Open Rithwikksvr opened 4 years ago
@rishi-a Can you please put a tick mark for all the issues you addressed?
@nipunbatra sir, what do you mean by fat? I took a look at the article and didn't find anything like that. Can you please be more specific?
We will also think about how to make the resampling bit look cooler.
I apologize for asking these trivial questions.
The slides for pi was very wide for the amount of information it carried. The current version looks better.
[x] can you use the same theme as distill.pub? The text in your article seems much larger than the text in Apoorv's article.
[x] Ever wondered how part of .. should end with a question mark
[x] how about how is odd. Make it more direct - Or, how your smart watch counts the number of st..
[x] each of these are examples --> each of these are applications of time-series machine learning
[x] the text in conditional or depends .. this entire para needs to be rewritten to be clear
[x] You might say should be followed with " "
[x] Put the vector drawing of sunny and rainfall. For all your diagrams you should have vector drawing. Use Inkscape or any online vector drawing tool for this purpose. All the non-interactive graphical models can be made using this or using sigma.js. But, should look pretty.
[x] Thus effectively --> Thus, effectively. While I know that this is an initial draft, it is important to write well for forming good habits.
[x] predicting today's weather does not end with a "."
[x] The Markov Model starts with "As explained before.." but that relevant text is currently missing. You may also write that the MM is also called a memory less system.
[x] in the graphical model shown, please shade the observed nodes in light gray. Also, you may to redraw this using sigma.js or any vector tool. You should look at Figure 10.3 and Figure 10.4 from Murphy ML book as inspiration.
[x] There are many editorial mistakes in second order Markov chain too. Like, using X{t-1} instead of x{t-1} and others. Also, see Figure 10.3 (b) from Murphy on how to draw second order Markov chain.
[x] Interactive example of Markov Chain: looking much better than before. Can we chose a different colour for biased and fair? The current ones: red and pink are similar. May I suggest violet and red (biased)? (based on Vibgyor) None of these colours that you use for nodes (biased, fair, start state) should appear in the transition or pi matrix. Also, the start state colour is similar to Pi (fair). Pi looks very fat. It is fine to make it thin IMO. Similarly, can we try out some other prettier ways to have resampling?
[x] Draw the diagram for HMM following above comments.
[x] Convention: make sure to use "t" from 1 to T everywhere for convention and do not use "n"
[x] Use the same colours for HMM A and B as used for MM A and B
[ ] Creating a text generation example for markov chain
[x] Highlighting Psuedo-code corresponding to sampling