Chung-chieh Shan

Using commit 131e0cf370272e7f85fc7dbd1a678c1f3566a6dd, I was able to track down the necessary preconditions to add to `lda.hk` to coax the trivial products into reducing (https://github.com/hakaru-dev/tcp/commit/f09704390995c0986f812c904d38b45a644b1704). So this issue is less urgent...

Well, the trivial product issue is not completely fixed, merely worked around by adding `if w[wordUpdate] < size(word_prior) && doc[wordUpdate] < numDocs` to the input `lda.hk`. Those added conditions are...

> I don't understand what it means for the KB to discharge a condition. Please understand "discharge" as "entail". > From what I do understand, we want to simplify `product(piecewise(i=K,f(i),1),i=A..B)`...

Indeed the presence of `C` should not prevent the simplification. In the worst case we should get `piecewise(And(K

(Similarly, replacing the first argument to `simplify_assuming` above with `ceil(idx(w,0))` or `floor(idx(w,0))` should produce just `idx(w,0)`.)

I just confirmed that `naive_bayes_gibbs.hk` is now simplified better as desired. (For future reference, here's how I confirmed it: ``` git clone github.com:hakaru-dev/tcp.git cd tcp make -B src/NaiveBayes/naive_bayes_simp.hk git diff...

Thanks! Regarding the burglary example, I would recommend (1) documenting the optional use of simplification (rather than sampling) after disintegration to figure an exact (rather than approximate) answer to the...

Is this done?

That output is correct!

Let's simplify the description above by letting `X` contain just one hole (filled initially with `kh == docUpdate`) and getting rid of `Y` by merging it into `Phi` itself. So...