| 1943____ | |
| ### ## #### ### # # | |
| ## ## #### ### # # | |
| ## ## #### ### # # | |
| ## ## ## #### #### ### # # | |
| ## ## ## ## #### ### # # | |
| ## ## ### #### #### ### # # | |
| ### ## ## ## #### ### # # | |
| ## ## #### ### # # |
| # Install plug.vim in ~/.vim/autoload | |
| mkdir -p ~/.vim/autoload | |
| wget https://raw.githubusercontent.com/junegunn/vim-plug/master/plug.vim -O ~/.vim/autoload/plug.vim | |
| # Add plugins to ~/.vim/vimrc | |
| echo -e " | |
| call plug#begin() | |
| Plug 'luiarthur/tmux.vim' | |
| Plug 'luiarthur/red.vim' | |
| call plug#end() |
| #!/bin/bash | |
| # GSOC instance. | |
| aws_instance_id=i-01234567d # dummy id. | |
| # Start a stopped aws ec2 instance. | |
| alias start_aws_instance="aws ec2 start-instances --instance-ids $gsoc_id" | |
| # Stop a running aws ec2 instance. | |
| alias stpo_aws_instance="aws ec2 stop-instances --instance-ids $gsoc_id" |
| # See: https://rufflewind.com/2016-12-30/reverse-mode-automatic-differentiation | |
| import numpy as np | |
| # NOTE: Reverse AD is generally favored in statistics / ML because we usually | |
| # optimize a likelihood / negative cost function, which is a scalar, in which | |
| # case reverse AD requires only one evaluation of the objective, which could | |
| # be expensive. Forward AD, while simpler to implement, requires as many | |
| # evaluations of the objective as the number of parameters. So, it tends to be | |
| # slower than Reverse AD when the number of parameters is > 10 or the objective | |
| # is costly to compute.. |
| # Number of cores to use. | |
| NCORES=3 | |
| # Use parenthesis to suppress superfluous output. | |
| ( | |
| for x in `seq 8` | |
| do | |
| ((i%NCORES)) && ((i++==0)) && wait; | |
| echo $x && sleep 2 & # comomand to execute. | |
| done |
| """ | |
| Basic progress bar. | |
| """ | |
| from datetime import timedelta, datetime | |
| import time | |
| def pbrange(*args, **kwargs): | |
| return pbar(range(*args), **kwargs) |
| import numpy as np | |
| def gaussian_kernel(X, knots, sd): | |
| """ | |
| X: points to evaluate kernel (n x q) | |
| knots: kernel locations (m x q) | |
| sd: kernel width | |
| """ | |
| diff = X[..., None] - knots.T[None, ...] # n x q x m | |
| ss = np.sum(diff ** 2, axis=1) # n x m |
url field.nbviewer.javascript: (() => {
const url = window.location.href;
const nbviewer_url = "https://nbviewer.org/github/";
const new_url = nbviewer_url + url.split("https://github.com/")[1];
window.open(new_url);
| import numpy as np | |
| def correlation_from_covariance(covariance): | |
| v = np.sqrt(np.diag(covariance)) | |
| outer_v = np.outer(v, v) | |
| correlation = covariance / outer_v | |
| correlation[covariance == 0] = 0 | |
| return correlation |