diff --git a/AI/Neural Networks/MLP.md b/AI/Neural Networks/MLP.md
index d077d1c..7d548b3 100644
--- a/AI/Neural Networks/MLP.md	
+++ b/AI/Neural Networks/MLP.md	
@@ -3,13 +3,13 @@
 	-   Universal approximation theorem
 -   Each hidden layer can operate as a different feature extraction layer
 -   Lots of weights to learn
--   Backpropagation is supervised
+-   [[Back-Propagation]] is supervised
 
 ![[mlp-arch.png]]
 
 # Universal Approximation Theory
 A finite feed-forward MLP with 1 hidden layer can in theory approximate any mathematical function
--   In practice not trainable with BP
+-   In practice not trainable with [[Back-Propagation|BP]]
 
 ![[activation-function.png]]
 ![[mlp-arch-diagram.png]]
\ No newline at end of file
diff --git a/AI/Neural Networks/MLP/Activation Functions.md b/AI/Neural Networks/MLP/Activation Functions.md
new file mode 100644
index 0000000..e69de29
diff --git a/AI/Neural Networks/MLP/Back-Propagation.md b/AI/Neural Networks/MLP/Back-Propagation.md
index 8016138..a4035e7 100644
--- a/AI/Neural Networks/MLP/Back-Propagation.md	
+++ b/AI/Neural Networks/MLP/Back-Propagation.md	
@@ -3,7 +3,104 @@ Error signal graph
 ![[mlp-arch-graph.png]]
 
 1. Error Signal
+	- $e_j(n)=d_j(n)-y_j(n)$
 2. Net Internal Sum
+	- $v_j(n)=\sum_{i=0}^mw_{ji}(n)y_i(n)$
 3. Output
+	- $y_j(n)=\varphi_j(v_j(n))$
 4. Instantaneous Sum of Squared Errors
-5. Average Squared Error
\ No newline at end of file
+	- $\mathfrak{E}(n)=\frac 1 2 \sum_{j\in C}e_j^2(n)$
+	- $C$ = o/p layer nodes
+5. Average Squared Error
+	- $\mathfrak E_{av}=\frac 1 N\sum_{n=1}^N\mathfrak E (n)$
+
+$$\frac{\partial\mathfrak E(n)}{\partial w_{ji}(n)}=
+\frac{\partial\mathfrak E(n)}{\partial e_j(n)}
+\frac{\partial e_j(n)}{\partial y_j(n)}
+\frac{\partial y_j(n)}{\partial v_j(n)}
+\frac{\partial v_j(n)}{\partial w_{ji}(n)}
+$$
+
+#### From 4
+$$\frac{\partial\mathfrak E(n)}{\partial e_j(n)}=
+e_j(n)$$
+#### From 1
+$$\frac{\partial e_j(n)}{\partial y_j(n)}=-1$$
+#### From 3 (note prime)
+$$\frac{\partial y_j(n)}{\partial v_j(n)}=
+\varphi_j'(v_j(n))$$
+#### From 2
+$$\frac{\partial v_j(n)}{\partial w_{ji}(n)}=
+y_i(n)$$
+
+## Composite
+$$\frac{\partial\mathfrak E(n)}{\partial w_{ji}(n)}=
+-e_j(n)\cdot
+\varphi_j'(v_j(n))\cdot
+y_i(n)
+$$
+
+$$\Delta w_{ji}(n)=
+-\eta\frac{\partial\mathfrak E(n)}{\partial w_{ji}(n)}$$
+$$\Delta w_{ji}(n)=
+\eta\delta_j(n)y_i(n)$$
+## Gradients
+#### Output
+$$\delta_j(n)=-\frac{\partial\mathfrak E (n)}{\partial v_j(n)}$$
+$$=-
+\frac{\partial\mathfrak E(n)}{\partial e_j(n)}
+\frac{\partial e_j(n)}{\partial y_j(n)}
+\frac{\partial y_j(n)}{\partial v_j(n)}$$
+$$=
+e_j(n)\cdot
+\varphi_j'(v_j(n))
+$$
+
+#### Local
+$$\delta_j(n)=-
+\frac{\partial\mathfrak E (n)}{\partial y_j(n)}
+\frac{\partial y_j(n)}{\partial v_j(n)}$$
+$$=-
+\frac{\partial\mathfrak E (n)}{\partial y_j(n)}
+\cdot
+\varphi_j'(v_j(n))$$
+$$\delta_j(n)=
+\varphi_j'(v_j(n))
+\cdot
+\sum_k \delta_k(n)\cdot w_{kj}(n)$$
+
+## Weight Correction
+$$\text{weight correction = learning rate $\cdot$ local gradient $\cdot$ input signal of neuron $j$}$$
+$$\Delta w_{ji}(n)=\eta\cdot\delta_j(n)\cdot y_i(n)$$
+
+-   Looking for partial derivative of error with respect to each weight
+-   4 partial derivatives
+	1.  Sum of squared errors WRT error in one output node
+	2.  Error WRT output $y$
+	3.  Output Y WRT Pre-activation function sum
+	4.  Pre-activation function sum WRT weight
+		-   Other weights constant, goes to zero
+		-   Leaves just $y_i$
+	-   Collect 3 boxed terms as delta $j$
+		-   Local gradient
+-   Weight correction can be too slow raw
+	-   Gets stuck
+	-   Add momentum
+
+![[mlp-local-hidden-grad.png]]
+
+-   Nodes further back
+	-   More complicated
+	-   Sum of later local gradients multiplied by backward weight (orange)
+	-   Multiplied by differential of activation function at node
+
+## Global Minimum
+-   Much more complex error surface than least-means-squared
+-   No guarantees of convergence
+	-   Non-linear optimisation
+-   Momentum
+	- $+\alpha\Delta w_{ji}(n-1), 0\leq|\alpha|<1$
+	-   Proportional to the change in weights last iteration
+		-   Can shoot past local minima if descending quickly
+
+![[mlp-global-minimum.png]]
\ No newline at end of file
diff --git a/AI/Neural Networks/SLP.md b/AI/Neural Networks/SLP.md
index ce9c89a..d12b757 100644
--- a/AI/Neural Networks/SLP.md	
+++ b/AI/Neural Networks/SLP.md	
@@ -4,4 +4,4 @@ $$=w^T(n)x(n)$$
 ![[slp-hyperplane.png]]
 Perceptron learning is performed for a finite number of iteration and then stops
 
-LMS is continuous learning that doesn't stop
\ No newline at end of file
+[[Least Mean Square|LMS]] is continuous learning that doesn't stop
\ No newline at end of file
diff --git a/img/mlp-global-minimum.png b/img/mlp-global-minimum.png
new file mode 100644
index 0000000..5b524cf
Binary files /dev/null and b/img/mlp-global-minimum.png differ
diff --git a/img/mlp-local-hidden-grad.png b/img/mlp-local-hidden-grad.png
new file mode 100644
index 0000000..1545f3c
Binary files /dev/null and b/img/mlp-local-hidden-grad.png differ