This is just a quick list of resourses on TDA that I put together for @rickasaurus after he was asking for links to papers, books, etc on Twitter and is by no means an exhaustive list.
Both Carlsson's and Ghrist's survey papers offer a very good introduction to the subject
Mapper algorithm.
| # BEFORE STARTING PUT libcudnn5_5.1.10-1+cuda8.0_amd64.deb AND libcudnn5-dev_5.1.10-1+cuda8.0_amd64.deb IN /tmp | |
| # OTHERWISE THIS WON'T WORK | |
| # This has been tested only on EC2 P2 xlarge instance with 16 GB storage and stock Ubuntu 16.04 | |
| # It's a setup for a playground EC2 machine to perform workshops with Jupyter Notebook on GPU. | |
| # It might contain some unnecessary crap | |
| # The process takes like 15 minutes | |
| pushd /tmp && | |
| stat libcudnn5_5.1.10-1+cuda8.0_amd64.deb && | |
| stat libcudnn5-dev_5.1.10-1+cuda8.0_amd64.deb && |
| from math import sqrt | |
| def put_kernels_on_grid (kernel, pad = 1): | |
| '''Visualize conv. features as an image (mostly for the 1st layer). | |
| Place kernel into a grid, with some paddings between adjacent filters. | |
| Args: | |
| kernel: tensor of shape [Y, X, NumChannels, NumKernels] | |
| (grid_Y, grid_X): shape of the grid. Require: NumKernels == grid_Y * grid_X |