Liang Xu

As my research always involves a huge amount of computation, parallel technique is a super fancy way to save my time. But the resource of cluster at my university is limited. I recently received a warning that CIT has detected multiple-account usage under my IP. To avoid the risk of being fired, I have to figure out another way to parallelize my job instead of parallelizing accounts. Right on time, GPU programming came out as an ideal option. I hope the GPU resource of my university is not full loaded yet.

Due to some incompatibilities of CUDA and Visual Studio 2017, I spent half a day to figure out the solutions and finally succeeded building up GPU programming environment on Windows 10 with Pycuda installed in Python. Here, I post the procedure of the build and some solutions to the incompatibilities.

In this blog, the details of the arguments used in ggradar2 are provided.

ggradar2 is now available. A large amount of features have been added to make your radar chart powerful. See ggradar2 on my Github

Finally, I have done visualizing the evolution of SMC in R instead of in Python. Here, two illustrations are given.

Share a fancy website that shows “Curated color palette”.

A convenient way to update all the packages.

In the third project, a bunch of mega data are generated for analysis. However, the simulated data is too big to load in memory. Thus, I need to extract some information without opening the file. After quite a few time searching for a solution, sed command in bash language helps me out. More details about the third project is coming soon.

I found the following interpretation to partial least square regression is much better than mine. So I cited it as below:

“Partial least squares regression (PLS regression) is a statistical method that bears some relation to principal components regression; instead of finding hyperplanes of maximum variance between the response and independent variables, it finds a linear regression model by projecting the predicted variables and the observable variables to a new space. Because both the X and Y data are projected to new spaces, the PLS family of methods are known as bilinear factor models. Partial least squares Discriminant Analysis (PLS-DA) is a variant used when the Y is categorical.”

—from Wikipedia

Writing a periodic progress report is a good way to record the process that you gain knowledge. Reviewing my progress reports in the last 3 years, it recovers my memory about the knowledge and technologies that I almost forgot as I rarely use them. From this post on, I would like to give them a review and share on the blog. Hope it can help people who are interest in and one can help me if I am wrong at somewhere.

I guess this function is specially useful to our group in which we play with L table.

L table is an alternative way to a phylo class for phylogenetic information storage. The function L2phylo has been implemented in the DDD package that converts an L table to a phylo class. This function phylo2L does the conversion the other way around. Thus, if you want to apply your model to an empirical data. This may be useful to you.