A Step by Step Backpropagation Example


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 and their applications, I highly recommend checking out Adrian Rosebrock’s excellent tutorial on Getting Started with Deep Learning and Python.


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:

neural_network (7)

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

neural_network (9)

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.

Total net input is also referred to as just net input by some sources.

Here’s how we calculate the total net input for h_1:

net_{h1} = w_1 * i_1 + w_2 * i_2 + b_1 * 1

net_{h1} = 0.15 * 0.05 + 0.2 * 0.1 + 0.35 * 1 = 0.3775

We then squash it using the logistic function to get the output of h_1:

out_{h1} = \frac{1}{1+e^{-net_{h1}}} = \frac{1}{1+e^{-0.3775}} = 0.593269992

Carrying out the same process for h_2 we get:

out_{h2} = 0.596884378

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

Here’s the output for o_1:

net_{o1} = w_5 * out_{h1} + w_6 * out_{h2} + b_2 * 1

net_{o1} = 0.4 * 0.593269992 + 0.45 * 0.596884378 + 0.6 * 1 = 1.105905967

out_{o1} = \frac{1}{1+e^{-net_{o1}}} = \frac{1}{1+e^{-1.105905967}} = 0.75136507

And carrying out the same process for o_2 we get:

out_{o2} = 0.772928465

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:

E_{total} = \sum \frac{1}{2}(target - output)^{2}

Some sources refer to the target as the ideal and the output as the actual.
The \frac{1}{2} is included so that exponent is cancelled when we differentiate later on. The result is eventually multiplied by a learning rate anyway so it doesn’t matter that we introduce a constant here [1].

For example, the target output for o_1 is 0.01 but the neural network output 0.75136507, therefore its error is:

E_{o1} = \frac{1}{2}(target_{o1} - out_{o1})^{2} = \frac{1}{2}(0.01 - 0.75136507)^{2} = 0.274811083

Repeating this process for o_2 (remembering that the target is 0.99) we get:

E_{o2} = 0.023560026

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

E_{total} = E_{o1} + E_{o2} = 0.274811083 + 0.023560026 = 0.298371109

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 w_5. We want to know how much a change in w_5 affects the total error, aka \frac{\partial E_{total}}{\partial w_{5}}.

\frac{\partial E_{total}}{\partial w_{5}} is read as “the partial derivative of E_{total} with respect to w_{5}“. You can also say “the gradient with respect to w_{5}“.

By applying the chain rule we know that:

\frac{\partial E_{total}}{\partial w_{5}} = \frac{\partial E_{total}}{\partial out_{o1}} * \frac{\partial out_{o1}}{\partial net_{o1}} * \frac{\partial net_{o1}}{\partial w_{5}}

Visually, here’s what we’re doing:

output_1_backprop (4)

We need to figure out each piece in this equation.

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

E_{total} = \frac{1}{2}(target_{o1} - out_{o1})^{2} + \frac{1}{2}(target_{o2} - out_{o2})^{2}

\frac{\partial E_{total}}{\partial out_{o1}} = 2 * \frac{1}{2}(target_{o1} - out_{o1})^{2 - 1} * -1 + 0

\frac{\partial E_{total}}{\partial out_{o1}} = -(target_{o1} - out_{o1}) = -(0.01 - 0.75136507) = 0.74136507

-(target - out) is sometimes expressed as out - target
When we take the partial derivative of the total error with respect to out_{o1}, the quantity \frac{1}{2}(target_{o2} - out_{o2})^{2} becomes zero because out_{o1} does not affect it which means we’re taking the derivative of a constant which is zero.

Next, how much does the output of o_1 change with respect to its total net input?

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

out_{o1} = \frac{1}{1+e^{-net_{o1}}}

\frac{\partial out_{o1}}{\partial net_{o1}} = out_{o1}(1 - out_{o1}) = 0.75136507(1 - 0.75136507) = 0.186815602

Finally, how much does the total net input of o1 change with respect to w_5?

net_{o1} = w_5 * out_{h1} + w_6 * out_{h2} + b_2 * 1

\frac{\partial net_{o1}}{\partial w_{5}} = 1 * out_{h1} * w_5^{(1 - 1)} + 0 + 0 = out_{h1} = 0.593269992

Putting it all together:

\frac{\partial E_{total}}{\partial w_{5}} = \frac{\partial E_{total}}{\partial out_{o1}} * \frac{\partial out_{o1}}{\partial net_{o1}} * \frac{\partial net_{o1}}{\partial w_{5}}

\frac{\partial E_{total}}{\partial w_{5}} = 0.74136507 * 0.186815602 * 0.593269992 = 0.082167041

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

\frac{\partial E_{total}}{\partial w_{5}} = -(target_{o1} - out_{o1}) * out_{o1}(1 - out_{o1}) * out_{h1}

Alternatively, we have \frac{\partial E_{total}}{\partial out_{o1}} and \frac{\partial out_{o1}}{\partial net_{o1}} which can be written as \frac{\partial E_{total}}{\partial net_{o1}}, aka \delta_{o1} (the Greek letter delta) aka the node delta. We can use this to rewrite the calculation above:

\delta_{o1} = \frac{\partial E_{total}}{\partial out_{o1}} * \frac{\partial out_{o1}}{\partial net_{o1}} = \frac{\partial E_{total}}{\partial net_{o1}}

\delta_{o1} = -(target_{o1} - out_{o1}) * out_{o1}(1 - out_{o1})


\frac{\partial E_{total}}{\partial w_{5}} = \delta_{o1} out_{h1}

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

\frac{\partial E_{total}}{\partial w_{5}} = -\delta_{o1} out_{h1}

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):

w_5^{+} = w_5 - \eta * \frac{\partial E_{total}}{\partial w_{5}} = 0.4 - 0.5 * 0.082167041 = 0.35891648

Some sources use \alpha (alpha) to represent the learning rate, others use \eta (eta), and others even use \epsilon (epsilon).

We can repeat this process to get the new weights w_6, w_7, and w_8:

w_6^{+} = 0.408666186

w_7^{+} = 0.511301270

w_8^{+} = 0.561370121

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 w_1, w_2, w_3, and w_4.

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

\frac{\partial E_{total}}{\partial w_{1}} = \frac{\partial E_{total}}{\partial out_{h1}} * \frac{\partial out_{h1}}{\partial net_{h1}} * \frac{\partial net_{h1}}{\partial w_{1}}



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 out_{h1} affects both out_{o1} and out_{o2} therefore the \frac{\partial E_{total}}{\partial out_{h1}} needs to take into consideration its effect on the both output neurons:

\frac{\partial E_{total}}{\partial out_{h1}} = \frac{\partial E_{o1}}{\partial out_{h1}} + \frac{\partial E_{o2}}{\partial out_{h1}}

Starting with \frac{\partial E_{o1}}{\partial out_{h1}}:

\frac{\partial E_{o1}}{\partial out_{h1}} = \frac{\partial E_{o1}}{\partial net_{o1}} * \frac{\partial net_{o1}}{\partial out_{h1}}

We can calculate \frac{\partial E_{o1}}{\partial net_{o1}} using values we calculated earlier:

\frac{\partial E_{o1}}{\partial net_{o1}} = \frac{\partial E_{o1}}{\partial out_{o1}} * \frac{\partial out_{o1}}{\partial net_{o1}} = 0.74136507 * 0.186815602 = 0.138498562

And \frac{\partial net_{o1}}{\partial out_{h1}} is equal to w_5:

net_{o1} = w_5 * out_{h1} + w_6 * out_{h2} + b_2 * 1

\frac{\partial net_{o1}}{\partial out_{h1}} = w_5 = 0.40

Plugging them in:

\frac{\partial E_{o1}}{\partial out_{h1}} = \frac{\partial E_{o1}}{\partial net_{o1}} * \frac{\partial net_{o1}}{\partial out_{h1}} = 0.138498562 * 0.40 = 0.055399425

Following the same process for \frac{\partial E_{o2}}{\partial out_{o1}}, we get:

\frac{\partial E_{o2}}{\partial out_{h1}} = -0.019049119


\frac{\partial E_{total}}{\partial out_{h1}} = \frac{\partial E_{o1}}{\partial out_{h1}} + \frac{\partial E_{o2}}{\partial out_{h1}} = 0.055399425 + -0.019049119 = 0.036350306

Now that we have \frac{\partial E_{total}}{\partial out_{h1}}, we need to figure out \frac{\partial out_{h1}}{\partial net_{h1}} and then \frac{\partial net_{h1}}{\partial w} for each weight:

out_{h1} = \frac{1}{1+e^{-net_{h1}}}

\frac{\partial out_{h1}}{\partial net_{h1}} = out_{h1}(1 - out_{h1}) = 0.59326999(1 - 0.59326999 ) = 0.241300709

We calculate the partial derivative of the total net input to h_1 with respect to w_1 the same as we did for the output neuron:

net_{h1} = w_1 * i_1 + w_2 * i_2 + b_1 * 1

\frac{\partial net_{h1}}{\partial w_1} = i_1 = 0.05

Putting it all together:

\frac{\partial E_{total}}{\partial w_{1}} = \frac{\partial E_{total}}{\partial out_{h1}} * \frac{\partial out_{h1}}{\partial net_{h1}} * \frac{\partial net_{h1}}{\partial w_{1}}

\frac{\partial E_{total}}{\partial w_{1}} = 0.036350306 * 0.241300709 * 0.05 = 0.000438568

You might also see this written as:

\frac{\partial E_{total}}{\partial w_{1}} = (\sum\limits_{o}{\frac{\partial E_{total}}{\partial out_{o}} * \frac{\partial out_{o}}{\partial net_{o}} * \frac{\partial net_{o}}{\partial out_{h1}}}) * \frac{\partial out_{h1}}{\partial net_{h1}} * \frac{\partial net_{h1}}{\partial w_{1}}

\frac{\partial E_{total}}{\partial w_{1}} = (\sum\limits_{o}{\delta_{o} * w_{ho}}) * out_{h1}(1 - out_{h1}) * i_{1}

\frac{\partial E_{total}}{\partial w_{1}} = \delta_{h1}i_{1}

We can now update w_1:

w_1^{+} = w_1 - \eta * \frac{\partial E_{total}}{\partial w_{1}} = 0.15 - 0.5 * 0.000438568 = 0.149780716

Repeating this for w_2, w_3, and w_4

w_2^{+} = 0.19956143

w_3^{+} = 0.24975114

w_4^{+} = 0.29950229

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 0.000035085. 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!

237 thoughts on “A Step by Step Backpropagation Example

  1. Can you please do a tutorial for back propagation in Elmann recurrent neural networks!!…. It would be of a lot of help…

  2. Great article! I struggle with one aspect though and that is calculating the partial derivatives from out/net, using (output * (1.0 – output). In cases where output is 0 or 1, it effectively kills the pass through of error. How do you handle this when, especially at startup, the network can devolve (or even start) in this state?

    • How do you get an output of exactly 0 or 1? The only neuron that can output them is the input neuron (because they output what they take in), but we don’t do anything with them. For any other neuron, output is the result of sigmoid function so it cannot be exactly 0 or 1.

  3. In trying to understand NNs better I have produced a Google Docs spreadsheet that almost does what this link talks about:

    I was hoping that going through that exercise will give me a better mental picture of what a NN is doing when it is working. I should be able to see easily by direct comparison what happens to the set of numbers with each iteration (ie each new pair of FP and BP sheets). I think I have got it nearly working except for the stuff in the dashed purple box. I want to confirm that the purple arrow is pointing to the wrong blue arrow? I also want someone to tell me how to implement the purple box . . If I could work that out I think I could then repeat the FR and BP sheets and see how the Diff column evolves . . here is the current SS:


    I still found your description above heavy going – could you help me finish my spreadsheet or turn your example into a spreadsheet?


  4. I found this post very helpful but there is one thing that is confusing me. You apply the logistic function on the output nodes to compute out_o1 and out_o2 as the sigmoid function applied to the weighted sum of the hidden nodes. My question is, what if you are predicting an output that has a range wider than 0 to 1. I had thought the output layer was simply a weighted sum of the outputs of the final hidden layer. In that case you could output a value in any range, but this seems very limiting.

    • Yeah, this only works on a range of 0 to 1. If you have a range from say 0 to 100 you can divide by 100 to get it down to a range of 0 to 1 so that you can use this neural network.

      • Shouldn’t it be the weights connecting the last hidden layer and the output layer? If the network had an input of 200, hidden of 100, and another hidden of 50, and an output of 10; it wouldn’t work. Because the layer of 1 hundred would add the outputs multiplied by 10 of the 50 layer weights, right? Sorry if I am not very clear. What I’m just trying to ask is this: Is the Who referring to the weight connecting the last layer and output layer or is it connecting the current layer and next layer?

  5. You have used a squared error function. When using that as a cost function along with the sigmoid function, won’t your cost function have a lot of local minima ?

  6. Great Post with the step by step explanation.
    Based on your explanation I am able to implement the same for LOGICAL GATE solution where I have used 2 Input Nodes (for logical Inputs), 1 Hidden Layer with 2 Nodes and one Output Node.
    Thanks for this detailed explanation.

  7. It was a great explanation, Mazur. Thank you very much for the effort.
    It would be great if you could response my following query.
    I have been trying to train a model using neuralnet package. My model contains five input one output and a hidden layer with 10 nodes. logistic transfer function is selected as an activation function. All data are normalized in 0,1 range. As the logistic function is bounded between 0,1 the results are expected to be in the range 0,1. But some of the training results are negative. What is causing the model to compute negative output?
    H. Paudel

  8. 反向传播算法入门资源索引 | 我爱自然语言处理

  9. Building a multi-class text classifier from scratch using Neural Networks – Stokastik

  10. Many thanks for tutorial!

    But I have a problem, when im trying use more neurons (e.g. 20 inputs and 8 outputs) with more training data, NN total error is almost stagnates after few cycles. Any solution? (adaptive eta?)

  11. Hi – Thanks much , this is the only place I got some valid information in a way I can understand. I have few questions, if you don’t can you please check these out.

    Question 1:
    You have tried to find derivative of Etotal WRT to W5.

    In this how did you get this “-1”(Minus 1), doesn’t the derivate rule equate to following

    2*1/2(targeto1-out01) ( No Minus)

    Question 2:
    In chain rule. To find derivative of Etotal WRT to W5 following was used.

    DEtotal/Dw5 = Dnet01/Dw5 * Dout01/ Dnet01 * DEtotal/DOut01

    Here please note : DOut01/Dnet01 , Out01 was used and It makes sense.

    But to To find derivative of Etotal WRT to W1 following was used.

    DOut01/Dnet01 was not used, why is that.

    • Question 1: Using chain rule, since out_01 has a negative sign before it in the parenthesis, you need to propagate (no pun intended) that negative sign. In other words:

      \frac{d}{dx} {(1-x)}^2 = (-1) \times (2-1) \times {(1-x)}^{(2-1)}

      Question 2:

      Check out the image beneath the writing “Next, we’ll continue the backwards pass by calculating new values for w1, w2, w3, and w4”. He explains that D(E_total)/D(out_h1) = D(E_o1)/D(Out_h1) + D(E_o2)/D(Out_h1). And then he breaks out that first term, D(E_o1)/D(Out_h1), in the next images, which includes your term “DOut01/Dnet01”

      • Hi Hari – Thank you for your patience and answer. May be I am missing something here, on question 1.

        Here derivate of E total WRT Out01 is (target01 – Out01), so the correct value is “-0.74136507”.

        Minus sign stating derivative in decreasing , going in downward direction.

        I am still not sure how “-0.74136507” changed to “0.74136507”

  12. Mazur, Thank you for your explanation, but i have a little question, i have calculated exactly the same steps you described, but the second round i calculated the total error is 0.292165448
      different from what you have written, i have done something wrong in your opinion?

  13. 神经网络101 手把手快速入门神经网络基础 | David 9的博客 --- 没有"过拟合"

  14. If you read the original 1986 paper (Learning representations by back-propagating errors), you will find that they calculate the error function in the reverse (out – target). I believe the math works out the same, but the general consensus from a variety of sources is to calculate it as -(target – out).

    I would be curious what the original authors, or even Yann LeCun, has to say about this.

    You allude to this curiosity early in your rundown. I figured it would be worthwhile to point out its potential origin.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s