probmods/dippl

The Design and Implementation of Probabilistic Programming Languages

dippl's Issues

Compile and minify webppl when updating submodule

See probmods/webppl#41

Expand section on caching

This section (and the one above it) is really just a stub: http://dippl.org/chapters/03-enumeration.html#caching

Would be good to say a bit more about how caching works, and when you can (and should) do it. Including inserting additional marginalization ops.

sampleWithFactor boxes broken

http://dippl.org/chapters/04-factorseq.html

Enumerate can only be used with ERPs that have support function.

Update 'Decomposing and interleaving factors' section

Make sure text and program results cohere

reading input/output stream

How do I read from the input/output stream (or a file)?

update-webppl script broken

Running the script results in:

Resetting webppl to latest remote...
fatal: ambiguous argument 'origin/master': unknown revision or path not in the working tree.
Use '--' to separate paths from revisions, like this:
'git [...] -- [...]'

Add progress bars for inference algorithms

Remaining dependencies between boxes

• Semantic parsing example
• Pragmatics example
• Final box on particle filter page
• PCFG on factor seq page

WebPPL boxes broken in Safari and Firefox

ReferenceError: Can't find variable: webppl

Add an EPUB version

Update to latest version of webppl

This requires some adjustment - the live cps and naming transform forms break if I just slot in the latest dev version.

"too much recursion" error in firefox

In Firefox only

In

• Early, incremental evidence after the line "Similarly the PCFG can be written as:"
• and after the line "Similarly for the PCFG:"
• and also in Semantic parsing, on the line print(literalListener("all blond people are nice"))

we get the error:

too much recursion

Make it easy to create publicly shared webppl pages from editor

Once webppl-editor and webppl-viz are stabilized, use them

The addressing transform

I was reading about the automatic addressing transform based on stack-addresses. It seams to make sense in most cases but I was wondering how general this scheme is. It seems to work when you recurse or iterate over a list and you are always going over the same modeled individuals in the same order. But imagine you were modelling a bus full of tourists, trying to predict some attributes of each tourist such as hunger level after each stop (to determine when to stop to eat).

Say the tourists like to trade seats and you decided to first predict the order of seating after each stop and then recurse over the bus seats to predict other attributes. When trying to reuse sample from the previous trace, wouldn't it make more sense for the sample to be tied to named individual tourists instead of to seats in the bus even if the stack address name scheme correspond to seat order not tourists order?

It might just be a matter of being careful when writing generative models to not recurse over different order of individual things. Otherwise allowing some kind of manual naming could also be an option.

Or am I interpreting this wrong?

Fix layout on mobile

Margins too wide

Blank canvas images in particular examples

The guassian mixture, hidden semi-markov model, and particle filter for semi-markov random walk examples render blank images in 05-particlefilter.html

Make introduction of factor notation more intuitive

In webppl issue #18, Tomer wrote:

I'm not sure what level of readership you're aiming at, but taking probmods.org as a benchmark, it feels like towards the end of Chapter 2 you're assuming a great deal of background knowledge for a casual reader.

In particular, while the introduction of inference through the binomial function feels 'standard' and normal, this paragraph suddenly shows up after it:

"What if we wanted to adjust the above binomial computation to favor executions in which a or b was true? The factor keyword re-weights an execution by adding the given number to the log-probability of that execution. For instance:"

What 'given number' are we adding here? How much does a change of -2 correspond to? Why would we want to make that specific inference or assumption? In Church/probmods when inference was introduction the conditioning was on a pretty simple notion that 'the sum of the heads is 2 or greater'.

Performance of setTimeout

We currently use setTimeout to clear the stack... this may be bad wrt performance. Should clear stack less often or switch to a trampoline to avoid stack bloat.

Clean up examples folder

Remove examples that are not part of the book. Migrate to webppl repository, forest, or specific research repositories as appropriate.

Fix histograms with too many bars on pragmatics page

• Pragmatics example