## Background

Backpropagation is a common method for training a neural network. There is no shortage of papers online that attempt to explain how backpropagation works, but few that include an example with actual numbers. This post is my attempt to explain how it works with a concrete example that folks can compare their own calculations to in order to ensure they understand backpropagation correctly.

If this kind of thing interests you, you should sign up for my newsletter where I post about AI-related projects that I’m working on.

## Backpropagation in Python

You can play around with a Python script that I wrote that implements the backpropagation algorithm in this Github repo.

## Backpropagation Visualization

For an interactive visualization showing a neural network as it learns, check out my Neural Network visualization.

## Additional Resources

If you find this tutorial useful and want to continue learning about neural networks, machine learning, and deep learning, I highly recommend checking out Adrian Rosebrock’s new book, Deep Learning for Computer Vision with Python. I really enjoyed the book and will have a full review up soon.

## Overview

For this tutorial, we’re going to use a neural network with two inputs, two hidden neurons, two output neurons. Additionally, the hidden and output neurons will include a bias.

Here’s the basic structure:

In order to have some numbers to work with, here are the initial weights, the biases, and training inputs/outputs:

The goal of backpropagation is to optimize the weights so that the neural network can learn how to correctly map arbitrary inputs to outputs.

For the rest of this tutorial we’re going to work with a single training set: given inputs 0.05 and 0.10, we want the neural network to output 0.01 and 0.99.

## The Forward Pass

To begin, lets see what the neural network currently predicts given the weights and biases above and inputs of 0.05 and 0.10. To do this we’ll feed those inputs forward though the network.

We figure out the *total net input* to each hidden layer neuron, *squash* the total net input using an *activation function* (here we use the *logistic function*), then repeat the process with the output layer neurons.

*net input*by some sources.

Here’s how we calculate the total net input for :

We then squash it using the logistic function to get the output of :

Carrying out the same process for we get:

We repeat this process for the output layer neurons, using the output from the hidden layer neurons as inputs.

Here’s the output for :

And carrying out the same process for we get:

### Calculating the Total Error

We can now calculate the error for each output neuron using the squared error function and sum them to get the total error:

For example, the target output for is 0.01 but the neural network output 0.75136507, therefore its error is:

Repeating this process for (remembering that the target is 0.99) we get:

The total error for the neural network is the sum of these errors:

## The Backwards Pass

Our goal with backpropagation is to update each of the weights in the network so that they cause the actual output to be closer the target output, thereby minimizing the error for each output neuron and the network as a whole.

### Output Layer

Consider . We want to know how much a change in affects the total error, aka .

By applying the chain rule we know that:

Visually, here’s what we’re doing:

We need to figure out each piece in this equation.

First, how much does the total error change with respect to the output?

Next, how much does the output of change with respect to its total net input?

The partial derivative of the logistic function is the output multiplied by 1 minus the output:

Finally, how much does the total net input of change with respect to ?

Putting it all together:

You’ll often see this calculation combined in the form of the delta rule:

Alternatively, we have and which can be written as , aka (the Greek letter delta) aka the *node delta*. We can use this to rewrite the calculation above:

Therefore:

Some sources extract the negative sign from so it would be written as:

To decrease the error, we then subtract this value from the current weight (optionally multiplied by some learning rate, eta, which we’ll set to 0.5):

We can repeat this process to get the new weights , , and :

We perform the actual updates in the neural network *after* we have the new weights leading into the hidden layer neurons (ie, we use the original weights, not the updated weights, when we continue the backpropagation algorithm below).

### Hidden Layer

Next, we’ll continue the backwards pass by calculating new values for , , , and .

Big picture, here’s what we need to figure out:

Visually:

We’re going to use a similar process as we did for the output layer, but slightly different to account for the fact that the output of each hidden layer neuron contributes to the output (and therefore error) of multiple output neurons. We know that affects both and therefore the needs to take into consideration its effect on the both output neurons:

Starting with :

We can calculate using values we calculated earlier:

And is equal to :

Plugging them in:

Following the same process for , we get:

Therefore:

Now that we have , we need to figure out and then for each weight:

We calculate the partial derivative of the total net input to with respect to the same as we did for the output neuron:

Putting it all together:

You might also see this written as:

We can now update :

Repeating this for , , and

Finally, we’ve updated all of our weights! When we fed forward the 0.05 and 0.1 inputs originally, the error on the network was 0.298371109. After this first round of backpropagation, the total error is now down to 0.291027924. It might not seem like much, but after repeating this process 10,000 times, for example, the error plummets to 0.0000351085. At this point, when we feed forward 0.05 and 0.1, the two outputs neurons generate 0.015912196 (vs 0.01 target) and 0.984065734 (vs 0.99 target).

If you’ve made it this far and found any errors in any of the above or can think of any ways to make it clearer for future readers, don’t hesitate to drop me a note. Thanks!

nice explanation.

Hey Matt. I’m stuck here: I got dEo2/douth1=-0.017144288. you got -0.019049. Can you check again?

Nicely explained. Thanks :)

Very helpful. Thanks

Thank you a lot! Didn’t get on lecture.Here is the best explanation!

To get the net input of h1 you need to calculate w1*i1 + w3*i2 + b1*1. The weight w2 does not connect to h1

It can look like that in the pictures, however h1 is in fact from w1 and w2. The weight # is created based off of the hidden layer. For example h1 has w1 and w2 while h2 has w3 and w4.

If we want to extend this to a batch of training samples, do we calculate the output layer errors simply by averaging the error at each output over the training samples, then obtain the total by summing the averages?

e.g. E[O][1] = sum between 1 and N of E[O][1][n], where n is the training sample index and {1 <= n <= N} (Is that correct?)

EDIT – forgot to divide sum by N

I want to thank you for this write up. Best one on the net! Your breakdown of the math is incredibly helpful

Don’t we need to update b1 and b2?

For really basic NN’s, it is not needed and can be left alone. Think of the biases as shifting your graph while the weights change the slope. If you wish to update the biases, simply do the same thing as the weights but don’t add the previous node as it is not attached to it. For example b2 is the same as w5, but without the hiddens (w5 would be h1) because it is not connected to any hiddens. b1 is the same as w1 but without the hidden stuff like h1(1-h1). Hope this helps, if not just leave another question.

Sorry but I messed this up, the h1(1-h1) would be removed for b2 NOT b1, as b1 connects to hidden layer and the b2 does not. For b1 you would remove the input (i1) stuff.

This is really amazing.thankyou for the clear explanation.

Neuronal networking in VBA+Excel |

Excuse me. If I have 1000 training data, In bckpropagation, the loss function would be E=sum((target1-out1)^2+(target2-out2)^2+……(target1000-out1000)^2), should partial derivative of the weights to E be divided by 1000(the number of data) when update the weights

Wonderful explaination!, btw how to reduce biases at input and hidden layer?

Deep Learning Resources – Machine Learning

Bravo Matt, la migliore spiegazione!

A quick introduction to derivatives for machine learning people – Data Science Austria

Hey Matt, I’ve been trying to duplicate your results in a network I am building, and running into an issue. Using your examples as my test data, I have passed all tests except for the weight values in w6 and w7. My values are not matching.

From a sniff test perspective, w6 appears it should be increasing since the output is lower than the expected (0.7 -> ~0.9) , but your calculation has the weight being reduced (0.45 -> ~0.41). With w7 your example has the weight increasing (0.5 -> ~0.51), but it appears it should be decreasing since the output is higher than expected (~0.7 -> 0.1).

I know this is an older post, but if you could double check those values and let me know, that would help me a lot.

:)

Thank you Matt for your very good explanation! How do you create the figures of the basic structre, … etc. ?

Hi I need help, please have someone help me

What is your problem?

that was very nice

could the problem be solved linear programming ?