Sie sind auf Seite 1von 17

# 2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

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

Backpropagation in Python
email.

You can play around with a Python script that I wrote that implements the
Join 2,617 other followers
backpropagation algorithm in this Github repo.
Backpropagation Visualization
Follow
For an interactive visualization showing a neural network as it learns, check out
Hey there! I’m a data scientist
at Help Scout where I wrangle Additional Resources
data to gain insights into our
product and business. I also If you find this tutorial useful and want to continue learning about neural
built Lean Domain Search,
networks, machine learning, and deep learning, I highly recommend checking
Preceden and many other
out Adrian Rosebrock’s new book, Deep Learning for Computer Vision with
software products over the
years. Python. I really enjoyed the book and will have a full review up soon.

Overview

Search … 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

## Matt Mazur Retweeted

Dan Luu
@danluu
twitter on HN, a bunch of
devs drop in to say bot
https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 1/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur
removal is so easy twitter obv.
must not care at all.

## Turns out twitter is removing

has ~300M MAU, making the
error tolerance v. low. This
seems like a really hard
problem.

## Matt Mazur Retweeted

Sweatpants Cher
@House_Feminist In order to have some numbers to work with, here are the initial weights, the
I secretly hope that twitter biases, and training inputs/outputs:
keeps extending the
character limit as a social
experiment, slowly
conditioning our attention
spans until we’re able to read
actual books again

## Matt Mazur Retweeted

Alan White
@aljwhite
This is one of the most
terrifying things I’ve seen in
all my life

## The goal of backpropagation is to optimize the weights so that the neural

Matt Mazur Retweeted
network can learn how to correctly map arbitrary inputs to outputs.
Kitze
@thekitze
For the rest of this tutorial we’re going to work with a single training set: given
holy shit I'm hanging this on
my wall 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
Feb 12, 2018
net input using an activation function (here we use the logistic function), then

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 2/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

Matt Mazur Retweeted repeat the process with the output layer neurons.
ABC News
@ABC

NASA's Curiosity rover Total net input is also referred to as just net input by some sources.
captures an incredible
panoramic view of Mars from
the Gale Crater, showing Here’s how we calculate the total net input for :
dunes, buttes and ridges
across the Red Planet's
surface. abcn.ws/2GJl2NP

Feb 4, 2018

## Matt Mazur Retweeted

I Am Devloper
We repeat this process for the output layer neurons, using the output from the
@iamdevloper
hidden layer neurons as inputs.
Elon Musk: I'm putting people
on Mars!
Developers: Fantastic, more Here’s the output for :
timezones to support.

Feb 9, 2018

## 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:

Some sources refer to the target as the ideal and the output as the actual.

## The 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 is 0.01 but the neural network output
0.75136507, therefore its error is:
https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 3/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

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 .

## is read as “the partial derivative of with respect to “. You

can also say “the gradient with respect to “.

## We need to figure out each piece in this equation.

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

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 4/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

is sometimes expressed as

When we take the partial derivative of the total error with respect to ,
the quantity becomes zero because does not
affect it which means we’re taking the derivative of a constant which is
zero.

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:

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 5/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

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):

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

## 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:

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 6/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

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 :

## Following the same process for , we get:

Therefore:

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 7/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

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:

## 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

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 8/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

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!

Like
82 bloggers like this.

Related

## Experimenting with a The State of Emergent I'm going to write more

Neural Network-based Mind often. For real this time.
Poker Bot In "Emergent Mind" In "Writing"
In "Poker Bot"

Posted on March 17, 2015 by Mazur. This entry was posted in Machine Learning and tagged ai,
backpropagation, machine learning, neural networks. Bookmark the permalink.

## ← Introducing TetriNET Bot Source Code Published

ABTestCalculator.com, an Open on Github →
Source A/B Test
Significance Calculator

## 617 thoughts on “A Step by Step Backpropagation Example”

Pavel Koryakin
— December 4, 2017 at 10:25 pm

Great tutorial!
But how can I update biases using back propagation ?

Matus Moravcik
— January 28, 2018 at 9:52 am

## You can imagine biases as weights of a neuron that always outputs 1.

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 9/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

Jack
— December 7, 2017 at 4:03 am

Same way. Just take the partial with respect to the bias instead of the weight.

WangLu
— December 20, 2017 at 9:51 pm

It’s easy to generalize the method accroding to weight update. You just need
use the similar way to calculate dEtotal/dbias to get the step for bias and update
it.

shikha purwar
— December 8, 2017 at 4:08 am

very good

HansMueller
— December 10, 2017 at 1:32 pm

The best explanaition of backpropagation I’ve read! Really helped me, thank you!

## A quick overview of back propagation | Panthmas

!g
ni
P

mobymotion
— December 14, 2017 at 6:16 am

## Thank you thank you thank you.

Just about to give up trying to understand back prop, before I saw this.

— December 14, 2017 at 6:17 am

## Thank you thank you thank you.

Just about to give up trying to understand back prop, before I saw this.

## How to create a Neural Network in JavaScript in only 30 lines of code – Learning

!g
ni
P

JavaScript

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 10/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

Ilya Nonename
— December 14, 2017 at 12:15 pm

Very, nice.
Did not succeed at first, because i was using 0 and 1 in one of inputs.
I figured out that 0 does not adjust weight.
So i shifted inputs by 0.5 (-0.5 and 0.5 instead 0 and 1) and it worked.

neuralnetwork963574777
— December 15, 2017 at 1:04 am

Hi Matt,
One question. In this article you calculate the Error function at the end of the forward
propagation process as

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

I understand that this way of calculating the Error function was mostly used in the past
and now we should use cross entropy. However, getting back to the squared error
function – because the difference between the target and output is power 2, the result is
always positive (regardless whether target > output or vice versa). That means that
regardless that the actual network output result (target – output) can be positive error or
negative error, we always back propagate the positive E function and eventually use the
fractions of it at any neuron to adjust its weight and bias.
So, the adjustment goes always in one direction. Since it was successfully used in the
past, how that worked? Or getting back to your example, we can use different input
number and come up with negative (target – output) but the Error function will still be
positive, and so the weight and bios adjustments for each neuron.

Regards
Igor

Henry Henri
— January 2, 2018 at 12:43 pm

The squared error is always positive. But for backpropagation you use the
(partial) derivative of the error function, which is linear and hence can be
positive or negative.

Wowza
— December 17, 2017 at 11:28 pm

Man I’m studying for a final and this explained the algorithm better than the textbook.
you’re actively the best

Shashank

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 11/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur
— December 18, 2017 at 1:00 pm

Amazing explanation!! can you please explain in the similar fashion about updating the
bias. As i am confused bias is only for a layer how can we update for every neuron??

Vinicius Silva
— December 18, 2017 at 6:23 pm

So what happens next? What do you mean by “after repeating this process 10,000 times,
for example, the error plummets to 0.0000351085”? Should we keep using the same
input record in all these 10000 iterations? I think I understood what has been explained
by this text, but I wish you could ellaborate a bit more on the whole neural network
learning process.

Also, can you provide a general idea on what is happening in your neural network
visualization example? What were you feeding the network during all those many
iterations?

Sorry for asking so many questions, it’s just I’m trying to get a deep understanding on
this topic, but failing to find quality material that isn’t too difficult for beginner like me.

Thank you.

oliveiravini1994
— December 18, 2017 at 6:25 pm

So what happens next? What do you mean by “after repeating this process 10,000 times,
for example, the error plummets to 0.0000351085”? Should we keep using the same
input record in all these 10000 iterations? I think I understood what has been explained
by this text, but I wish you could ellaborate a bit more on the whole neural network
learning process.

Also, can you provide a general idea on what is happening in your neural network
visualization example? What were you feeding the network during all those many
iterations?

Sorry for asking so many questions, it’s just I’m trying to get a deep understanding on
this topic, but failing to find quality material that isn’t too difficult for beginner like me.

Thank you.

Daniel
— December 23, 2017 at 10:18 pm

this network doesn’t work well, my outputs are exactly like the example above,
but training 16000 inputs for xor problem, the error is still very big, with another
net I got very small error with 2000 inputs and i didnt even touch eta

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 12/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

Henry Henri
— January 2, 2018 at 12:54 pm

If I get your question correctly you’re asking, whether to keep using the same
input for all training cycles, then the answer is no. The network learns by
example. The more examples you show the more it will learn. If you only show
one example, that’s the only case it will be able to work with. Depending on the
complexity of the task you might need to teach tens of thousands of different
samples throughout training. In some cases, it is appropriate to show the same
sample multiple times. E.g. if you want to train XOR (you need at least one
hidden layer for this) you have the possible samples (0,0 => 0) (1,0 => 1) (0,1 => 1)
and (1,1 => 0). You run them in some random order until you have reached an
error rate you’re comfortable with.

Daniel
— December 19, 2017 at 2:34 pm

why should it be called backpropagation if you don’t update the weights after you
calculed them? you can easily perform this operation from the input layer to the output
layer and get the same result… are you sure about “we use the original weights, not the
updated weights, when we continue the backpropagation algorithm below” ?

Henry Henri
— January 2, 2018 at 12:47 pm

## Updating and backpropagation are two separate steps. Backpropagation

pushes the error back through the network to find out how much responsibility
to assign to each weight. This responsibility is then used to update the weights.
You can interleave backpropagation and update for each layer, but first, you
have to calculate the error for the next layer before you can do the update.

— December 20, 2017 at 1:24 pm

## Great explanation! Short and clear!

WangLu
— December 20, 2017 at 9:44 pm

Great simpification a NN to a two layer single data structure. It’s really easy to learn for
beginner. Thanks man.

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 13/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur

Daniel Zheng
— December 23, 2017 at 10:10 pm

Great tutorial just finished going through the math and managed to reproduce the
calculation. Only a matter of time until I master it.

rspurge2
— December 25, 2017 at 3:17 pm

## This was seriously helpful. Thanks for writing it!

vinith kumar
— December 28, 2017 at 12:22 am

## why bias weight is not updated?

Saurabh
— December 28, 2017 at 4:29 am

Simple and intuitive explanation !! This is what I was looking for. Thank you.

Wallen Tan
— December 30, 2017 at 7:38 pm

Really good tutorial! One of the most helpful ones I’ve come across.

Does the bias unit in each layer have only one weight or would there be a separate
weight per connection with the nodes?

Willy Wonka
— January 2, 2018 at 9:43 am

Man you explained it all, i have finally succeeded to implement hidden layer
backpropagation now. THANKS!

Henry Henri
— January 2, 2018 at 12:49 pm

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 14/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur
Maybe I missed it, but I think you left out the update of the biases. It’s simple but still
might not be obvious to everyone new to the topic.

James Liu
— January 4, 2018 at 1:49 am

## hi, i have two questions:

1) how do u ensure differentiating give the minimum value? instead of giving the
maximum value?
2) Won’t differentiating it once give the lowest maximum value? which is the smallest
error. why do we have to differentiate it 10000 times?

## thanks, from singapore here!

Priyank
— January 4, 2018 at 11:38 am

Dear sir,
Error should be:
Error=(1/2)(out-target)^2
isn’t it?

Priyank
— January 4, 2018 at 11:54 am

## sorry it’s correct

Rashmi G
— January 8, 2018 at 5:26 am

Hi..A very useful article..I have a doubt with calculating error for o1 . Here I am getting a
negative value (-0.2747). Kindly help me here. Thanks in advance.

— January 8, 2018 at 10:17 am

## Thank you for this great tutorial.

— January 8, 2018 at 10:20 am

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 15/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur
for those who ask about bias updating you may assume the weights of bias as
W’ 1 , W’ 2 , W’ 3 , W’ 4 and then apply same process on them

Draft

!g
ni
P
How To Build an Artificial Neural Network From Scratch

!g
ni
P
Deept
— January 13, 2018 at 4:43 am

## This is the best explanation of backward propagation I ever read.!!

Manas
— January 17, 2018 at 9:49 pm

This is so far the best explanation of backprop I read so far. Thankyou so much for this!

felipe
— January 18, 2018 at 2:39 pm

“We can calculate \frac{\partial E_{o1}}{\partial net_{o1}} using values we calculated earlier:
” whats going on here? how do i get the result?

## Great explanation btw

Kashyap Mahanta
— January 21, 2018 at 1:21 am

## The best one so far !!!

Ankit Bhaukajee
— January 25, 2018 at 9:54 am

This blog is the best of the best explanation I have found in decoding backprop.
Everybody was giving their own formula and I was not able to grasp the intuition but this
post really helped me what is happening inside. Thank you for helping people.

Thomas
— January 28, 2018 at 1:41 pm

https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ 16/17
2/19/2018 A Step by Step Backpropagation Example – Matt Mazur
Amazing explanation!
Though I used some random inputs and set the target values to double the input values
(so the first output of the network is double the value of the first input, and the second
output is double the value of the second input). It worked perfectly for specific input
values. For example [0.03,0.09] would output very close to [0.06,0.18].
Though when I ran the algorithm in a loop many times, then tested the network (i.e.
without using backprop and target values) it just outputted the same values that were
outputted in the last iteration of the loop, rather than doubling the new values I inputted
into the network.
So basically it only worked when I ran the backprop with the target values – though I
want it to work without the target values!
Can anyone suggest anything? Sorry if I’m not being very clear, I’d be happy to explain
myself if anyone is confused about what I mean.

Thomas
— January 30, 2018 at 1:28 pm

Don’t worry, I worked it out! I was backpropagating too much for each pair of
inputs, and not putting enough test inputs in. I should’ve been backpropagating
alot less and using alot more test inputs!

Christophe Schnitzler
— January 30, 2018 at 6:02 am

Hi
I really don’t get how you calculate this line
\frac{\partial E_{total}}{\partial out_{o1}} = 2 * \frac{1}{2}(target_{o1} – out_{o1})^{2 – 1} * -1 + 0
I’m sure it is really simple, but cannot figure it out. Thanks for your help!