Remove optimizers (#103)

* Changes..

* tests/optimization now passes

* Remove _optimizer.py

* Fix config class for early stoppable recommenders.

* max_epoch -> train_epochs for recommender's parameter.

* Fix tests

* Mv optimizer under recommenders/

* fixing examples

* Improved example, fix import failure

Author: tohtsky

.gitignore

@@ -27,3 +27,4 @@ pybind11
 docs/build/
 docs/source/api_reference/
 .cache/
+.coverage

Readme.md

@@ -82,7 +82,7 @@ recommender.learn()
 recommender.get_score_remove_seen([0])
 ```
 
-## Step 2. Evaluate on a validation set
+## Step 2. Evaluation on a validation set
 
 To evaluate the performance of a recommenderm we have to split the dataset to train and validation sets:
 
@@ -120,15 +120,12 @@ This will print something like
 }
 ```
 
-## Step 3. Optimize the Hyperparameter
+## Step 3. Hyperparameter optimization
 
-Now that we can evaluate the recommenders' performance against the validation set, we can use [optuna](https://github.com/optuna/optuna)-backed hyperparameter optimizer.
+Now that we can evaluate the recommenders' performance against the validation set, we can use [optuna](https://github.com/optuna/optuna)-backed hyperparameter optimization.
 
 ```Python
-from irspack import IALSOptimizer
-
-optimizer = IALSOptimizer(X_train, evaluator)
-best_params, trial_dfs  = optimizer.optimize(n_trials=20)
+best_params, trial_dfs  = IALSRecommender.tune(X_train, evaluator, n_trials=20)
 
 # maximal ndcg around 0.43 ~ 0.45
 trial_dfs["ndcg@10"].max()

docs/source/api_reference.rst

@@ -57,42 +57,6 @@ which requires ``jax``, ``jaxlib``, ``dm-haiku``, and ``optax``:
 
     MultVAERecommender
 
-
-.. currentmodule:: irspack.optimizers
-
-Optimizers
------------
-.. autosummary::
-    :toctree: api_reference
-    :nosignatures:
-
-    BaseOptimizer
-    TopPopOptimizer
-    IALSOptimizer
-    P3alphaOptimizer
-    RP3betaOptimizer
-    TruncatedSVDOptimizer
-    CosineKNNOptimizer
-    AsymmetricCosineKNNOptimizer
-    JaccardKNNOptimizer
-    TverskyIndexKNNOptimizer
-    CosineUserKNNOptimizer
-    AsymmetricCosineUserKNNOptimizer
-    SLIMOptimizer
-    DenseSLIMOptimizer
-    MultVAEOptimizer
-    get_optimizer_class
-
-
-Autopilot
----------
-.. autosummary::
-    :toctree: api_reference
-    :nosignatures:
-
-    autopilot
-
-
 .. currentmodule:: irspack.split
 
 Split Functions