latexify hmm_poisson_changepoint.ipynb

[TakshPanchal]

Description

Figure Number

Figure 29.6, 29.7 and 29.8

Figures

Figure 29.6

Before PR	After PR

Figure 29.7

Before PR	After PR

Figure 29.8

Before PR	After PR

Issue

#113

Steps

• Latexify

Checklist:

• Performed a self-review of the code
• Tested on Google Colab.

Potential problems/Important remarks

When I tried to add legends in fig 29.7, it was too cumbersome. Should I add a legend to fig 29.7 figures?

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[karm-patel]

Regarding the legend in Fig 29.7

1. In the case of the same legend in every plot, we usually add a legend in only one plot, the ideally first plot should have a legend, but if the first plot doesn't have enough space for the legend, then we can choose any plot which has more space.
2. In case if legend's boundary is overlapping with plot you can use ax.legend(..., frameon=False)

[karm-patel]

Fig 29.6

1. You can put the title of the 2nd subplot in subcaption like (b) Inferred
2. Because we need the same ylim for both the subplots, the first comment would resolve this issue.

[karm-patel]

@TakshPanchal , Other things look good to me, you can create a PR in the main repo after resolving the above comments.

[TakshPanchal]

Removing the frame still didn't work out so I set fontsize to xx-small

[TakshPanchal]

Fig 29.6

[TakshPanchal]

• You can put the title of the 2nd subplot in subcaption like (b) Inferred

It will be the change in the latex file of the book right? 🤔

[karm-patel]

• You can put the title of the 2nd subplot in subcaption like (b) Inferred

It will be the change in the latex file of the book right? thinking

Yes right, I'll do that. your SS will remind me to edit latex :)

[karm-patel]

Removing the frame still didn't work out so I set fontsize to xx-small

Try fontsize in numbers always, so we can get an idea, start with 9 and then decrease, ideally 7 should be minimum
Here you can apply some hacks, like add "\n" - plt.plot(..., label="inferred\nrate")

[karm-patel]

And I prefer to add a legend in the first subplot, otherwise, it would not quickly visible.

[TakshPanchal]

Okay. understood.

[TakshPanchal]

How's this looking?

[karm-patel]

How's this looking?

Ummm, looks a little crowdy,
Maybe add a legend in 4th plot. Because 4th plot is also the first plot. And please tell me fontsize also

[karm-patel]

One more thing, Now I think you don't need to use reverse style legend labels, right?

[karm-patel]

@TakshPanchal
try out this comment also :)

[karm-patel]

Great @TakshPanchal , Go ahead with PR in the main repo.

[TakshPanchal]

Done :) and fontsize is 7

[karm-patel]

@TakshPanchal Wait,

Do you know the reason why this low spike is there? Previously it was not there

[karm-patel]

Line #1.    rng_key, rng_normal = jax.random.split(rng_key)

may be you can try changing rng_key to reproduce the results.

Reply via ReviewNB

[karm-patel]

@TakshPanchal, don't spend much time to reproduce 2nd plot, just try some rng_keys, otherwise, you can ask this to Dr. Kevin while creating PR, who knows this spike is obvious.

[TakshPanchal]

Okay got it.

[TakshPanchal]

I tried some values of seeds and rng_keys but no change in spike.