Alan Chang

Machine learning + systems biology. Into Bayesian methods, probabilistic models, deep learning, topology. Impulse book buyer. Has too many PDFs to read.

[Alan Chang @ChangAlanL]

@AnnaGHughes There are cellular membrane proteins whose main function is to move phospholipids to different sides of the bilayer. Two of these proteins are called "flippases" and "floppases".

[Alan Chang @ChangAlanL]

@KyleHaxWhy @elchefe @codeFreedomRitr Hugging at max velocity + tactical cuddling

[Alan Chang @ChangAlanL]

@Caroline_Bartma Would be interested!

[Alan Chang @ChangAlanL]

@JoshuaGrubbsPhD I'll join if available

[Alan Chang @ChangAlanL]

@VanAllenLab Although the work is potentially one of the greatest contributions to the field since the seminal work of O'Doyle T, et al. (1918), reviewer #2 was concerned by the "lack of previous publications at the highest [playoff] level."

[Alan Chang @ChangAlanL]

@ChelseaParlett "it's not trivial to me, damn it!" - me, staring at Eqns (14-16) and Corollary 7.

[Alan Chang @ChangAlanL]

@Corey_Yanofsky (2) [Question on my mind I have no answer to]. In high-dimensional 1K, 5K, 10K feature data, is there no hope for prob. models or causal models? Is it (A) doomed for difficult inference, (B) no performance gains or (C) the right Q has not been effectively formulated / explored?

[Alan Chang @ChangAlanL]

@Corey_Yanofsky (1) In -omics, knowledge graphs, single cell, and CyTOF we have a what could be a graph representation with many examples, but noisy labels of what the nodes or relations should be. It's an interesting case for self-supervised learning for feature representation and other uses.

[Alan Chang @ChangAlanL]

@frequencies https://t.co/c9K8dMcDI3 is pretty good all-around Python, I would probably skip some of the less relevant exercises. If you need to data analysis something like https://t.co/nGHCPCN5rz. For lots of ML, I like Sebastian Raschka's Machine Learning with PyTorch and Scikit-Learn

[Alan Chang @ChangAlanL]

Journal Commentary: Although the long-held paradigm is that there is no evidence that Bunny plays the role of a rider (Barnes, et al. 2007), new experimental results show that, contrary to previous understanding, Bunny does indeed have rider properties (Polachek, et al. 2021).

[Alan Chang @ChangAlanL]

@frequencies Oh and a deeper cut anime from the same personal era - Wolf's Rain! Very cool world building...much less interesting story.

[Alan Chang @ChangAlanL]

@frequencies I think I rented those DVDs from Albertson's (a truly early 2000s sentence) or maybe from the library. I thought it was cool, but was also watching Trigun, Cowboy Bebop, and Ghost in the Shell: SAC around this time. Maybe worth a rewatch of all of these series?

[Alan Chang @ChangAlanL]

@DynamicWebPaige @code Both, but I've had some issues with my setup for pair programming using VSCode Remote Development and Live Share with Jupyter notebooks. Getting kernels and the shared interactive window to work as expected has been a pain point for some time, but super happy to chat about it!

[Alan Chang @ChangAlanL]

I realize this doesn't follow the typical naming convention for R releases, but I would have made an exception for R version 3.6.9 "Damn You Fine".

[Alan Chang @ChangAlanL]

Looking forward to binge-watching it all!

[Machine Learning Dept. at Carnegie Mellon @mldcmu]

You can access NeurIPS live videos, even without registering! Only live interaction requires registration → https://t.co/zJhjs05UZo

[Alan Chang @ChangAlanL]

A trove of interesting papers! I also highly recommend @Pseudomanifold's workshop lectures at https://t.co/VoULp4ZNYx for an effective introduction/background.

[Bastian Grossenbacher-Rieck @Pseudomanifold]

Revisions are done, posters are ready!

The #NeurIPS2020 Workshop on Topological Data Analysis & Beyond is happy to present its accepted submissions @ https://t.co/h43cV2ZraC.

Congrats to all authors and big kudos to the reviewers!

#Topology + #MachineLearning = 🔴 = ❤️

[Alan Chang @ChangAlanL]

@Caroline_Bartma Indignant about it?