I’m having difficulty determining whether it(src/tree/updater_quantile_hist.cc and so on) operates under tree_method='approx' or tree_method='hist'.
This has made it challenging to analyze the exact algorithm in detail when tree_method='hist' is set, so I’m posting my questions here in the hope of getting some clarification.

I would really appreciate any insights on the following:

1. Histogram Bin Splitting Criterion:

   • When creating the histogram, are the bins divided based on the number of datapoints or the sum of Hessian values?

2. Candidate Split Point Calculation:

   • When determining candidate split points, is the value set as:
     • The average of the maximum value of bin1 and the minimum value of bin2?
     • The maximum value of bin1?
     • The minimum value of bin2?

3. Handling Uneven Bin Sizes:

• If the answer to question 1 is that bins are split based on the number of datapoints, how is binning handled when the datapoints do not divide evenly? For example, if max_bin = 4 and there are 10 datapoints, how would they be distributed?

I would greatly appreciate any insights or guidance on this. Thank you in advance for your help!