmN8dt8WeQ
    • https://www.youtube.com/watch?v=otRoAQtc5Dk
    • https://www.youtube.com/watch?v=_H0i0IhEO2g
    • https://www.youtube.com/watch?v=r5XKzjTFCZQ
    • https://www.youtube.com/watch?v=K14VNejrgmc
  • Github
    • https://github.com/ixaxaar/pytorch-dnc
    • https://github.com/wills2/tf-DNC
    • https://github.com/bgavran/DNC
    • https://github.com/claymcleod/tf-differentiable-neural-computer
  • Block Diagram
    • https://github.com/Mostafa-Samir/DNC-tensorflow/blob/master/docs/data-flow.md
    • https://github.com/Mostafa-Samir/DNC-tensorflow/tree/master/docs
• It’s application in SLAM.
  • https://arxiv.org/abs/1702.08360​
  • https://arxiv.org/abs/1706.09520
• Related video to this.
  • https://vimeo.com/252185932

Meeting summary

Had discussion on DNC memory read, wright operation.

Next Meeting

• Meeting Date: 11.07.2018
• Paper: Hybrid computing using a neural network with dynamic external memory