Finally switched to Jekyll with Mathjax!

3 minute read

I finally decided to migrate various of my old writing in my faculty website at UC Irvine (deactivated as of Oct 2020) a static website generator Jekyll. Apparently Jekyll has a much more flexible and pleasant editing experience than WordPress, as I am using VSCode to do everything nowadays (teaching numpy, $\LaTeX$, MATLAB, and some markdown writings).

Example of Python snippets

The following function, which is an explicit implementation of 3-layer neural net backpropagation in numpy , was used in UC Irvine Math 10, an introductory machine learning class I designed circa 2019. We can use liquid syntax to write code block in Markdown, while enabling line numbers if we set the kramdown syntax highlighter engine to rouge.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
def backprop(W,b,X,y):
    '''
    W: weight, b: bias
    X: features, y: target
    Step 1: forward pass
    Step 2: backprop
    '''
    N = X.shape[0]
    # layer 1 = input layer
    a1 = X
    # layer 1 (input layer) -> layer 2 (hidden layer)
    z2 = np.matmul(X, W[0]) + b[0]
    # layer 2 activation
    a2 = relu(z2)
    # layer 2 (hidden layer) -> layer 3 (output layer)
    z3 = np.matmul(a2, W[1]) + b[1]
    # output layer activation
    output = z3
    
    # layer 3's derivative
    delta3 = -(y - output)
    # layer 2's derivative
    delta2 = np.matmul(delta3, W[1].T)*relu_prime(z2)
    # no derivative for layer 1
    
    dW = [np.matmul(a1.T,delta2)/N + eps*W[0], np.matmul(a2.T,delta3)/N + eps*W[1]]
    db = [np.mean(delta2, axis=0), np.mean(delta3, axis=0)]
    # dW[0] is W[0]'s derivative, and dW[1] is W[1]'s derivative; similar for db
    return dW, db

Mathjax with LaTeX

The following instruction follows a post at Brendan A R Sechter’s Development Blog with some simplifications if we are either mathematicians, or want to add MathJax to every post semi-default. First we need to add a renderer in certain style file in _layout, by adding the following code block to default.html; Alternatively if we are using Minimal Mistakes template, we wanna add this block to scripts.html under _includes folder.


<script type="text/javascript" async
  src="https://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-MML-AM_CHTML">
</script>


<script type="text/x-mathjax-config">
  MathJax.Hub.Config({
    extensions: ["tex2jax.js"],
    jax: ["input/TeX", "output/HTML-CSS"],
    tex2jax: {
      inlineMath: [ ['$','$'], ["\\(","\\)"] ],
      displayMath: [ ['$$','$$'], ["\\[","\\]"] ],
      processEscapes: true
    },
    "HTML-CSS": { availableFonts: ["TeX"] }
  });
</script> 

Then either $ $ or \\( \\) gives us inline math as the html above suggests. Notice the engine here directly rips from mathjax.org, I have seen bloggers complained that it does not support external rendering anymore(?), yet based on my experience it is not the case.

Now we can write some famous formulae like:

\[E = mc^2, \tag{The most famous formula?}\]

or multiline equation without the \begin{aligned}\end{aligned} or equivalence environment needed for a $\LaTeX$ compiler like pdflatex or xelatex,

\[\displaystyle \int_{\Omega} \nabla \boldsymbol{\phi} : \nabla\boldsymbol{\psi} = \int_{\Omega}\Big( (\nabla\times \boldsymbol{\phi}) \cdot (\nabla\times \boldsymbol{\psi}) + (\nabla \cdot \boldsymbol{\phi}) (\nabla \cdot \boldsymbol{\psi}) \Big) \\ \displaystyle \quad + \int_{\partial \Omega} \Big( \underbrace{\nabla_{\Gamma}(\boldsymbol{n}\cdot \boldsymbol{\phi}_n)\cdot \boldsymbol{\psi}_{\Gamma}}_{\text{tangential}} - \underbrace{(\nabla\cdot\boldsymbol{\phi}_{\Gamma})(\boldsymbol{n}\cdot\boldsymbol{\psi}_{n})}_{\text{normal}} \Big). \tag{$\ast$}\label{eq:multi}\]

The tag and a linkable equation reference, such as \eqref{eq:multi} can be set by adding these near the end of an equation.

\tag{tag_you_like}\label{label_you_like}

and then reference it using \eqref{} like we would use in a regular $\LaTeX$ document.

However, there are two caveats, in a multiline displaystyle, we want to use $$ $$ delimiters instead of \\[ \\], the alignment toggle & is not applicable either, so we have to format the equation beautifully in the first place.

Minimal Mistakes template

The template I am using is called Minimal mistakes, notice that the term “minimal” is different from “minimum”, while “minimum” indicates certain non-surpassable lowest limit, “minimal” is more metaphoric, and accords more with the minimalism in arts.

Minimal: characterized by the use of simple form or structures.

Comments