PR fixes (v1)

aimclub · Apr 4, 2023 · 345ff5c · 345ff5c
1 parent 0f27918
commit 345ff5c
Show file tree

Hide file tree

Showing 9 changed files with 38 additions and 28 deletions.
diff --git a/docs/source/basics/concepts.rst b/docs/source/basics/concepts.rst
@@ -1,7 +1,7 @@
 Main Concepts
 =============
 
-The main concepts of the framework are the follows:
+The main framework concepts are as follows::
 
 - **Flexibility.** FEDOT can be used to automate the construction of solutions for various problems, data types (texts, images, tables), and models;
 - **Integrability.** FEDOT supports widely used ML libraries (Scikit-learn, CatBoost, XGBoost, etc.) and allows you to integrate custom ones;

diff --git a/docs/source/introduction/fedot_features/automation/data_nature.rst b/docs/source/introduction/fedot_features/automation/data_nature.rst
@@ -4,8 +4,10 @@ Data nature
 FEDOT uses specific data processing according to the source
 of the input data (whether it is pandas DataFrame, or numpy array, or just a path to the dataset, etc).
 
-..
-    Стоит ли тут написать, что datetime будет конвертиться в число?
+.. note::
+
+    Be careful with datetime features, as they are to be casted into a float type with milliseconds unit.
+
 
 Apart from that FEDOT is capable of working with multi-modal data.
 It means that you can pass it different types of datasets

diff --git a/docs/source/introduction/fedot_features/automation/dimensionality_reduction.rst b/docs/source/introduction/fedot_features/automation/dimensionality_reduction.rst
@@ -1,7 +1,16 @@
-Dimensionality reduction
-------------------------
+Dimensionality operations
+-------------------------
 
-Currently FEDOT supports only feature selection algorithms provided by scikit-learn, but there aren't extraction ones.  
+FEDOT supports bunch of dimension preprocessing operations that can be be added to the pipeline as a node.
 
-..
-    [А что по ITMO_FS у нас, кстати? Вроде нигде не используется на 28.03.2023]
+Feature selection
+^^^^^^^^^^^^^^^^^
+
+There are different linear and non-linear algorithms for regression and classification tasks
+which uses scikit-learn's Recursive Feature Elimination (RFE).
+
+Feature extraction
+^^^^^^^^^^^^^^^^^^
+
+Currently there are PCA (kernel methods supported), fast ICA and
+polynomial features algorithms for generating new features. 
diff --git a/docs/source/introduction/fedot_features/automation/learning.rst b/docs/source/introduction/fedot_features/automation/learning.rst
@@ -4,7 +4,7 @@ Learning
 FEDOT is capable of setting its 'automation rate' by omitting some of its parameters.
 For example, if you just create the FEDOT instance and call the ``fit`` method with the appropriate dataset on it,
 you will have a full automation of the learning process,
-see :doc:`automated composing <../../tutorial/composing_pipelines/automated_creation>`
+see :doc:`automated composing </introduction/tutorial/composing_pipelines/automated_creation>`
 
 At the same time, if you pass some of the parameters, you will have a partial automation,
-see :doc:`manual composing <../../tutorial/composing_pipelines/manual_creation>`
+see :doc:`manual composing </introduction/tutorial/composing_pipelines/manual_creation>`
diff --git a/docs/source/introduction/fedot_features/automation/models_used.rst b/docs/source/introduction/fedot_features/automation/models_used.rst
@@ -0,0 +1,14 @@
+Models used
+-----------
+
+Using the parameter ``preset`` of the :doc:`main API </api/api>` you can specify
+what models are available during the learning process. 
+
+It influences:
+
+* composing speed and quality
+* computational behaviour
+* task relevance
+
+.. note::
+    To make it simple, FEDOT uses ``auto`` by default to identify the best choice for you.
diff --git a/docs/source/introduction/fedot_features/automation/used_models.rst b/docs/source/introduction/fedot_features/automation/used_models.rst
diff --git a/docs/source/introduction/fedot_features/tasks.rst b/docs/source/introduction/fedot_features/tasks.rst
@@ -5,7 +5,4 @@ FEDOT is currently capable of solving:
 
 * Classification
 * Regression
-* Clustering (but not as the main one yet)
-* Time-series problems ([uni/multi]variate)
-
-The last is not true to many other frameworks and also that versatility is not true either.
+* Time-series forecasting ([uni/multi]variate)
diff --git a/docs/source/introduction/tutorial/quickstart.rst b/docs/source/introduction/tutorial/quickstart.rst
@@ -3,8 +3,7 @@ Quick start
 
 FEDOT Framework quick start guide.
 
-It does not matter which operating system you use. FEDOT works perfectly on either Windows, or macOS,
-or Linux distributions.
+Fedot has been perfectly supported on Windows, Linux, and MacOS systems.
 
 Firstly, you need to setup the environment to work with the relevant libs.
 Secondly, you must learn, how to build your solutions: you can do it both manually and automatically.

diff --git a/docs/source/introduction/what_is_fedot.rst b/docs/source/introduction/what_is_fedot.rst
@@ -27,7 +27,7 @@ FEDOT supports a full life-сyсle of machine learning task that includes prepro
     prediction = model.predict(features=x_test)
     metrics = model.get_metrics(target=y_test)
 
-Once FEDOT finds the best solution you have an opportunity :doc:`to save it <../basics/pipeline_save_load>` and look closer to the solution and optimization process if needed.
+Once FEDOT finds the best solution you have an opportunity :doc:`to save it </basics/pipeline_save_load>` and look closer to the solution and optimization process if needed.
 
 |Example of solution|