diff --git a/README.md b/README.md
index afdeaa3..5ce1ba8 100644
--- a/README.md
+++ b/README.md
@@ -1,17 +1,22 @@
 # Plotabit
-
 Dataset link: https://www.kaggle.com/datasets/fedesoriano/stellar-classification-dataset-sdss17
-
+Deep Learning: https://machinelearningmastery.com/tutorial-first-neural-network-python-keras/
 
 ## Columns
 
-|Keep         |Not sure    |Skip        |
-|:-----------:|:----------:|:----------:|
-|alpha        |field_ID    |obj_ID      |
-|delta        |            |run_ID      |
-|u            |            |rerun_ID    |
-|g            |            |plate       |
-|r            |            |MJD         |
-|i            |            |fiber_ID    |
-|z            |            |            |
-|redshift     |            |            |
+|Keep         |Skip        |
+|:-----------:|:----------:|
+|alpha        |obj_ID      |
+|delta        |run_ID      |
+|u            |rerun_ID    |
+|g            |plate       |
+|r            |MJD         |
+|i            |fiber_ID    |
+|z            |            |
+|redshift     |            |
+
+## Analysis
+
+- [ ] Train models with just 20000 "GALAXY" class (has an impact?)
+- [ ] Which model is the best, ratio learn_time/precision
+- [ ] Can we drop more categories and have same results (useless data?)