DQD Emitter batch_size description (#336)

## Description

<!-- Provide a brief description of the PR's purpose here. -->
We originally had batch_size for DQD emitter include the solution
returned by ask_dqd, i.e., if you want ask to return 36 solutions you
need to put batch_size=37. We reverted this decision and now assume that
ask_dqd only returns one solution, i.e., you will put batch_size=36 for
the above example. This make the batch_size description between DQD and
non-DQD emitters consistent.

However, at some point we must have forgot to change the docstrings
according. This PR address that.

**Furthermore, I find that some configuration in sphere.py is not
consistent with the original paper.** For example, the [CMA-MAE
paper](https://arxiv.org/pdf/2205.10752.pdf) "select[s] a batch size
$\lambda = 36$ following prior work" (Appendix A), while [this
line](https://github.com/icaros-usc/pyribs/blob/7099f1ab7679d557f41a9608b9cdaaf075d30981/examples/sphere.py#L490C1-L490C26)
clearly uses $\lambda=37$. Similarly, our sphere experiment with
CMA-MAEGA also uses $\lambda=37$.

I could be missing something here, so I figured I should try to clarify
in this PR in case other users have same confusion as me.

## TODO

<!-- Notable points that this PR has either accomplished or will
accomplish. -->

## Questions

<!-- Any concerns or points of confusion? -->

## Status

- [x] I have read the guidelines in
[CONTRIBUTING.md](https://github.com/icaros-usc/pyribs/blob/master/CONTRIBUTING.md)
- [x] I have formatted my code using `yapf`
- [x] I have tested my code by running `pytest`
- [x] I have linted my code with `pylint`
- [ ] ~I have added a one-line description of my change to the changelog
in `HISTORY.md`~
- [x] This PR is ready to go

examples/sphere.py

@@ -43,17 +43,9 @@
 - `cma_maega`: GridArchive (learning_rate = 0.01) with
   GradientArborescenceEmitter using ImprovementRanker.
 
-All algorithms use 15 emitters, each with a batch size of 37. Each one runs for
-4500 iterations for a total of 15 * 37 * 4500 ~= 2.5M evaluations.
-
-Notes:
-- `cma_mega` and `cma_mega_adam` use only one emitter and run for 10,000
-  iterations. This is to be consistent with the paper (`Fontaine 2021
-  <https://arxiv.org/abs/2106.03894>`_) in which these algorithms were proposed.
-- `cma_mae` and `cma_maega` run for 10,000 iterations as well.
-- CVTArchive in this example uses 10,000 cells, as opposed to the 250,000
-  (500x500) in the GridArchive, so it is not fair to directly compare
-  `cvt_map_elites` and `line_cvt_map_elites` to the other algorithms.
+The parameters for each algorithm are stored in CONFIG. The parameters
+reproduce the experiments presented in the paper in which each algorithm is
+introduced.
 
 Outputs are saved in the `sphere_output/` directory by default. The archive is
 saved as a CSV named `{algorithm}_{dim}_archive.csv`, while snapshots of the
@@ -431,7 +423,7 @@
         "archive_dims": (100, 100),
         "use_result_archive": False,
         "is_dqd": True,
-        "batch_size": 36,
+        "batch_size": 35,
         "archive": {
             "class": GridArchive,
             "kwargs": {
@@ -459,7 +451,7 @@
         "archive_dims": (100, 100),
         "use_result_archive": False,
         "is_dqd": True,
-        "batch_size": 36,
+        "batch_size": 35,
         "archive": {
             "class": GridArchive,
             "kwargs": {
@@ -487,7 +479,7 @@
         "archive_dims": (100, 100),
         "use_result_archive": True,
         "is_dqd": False,
-        "batch_size": 37,
+        "batch_size": 36,
         "archive": {
             "class": GridArchive,
             "kwargs": {
@@ -516,7 +508,7 @@
         "archive_dims": (100, 100),
         "use_result_archive": True,
         "is_dqd": True,
-        "batch_size": 37,
+        "batch_size": 35,
         "archive": {
             "class": GridArchive,
             "kwargs": {

ribs/emitters/_gradient_arborescence_emitter.py

@@ -9,7 +9,7 @@
 
 class GradientArborescenceEmitter(EmitterBase):
     """Generates solutions with a gradient arborescence, with coefficients
-    parameterized by an ES.
+    parameterized by an evolution strategy.
 
     This emitter originates in `Fontaine 2021
     <https://arxiv.org/abs/2106.03894>`_. It leverages the gradient information
@@ -100,10 +100,9 @@ class GradientArborescenceEmitter(EmitterBase):
             an open problem. Hence, this argument must be set to None.
         batch_size (int): Number of solutions to return in :meth:`ask`. If not
             passed in, a batch size will be automatically calculated using the
-            default CMA-ES rules. Note that `batch_size` **does not** include
-            the number of solutions returned by :meth:`ask_dqd`, but also note
-            that :meth:`ask_dqd` always returns one solution, i.e. the solution
-            point.
+            default CMA-ES rules. This **does not** account for the **one**
+            solution returned by :meth:`ask_dqd`, which is the solution point
+            maintained by the gradient optimizer.
         epsilon (float): For numerical stability, we add a small epsilon when
             normalizing gradients in :meth:`tell_dqd` -- refer to the
             implementation `here
@@ -252,12 +251,12 @@ def ask(self):
         The multivariate Gaussian is parameterized by the evolution strategy
         optimizer ``self._opt``.
 
-        Note that this method returns `batch_size - 1` solution as one solution
-        is returned via ask_dqd.
+        This method returns ``batch_size`` solutions, even though one solution
+        is returned via ``ask_dqd``.
 
         Returns:
-            (batch_size, :attr:`solution_dim`) array -- a batch of new solutions
-            to evaluate.
+            (:attr:`batch_size`, :attr:`solution_dim`) array -- a batch of new
+            solutions to evaluate.
         """
         coeff_lower_bounds = np.full(
             self._num_coefficients,