diff --git a/README.md b/README.md
index 6b1560c..307629a 100644
--- a/README.md
+++ b/README.md
@@ -14,6 +14,7 @@ Scalable Second-Order methods in PyTorch.
   - [[paper (preprint)](https://arxiv.org/abs/1906.02506)]
 
 ## Scalable Second-Order Optimization
+![](docs/overview.png)
 
 ### Optimizers
 
@@ -46,8 +47,6 @@ Refer Section 6 of [Optimization Methods for Large-Scale Machine Learning](https
 
 ### Approximation Methods
 
-![](docs/overview.png)
-
 PyTorch-SSO calculates the curvature as a layer-wise block-diagonal matrix. 
 
 You can specify the approximation method for the curvatures in each layer from the follwing.