Visitor by the balcony.
| Website | https://logarithmic.net/pfh |
| X (deprecated) | https://twitter.com/paulfharrison |
| Bluesky (maybe?) | https://bsky.app/profile/paulfharrison.bsky.social |
Paul Harrison
- 166 Followers
- 73 Following
- 212 Posts
Bioinformatician with the Monash Genomics and Bioinformatics Platform, Monash University, Melbourne, Australia. I aim to control my false statement rate.
Happy little accidents...
I'm really liking this course on generative diffusion models. They seem to have boiled many years of confusing development of ideas down to a simple approach.
Second, sampling from the distribution with a Langevin Dynamics simulation. The algorithm is almost identical to gradient descent with momentum, but we add just the right amount of noise to the momentum at each step.
Comparison of optimization and sampling from a distribution defined by an energy function. I use a continuous version of the Ising model spin lattice energy.
First, optimization from a random initial state using gradient descent with momentum, using the SGD optimizer in PyTorch.
This week the Monash Genomics and Bioinformatics Platform did a bulk RNA-Seq workshop, covering end-to-end from experimental design, through library preparation, then using web tools to run an analysis pipeline and dig into differential expression.
I made a short video of the strange things UMAP and t-SNE can do to your data. The algorithms are shown mostly working as intended, yet with some surprising consequences.
What will UMAP do? What will t-SNE do?
pip install langevitour
Interactively tour your high-dimensional data directly from your Jupyter notebooks. Huge thanks to Wytamma Wirth @[email protected] for contributing the Python wrapper for langevitour.
https://colab.research.google.com/github/pfh/langevitour/blob/main/py/examples/langevitour.ipynb
