Bitcoin Is Evil

Finance 2017.05.08 23:30


Bitcoin Is Evil

It’s always important, and always hard, to distinguish positive economics — how things work — from normative economics — how things should be. Indeed, on many of the macro issues I’ve written about it has been obvious that large numbers of economists can’t bring themselves to make that distinction; they dislike activist government on political grounds, and this leads them to make really bad arguments about why fiscal stimulus can’t work and monetary stimulus will be disastrous. I don’t, by the way, think that this effect is symmetric: although people like Robert Lucas were quick to accuse people like Christy Romer of fabricating macro arguments to support a big-government agenda, this didn’t actually happen.

But I come now to talk not about macro but about money — specifically, about Bitcoin and all that.

So far almost all of the Bitcoin discussion has been positive economics — can this actually work? And I have to say that I’m still deeply unconvinced. To be successful, money must be both a medium of exchange and a reasonably stable store of value. And it remains completely unclear why BitCoin should be a stable store of value. Brad DeLong puts it clearly:

Underpinning the value of gold is that if all else fails you can use it to make pretty things. Underpinning the value of the dollar is a combination of (a) the fact that you can use them to pay your taxes to the U.S. government, and (b) that the Federal Reserve is a potential dollar sink and has promised to buy them back and extinguish them if their real value starts to sink at (much) more than 2%/year (yes, I know).

Placing a ceiling on the value of gold is mining technology, and the prospect that if its price gets out of whack for long on the upside a great deal more of it will be created. Placing a ceiling on the value of the dollar is the Federal Reserve’s role as actual dollar source, and its commitment not to allow deflation to happen.

Placing a ceiling on the value of bitcoins is computer technology and the form of the hash function… until the limit of 21 million bitcoins is reached. Placing a floor on the value of bitcoins is… what, exactly?

I have had and am continuing to have a dialogue with smart technologists who are very high on BitCoin — but when I try to get them to explain to me why BitCoin is a reliable store of value, they always seem to come back with explanations about how it’s a terrific medium of exchange. Even if I buy this (which I don’t, entirely), it doesn’t solve my problem. And I haven’t been able to get my correspondents to recognize that these are different questions.

But as I said, this is a positive discussion. What about the normative economics? Well, you should read Charlie Stross:

BitCoin looks like it was designed as a weapon intended to damage central banking and money issuing banks, with a Libertarian political agenda in mind—to damage states ability to collect tax and monitor their citizens financial transactions.

Go read the whole thing.

Stross doesn’t like that agenda, and neither do I; but I am trying not to let that tilt my positive analysis of BitCoin one way or the other. One suspects, however, that many BitCoin enthusiasts are, in fact, enthusiastic because, as Stross says, “it pushes the same buttons as their gold fetish.”

So let’s talk both about whether BitCoin is a bubble and whether it’s a good thing — in part to make sure that we don’t confuse these questions with each other.

Posted by steloflute


Summary: I learn best with toy code that I can play with. This tutorial teaches backpropagation via a very simple toy example, a short python implementation.

Edit: Some folks have asked about a followup article, and I'm planning to write one. I'll tweet it out when it's complete at @iamtrask. Feel free to follow if you'd be interested in reading it and thanks for all the feedback!

Just Give Me The Code:

01.= np.array([ [0,0,1],[0,1,1],[1,0,1],[1,1,1] ])
02.= np.array([[0,1,1,0]]).T
03.syn0 = 2*np.random.random((3,4)) - 1
04.syn1 = 2*np.random.random((4,1)) - 1
05.for in xrange(60000):
06.l1 = 1/(1+np.exp(-(,syn0))))
07.l2 = 1/(1+np.exp(-(,syn1))))
08.l2_delta = (y - l2)*(l2*(1-l2))
09.l1_delta = * (l1 * (1-l1))
10.syn1 +=
11.syn0 +=

Other Languages: D

However, this is a bit terse…. let’s break it apart into a few simple parts.

Part 1: A Tiny Toy Network

A neural network trained with backpropagation is attempting to use input to predict output.


Consider trying to predict the output column given the three input columns. We could solve this problem by simply measuring statistics between the input values and the output values. If we did so, we would see that the leftmost input column is perfectly correlated with the output. Backpropagation, in its simplest form, measures statistics like this to make a model. Let's jump right in and use it to do this.

2 Layer Neural Network:

01.import numpy as np
03.# sigmoid function
04.def nonlin(x,deriv=False):
06.return x*(1-x)
07.return 1/(1+np.exp(-x))
09.# input dataset
10.= np.array([  [0,0,1],
13.[1,1,1] ])
15.# output dataset           
16.= np.array([[0,0,1,1]]).T
18.# seed random numbers to make calculation
19.# deterministic (just a good practice)
22.# initialize weights randomly with mean 0
23.syn0 = 2*np.random.random((3,1)) - 1
25.for iter in xrange(10000):
27.# forward propagation
28.l0 = X
29.l1 = nonlin(,syn0))
31.# how much did we miss?
32.l1_error = - l1
34.# multiply how much we missed by the
35.# slope of the sigmoid at the values in l1
36.l1_delta = l1_error * nonlin(l1,True)
38.# update weights
39.syn0 +=,l1_delta)
41.print "Output After Training:"
42.print l1
Output After Training:
[[ 0.00966449]
 [ 0.00786506]
 [ 0.99358898]
 [ 0.99211957]]
XInput dataset matrix where each row is a training example
yOutput dataset matrix where each row is a training example
l0First Layer of the Network, specified by the input data
l1Second Layer of the Network, otherwise known as the hidden layer
syn0First layer of weights, Synapse 0, connecting l0 to l1.
*Elementwise multiplication, so two vectors of equal size are multiplying corresponding values 1-to-1 to generate a final vector of identical size.
-Elementwise subtraction, so two vectors of equal size are subtracting corresponding values 1-to-1 to generate a final vector of identical size. x and y are vectors, this is a dot product. If both are matrices, it's a matrix-matrix multiplication. If only one is a matrix, then it's vector matrix multiplication.

As you can see in the "Output After Training", it works!!! Before I describe processes, I recommend playing around with the code to get an intuitive feel for how it works. You should be able to run it "as is" in an ipython notebook (or a script if you must, but I HIGHLY recommend the notebook). Here are some good places to look in the code: 

• Compare l1 after the first iteration and after the last iteration. 
• Check out the "nonlin" function. This is what gives us a probability as output.
• Check out how l1_error changes as you iterate. 
• Take apart line 36. Most of the secret sauce is here. 
• Check out line 39. Everything in the network prepares for this operation. 

Let's walk through the code line by line.

Recommendation: open this blog in two screens so you can see the code while you read it. That's kinda what I did while I wrote it. :)

Line 01: This imports numpy, which is a linear algebra library. This is our only dependency.

Line 04: This is our "nonlinearity". While it can be several kinds of functions, this nonlinearity maps a function called a "sigmoid". A sigmoid function maps any value to a value between 0 and 1. We use it to convert numbers to probabilities. It also has several other desirable properties for training neural networks.

Line 05: Notice that this function can also generate the derivative of a sigmoid (when deriv=True). One of the desirable properties of a sigmoid function is that its output can be used to create its derivative. If the sigmoid's output is a variable "out", then the derivative is simply out * (1-out). This is very efficient. 

If you're unfamililar with derivatives, just think about it as the slope of the sigmoid function at a given point (as you can see above, different points have different slopes). For more on derivatives, check out this derivatives tutorialfrom Khan Academy.

Line 10: This initializes our input dataset as a numpy matrix. Each row is a single "training example". Each column corresponds to one of our input nodes. Thus, we have 3 input nodes to the network and 4 training examples.

Line 16: This initializes our output dataset. In this case, I generated the dataset horizontally (with a single row and 4 columns) for space. ".T" is the transpose function. After the transpose, this y matrix has 4 rows with one column. Just like our input, each row is a training example, and each column (only one) is an output node. So, our network has 3 inputs and 1 output.

Line 20: It's good practice to seed your random numbers. Your numbers will still be randomly distributed, but they'll be randomly distributed in exactly the same way each time you train. This makes it easier to see how your changes affect the network.

Line 23: This is our weight matrix for this neural network. It's called "syn0" to imply "synapse zero". Since we only have 2 layers (input and output), we only need one matrix of weights to connect them. Its dimension is (3,1) because we have 3 inputs and 1 output. Another way of looking at it is that l0 is of size 3 and l1 is of size 1. Thus, we want to connect every node in l0 to every node in l1, which requires a matrix of dimensionality (3,1). :) 

Also notice that it is initialized randomly with a mean of zero. There is quite a bit of theory that goes into weight initialization. For now, just take it as a best practice that it's a good idea to have a mean of zero in weight initialization. 

Another note is that the "neural network" is really just this matrix. We have "layers" l0 and l1 but they are transient values based on the dataset. We don't save them. All of the learning is stored in the syn0 matrix.

Line 25: This begins our actual network training code. This for loop "iterates" multiple times over the training code to optimize our network to the dataset.

Line 28: Since our first layer, l0, is simply our data. We explicitly describe it as such at this point. Remember that X contains 4 training examples (rows). We're going to process all of them at the same time in this implementation. This is known as "full batch" training. Thus, we have 4 different l0 rows, but you can think of it as a single training example if you want. It makes no difference at this point. (We could load in 1000 or 10,000 if we wanted to without changing any of the code).

Line 29: This is our prediction step. Basically, we first let the network "try" to predict the output given the input. We will then study how it performs so that we can adjust it to do a bit better for each iteration. 

This line contains 2 steps. The first matrix multiplies l0 by syn0. The second passes our output through the sigmoid function. Consider the dimensions of each:

(4 x 3) dot (3 x 1) = (4 x 1) 

Matrix multiplication is ordered, such the dimensions in the middle of the equation must be the same. The final matrix generated is thus the number of rows of the first matrix and the number of columns of the second matrix.

Since we loaded in 4 training examples, we ended up with 4 guesses for the correct answer, a (4 x 1) matrix. Each output corresponds with the network's guess for a given input. Perhaps it becomes intuitive why we could have "loaded in" an arbitrary number of training examples. The matrix multiplication would still work out. :)

Line 32: So, given that l1 had a "guess" for each input. We can now compare how well it did by subtracting the true answer (y) from the guess (l1). l1_error is just a vector of positive and negative numbers reflecting how much the network missed.

Line 36: Now we're getting to the good stuff! This is the secret sauce! There's a lot going on in this line, so let's further break it into two parts.

First Part: The Derivative


If l1 represents these three dots, the code above generates the slopes of the lines below. Notice that very high values such as x=2.0 (green dot) and very low values such as x=-1.0 (purple dot) have rather shallow slopes. The highest slope you can have is at x=0 (blue dot). This plays an important role. Also notice that all derivatives are between 0 and 1.

Entire Statement: The Error Weighted Derivative

1.l1_delta = l1_error * nonlin(l1,True)

There are more "mathematically precise" ways than "The Error Weighted Derivative" but I think that this captures the intuition. l1_error is a (4,1) matrix. nonlin(l1,True) returns a (4,1) matrix. What we're doing is multiplying them "elementwise". This returns a (4,1) matrix l1_delta with the multiplied values. 

When we multiply the "slopes" by the error, we are reducing the error of high confidence predictions. Look at the sigmoid picture again! If the slope was really shallow (close to 0), then the network either had a very high value, or a very low value. This means that the network was quite confident one way or the other. However, if the network guessed something close to (x=0, y=0.5) then it isn't very confident. We update these "wishy-washy" predictions most heavily, and we tend to leave the confident ones alone by multiplying them by a number close to 0.

Line 39: We are now ready to update our network! Let's take a look at a single training example.In this training example, we're all setup to update our weights. Let's update the far left weight (9.5).

weight_update = input_value * l1_delta 

For the far left weight, this would multiply 1.0 * the l1_delta. Presumably, this would increment 9.5 ever so slightly. Why only a small ammount? Well, the prediction was already very confident, and the prediction was largely correct. A small error and a small slope means a VERY small update. Consider all the weights. It would ever so slightly increase all three.

However, because we're using a "full batch" configuration, we're doing the above step on all four training examples. So, it looks a lot more like the image above. So, what does line 39 do? It computes the weight updates for each weight for each training example, sums them, and updates the weights, all in a simple line. Play around with the matrix multiplication and you'll see it do this!


So, now that we've looked at how the network updates, let's look back at our training data and reflect. When both an input and a output are 1, we increase the weight between them. When an input is 1 and an output is 0, we decrease the weight between them.


Thus, in our four training examples below, the weight from the first input to the output would consistently increment or remain unchanged, whereas the other two weights would find themselves both increasing and decreasing across training examples (cancelling out progress). This phenomenon is what causes our network to learn based on correlations between the input and output.

Part 2: A Slightly Harder Problem


Consider trying to predict the output column given the two input columns. A key takeway should be that neither columns have any correlation to the output. Each column has a 50% chance of predicting a 1 and a 50% chance of predicting a 0.

So, what's the pattern? It appears to be completely unrelated to column three, which is always 1. However, columns 1 and 2 give more clarity. If either column 1 or 2 are a 1 (but not both!) then the output is a 1. This is our pattern.

This is considered a "nonlinear" pattern because there isn't a direct one-to-one relationship between the input and output. Instead, there is a one-to-one relationship between a combination of inputs, namely columns 1 and 2.

Believe it or not, image recognition is a similar problem. If one had 100 identically sized images of pipes and bicycles, no individual pixel position would directly correlate with the presence of a bicycle or pipe. The pixels might as well be random from a purely statistical point of view. However, certain combinations of pixels are not random, namely the combination that forms the image of a bicycle or a person.

Our Strategy

In order to first combine pixels into something that can then have a one-to-one relationship with the output, we need to add another layer. Our first layer will combine the inputs, and our second layer will then map them to the output using the output of the first layer as input. Before we jump into an implementation though, take a look at this table.

Inputs (l0)Hidden Weights (l1)Output (l2)

If we randomly initialize our weights, we will get hidden state values for layer 1. Notice anything? The second column (second hidden node), has a slight correlation with the output already! It's not perfect, but it's there. Believe it or not, this is a huge part of how neural networks train. (Arguably, it's the only way that neural networks train.) What the training below is going to do is amplify that correlation. It's both going to update syn1 to map it to the output, and update syn0 to be better at producing it from the input!

Note: The field of adding more layers to model more combinations of relationships such as this is known as "deep learning" because of the increasingly deep layers being modeled.

3 Layer Neural Network:

01.import numpy as np
03.def nonlin(x,deriv=False):
05.return x*(1-x)
07.return 1/(1+np.exp(-x))
09.= np.array([[0,0,1],
14.= np.array([[0],
21.# randomly initialize our weights with mean 0
22.syn0 = 2*np.random.random((3,4)) - 1
23.syn1 = 2*np.random.random((4,1)) - 1
25.for in xrange(60000):
27.# Feed forward through layers 0, 1, and 2
28.l0 = X
29.l1 = nonlin(,syn0))
30.l2 = nonlin(,syn1))
32.# how much did we miss the target value?
33.l2_error = - l2
35.if (j% 10000== 0:
36.print "Error:" + str(np.mean(np.abs(l2_error)))
38.# in what direction is the target value?
39.# were we really sure? if so, don't change too much.
40.l2_delta = l2_error*nonlin(l2,deriv=True)
42.# how much did each l1 value contribute to the l2 error (according to the weights)?
43.l1_error =
45.# in what direction is the target l1?
46.# were we really sure? if so, don't change too much.
47.l1_delta = l1_error * nonlin(l1,deriv=True)
49.syn1 +=
50.syn0 +=
XInput dataset matrix where each row is a training example
yOutput dataset matrix where each row is a training example
l0First Layer of the Network, specified by the input data
l1Second Layer of the Network, otherwise known as the hidden layer
l2Final Layer of the Network, which is our hypothesis, and should approximate the correct answer as we train.
syn0First layer of weights, Synapse 0, connecting l0 to l1.
syn1Second layer of weights, Synapse 1 connecting l1 to l2.
l2_errorThis is the amount that the neural network "missed".
l2_deltaThis is the error of the network scaled by the confidence. It's almost identical to the error except that very confident errors are muted.
l1_errorWeighting l2_delta by the weights in syn1, we can calculate the error in the middle/hidden layer.
l1_deltaThis is the l1 error of the network scaled by the confidence. Again, it's almost identical to the l1_error except that confident errors are muted.

Recommendation: open this blog in two screens so you can see the code while you read it. That's kinda what I did while I wrote it. :)

Everything should look very familiar! It's really just 2 of the previous implementation stacked on top of each other. The output of the first layer (l1) is the input to the second layer. The only new thing happening here is on line 43.

Line 43: uses the "confidence weighted error" from l2 to establish an error for l1. To do this, it simply sends the error across the weights from l2 to l1. This gives what you could call a "contribution weighted error" because we learn how much each node value in l1 "contributed" to the error in l2. This step is called "backpropagating" and is the namesake of the algorithm. We then update syn0 using the same steps we did in the 2 layer implementation.

Part 3: Conclusion and Future Work

My Recommendation:

If you're serious about neural networks, I have one recommendation. Try to rebuild this network from memory. I know that might sound a bit crazy, but it seriously helps. If you want to be able to create arbitrary architectures based on new academic papers or read and understand sample code for these different architectures, I think that it's a killer exercise. I think it's useful even if you're using frameworks like TorchCaffe, or Theano. I worked with neural networks for a couple years before performing this exercise, and it was the best investment of time I've made in the field (and it didn't take long).

Future Work

This toy example still needs quite a few bells and whistles to really approach the state-of-the-art architectures. Here's a few things you can look into if you want to further improve your network. (Perhaps I will in a followup post.)

• Alpha 
• Bias Units
• Mini-Batches
• Delta Trimming 
• Parameterized Layer Sizes
• Regularization
• Dropout
• Momentum
• Batch Normalization 
• GPU Compatability
• Other Awesomeness You Implement

Posted by steloflute


An Intuitive Explanation of Convolutional Neural Networks

What are Convolutional Neural Networks and why are they important?

Convolutional Neural Networks (ConvNets or CNNs) are a category of Neural Networks that have proven very effective in areas such as image recognition and classification. ConvNets have been successful in identifying faces, objects and traffic signs apart from powering vision in robots and self driving cars.

convnets use.png

Figure 1: Source [1]

In Figure 1 above, a ConvNet is able to recognize scenes and the system is able to suggest relevant tags such as ‘bridge’, ‘railway’ and ‘tennis’ while Figure 2 shows an example of ConvNets being used for recognizing everyday objects, humans and animals. Lately, ConvNets have been effective in several Natural Language Processing tasks (such as sentence classification) as well.

Screen Shot 2016-08-07 at 4.17.11 PM.png

Figure 2: Source [2]

ConvNets, therefore, are an important tool for most machine learning practitioners today. However, understanding ConvNets and learning to use them for the first time can sometimes be an intimidating experience. The primary purpose of this blog post is to develop an understanding of how Convolutional Neural Networks work on images.

If you are new to neural networks in general, I would recommend reading this short tutorial on Multi Layer Perceptrons to get an idea about how they work, before proceeding. Multi Layer Perceptrons are referred to as “Fully Connected Layers” in this post.

The LeNet Architecture (1990s)

LeNet was one of the very first convolutional neural networks which helped propel the field of Deep Learning. This pioneering work by Yann LeCun was named LeNet5 after many previous successful iterations since the year 1988 [3]. At that time the LeNet architecture was used mainly for character recognition tasks such as reading zip codes, digits, etc.

Below, we will develop an intuition of how the LeNet architecture learns to recognize images. There have been several new architectures proposed in the recent years which are improvements over the LeNet, but they all use the main concepts from the LeNet and are relatively easier to understand if you have a clear understanding of the former.

Screen Shot 2016-08-07 at 4.59.29 PM.png

Figure 3: A simple ConvNet. Source [5]

The Convolutional Neural Network in Figure 3 is similar in architecture to the original LeNet and classifies an input image into four categories: dog, cat, boat or bird (the original LeNet was used mainly for character recognition tasks). As evident from the figure above, on receiving a boat image as input, the network correctly assigns the highest probability for boat (0.94) among all four categories. The sum of all probabilities in the output layer should be one (explained later in this post).

There are four main operations in the ConvNet shown in Figure 3 above:

  1. Convolution
  2. Non Linearity (ReLU)
  3. Pooling or Sub Sampling
  4. Classification (Fully Connected Layer)

These operations are the basic building blocks of every Convolutional Neural Network, so understanding how these work is an important step to developing a sound understanding of ConvNets. We will try to understand the intuition behind each of these operations below.

Images are a matrix of pixel values

Essentially, every image can be represented as a matrix of pixel values.


Figure 4: Every image is a matrix of pixel values. Source [6]

Channel is a conventional term used to refer to a certain component of an image. An image from a standard digital camera will have three channels – red, green and blue – you can imagine those as three 2d-matrices stacked over each other (one for each color), each having pixel values in the range 0 to 255.

grayscale image, on the other hand, has just one channel. For the purpose of this post, we will only consider grayscale images, so we will have a single 2d matrix representing an image. The value of each pixel in the matrix will range from 0 to 255 – zero indicating black and 255 indicating white.

The Convolution Step

ConvNets derive their name from the “convolution” operator. The primary purpose of Convolution in case of a ConvNet is to extract features from the input image. Convolution preserves the spatial relationship between pixels by learning image features using small squares of input data. We will not go into the mathematical details of Convolution here, but will try to understand how it works over images.

As we discussed above, every image can be considered as a matrix of pixel values. Consider a 5 x 5 image whose pixel values are only 0 and 1 (note that for a grayscale image, pixel values range from 0 to 255, the green matrix below is a special case where pixel values are only 0 and 1):

Screen Shot 2016-07-24 at 11.25.13 PM

Also, consider another 3 x 3 matrix as shown below:

Screen Shot 2016-07-24 at 11.25.24 PM

Then, the Convolution of the 5 x 5 image and the 3 x 3 matrix can be computed as shown in the animation in Figure 5 below:Convolution_schematic

Figure 5: The Convolution operation. The output matrix is called Convolved Feature or Feature Map. Source [7]

Take a moment to understand how the computation above is being done. We slide the orange matrix over our original image (green) by 1 pixel (also called ‘stride’) and for every position, we compute element wise multiplication (between the two matrices) and add the multiplication outputs to get the final integer which forms a single element of the output matrix (pink). Note that the 3×3 matrix “sees” only a part of the input image in each stride.

In CNN terminology, the 3×3 matrix is called a ‘filter‘ or ‘kernel’ or ‘feature detector’ and the matrix formed by sliding the filter over the image and computing the dot product is called the ‘Convolved Feature’ or ‘Activation Map’ or the ‘Feature Map‘. It is important to note that filters acts as feature detectors from the original input image.

It is evident from the animation above that different values of the filter matrix will produce different Feature Maps for the same input image. As an example, consider the following input image:


In the table below, we can see the effects of convolution of the above image with different filters. As shown, we can perform operations such as Edge Detection, Sharpen and Blur just by changing the numeric values of our filter matrix before the convolution operation [8] – this means that different filters can detect different features from an image, for example edges, curves etc. More such examples are available in Section 8.2.4 here.

Screen Shot 2016-08-05 at 11.03.00 PM.png

Another good way to understand the Convolution operation is by looking at the animation in Figure 6 below:


Figure 6: The Convolution Operation. Source [9]

A filter (with red outline) slides over the input image (convolution operation) to produce a feature map. The convolution of another filter (with the green outline), over the same image gives a different feature map as shown. It is important to note that the Convolution operation captures the local dependencies in the original image. Also notice how these two different filters generate different feature maps from the same original image. Remember that the image and the two filters above are just numeric matrices as we have discussed above.

In practice, a CNN learns the values of these filters on its own during the training process (although we still need to specify parameters such as number of filtersfilter sizearchitecture of the network etc. before the training process). The more number of filters we have, the more image features get extracted and the better our network becomes at recognizing patterns in unseen images.

The size of the Feature Map (Convolved Feature) is controlled by three parameters [4] that we need to decide before the convolution step is performed:

  • Depth: Depth corresponds to the number of filters we use for the convolution operation. In the network shown in Figure 7, we are performing convolution of the original boat image using three distinct filters, thus producing three different feature maps as shown. You can think of these three feature maps as stacked 2d matrices, so, the ‘depth’ of the feature map would be three.

Screen Shot 2016-08-10 at 3.42.35 AM

Figure 7
  • Stride: Stride is the number of pixels by which we slide our filter matrix over the input matrix. When the stride is 1 then we move the filters one pixel at a time. When the stride is 2, then the filters jump 2 pixels at a time as we slide them around. Having a larger stride will produce smaller feature maps.
  • Zero-padding: Sometimes, it is convenient to pad the input matrix with zeros around the border, so that we can apply the filter to bordering elements of our input image matrix. A nice feature of zero padding is that it allows us to control the size of the feature maps. Adding zero-padding is also called wide convolution, and not using zero-padding would be a narrow convolution. This has been explained clearly in [14].

Introducing Non Linearity (ReLU)

An additional operation called ReLU has been used after every Convolution operation in Figure 3 above. ReLU stands for Rectified Linear Unit and is a non-linear operation. Its output is given by:

Screen Shot 2016-08-10 at 2.23.48 AM.png

Figure 8: the ReLU operation

ReLU is an element wise operation (applied per pixel) and replaces all negative pixel values in the feature map by zero. The purpose of ReLU is to introduce non-linearity in our ConvNet, since most of the real-world data we would want our ConvNet to learn would be non-linear (Convolution is a linear operation – element wise matrix multiplication and addition, so we account for non-linearity by introducing a non-linear function like ReLU).

The ReLU operation can be understood clearly from Figure 9 below. It shows the ReLU operation applied to one of the feature maps obtained in Figure 6 above. The output feature map here is also referred to as the ‘Rectified’ feature map.


Screen Shot 2016-08-07 at 6.18.19 PM.png

Figure 9: ReLU operation. Source [10]

Other non linear functions such as tanh or sigmoid can also be used instead of ReLU, but ReLU has been found to perform better in most situations.

The Pooling Step

Spatial Pooling (also called subsampling or downsampling) reduces the dimensionality of each feature map but retains the most important information. Spatial Pooling can be of different types: Max, Average, Sum etc.

In case of Max Pooling, we define a spatial neighborhood (for example, a 2×2 window) and take the largest element from the rectified feature map within that window. Instead of taking the largest element we could also take the average (Average Pooling) or sum of all elements in that window. In practice, Max Pooling has been shown to work better.

Figure 10 shows an example of Max Pooling operation on a Rectified Feature map (obtained after convolution + ReLU operation) by using a 2×2 window.

Screen Shot 2016-08-10 at 3.38.39 AM.png

Figure 10: Max Pooling. Source [4]

We slide our 2 x 2 window by 2 cells (also called ‘stride’) and take the maximum value in each region. As shown in Figure 10, this reduces the dimensionality of our feature map.

In the network shown in Figure 11, pooling operation is applied separately to each feature map (notice that, due to this, we get three output maps from three input maps).

Screen Shot 2016-08-07 at 6.19.37 PM.png

Figure 11: Pooling applied to Rectified Feature Maps

Figure 12 shows the effect of Pooling on the Rectified Feature Map we received after the ReLU operation in Figure 9 above.

Screen Shot 2016-08-07 at 6.11.53 PM.png

Figure 12: Pooling. Source [10]

The function of Pooling is to progressively reduce the spatial size of the input representation [4]. In particular, pooling

  • makes the input representations (feature dimension) smaller and more manageable
  • reduces the number of parameters and computations in the network, therefore, controlling overfitting [4]
  • makes the network invariant to small transformations, distortions and translations in the input image (a small distortion in input will not change the output of Pooling – since we take the maximum / average value in a local neighborhood).
  • helps us arrive at an almost scale invariant representation of our image (the exact term is “equivariant”). This is very powerful since we can detect objects in an image no matter where they are located (read [18] and [19] for details).

Story so far

Screen Shot 2016-08-08 at 2.26.09 AM.png

Figure 13

So far we have seen how Convolution, ReLU and Pooling work. It is important to understand that these layers are the basic building blocks of any CNN. As shown in Figure 13, we have two sets of Convolution, ReLU & Pooling layers – the 2nd Convolution layer performs convolution on the output of the first Pooling Layer using six filters to produce a total of six feature maps. ReLU is then applied individually on all of these six feature maps. We then perform Max Pooling operation separately on each of the six rectified feature maps.

Together these layers extract the useful features from the images, introduce non-linearity in our network and reduce feature dimension while aiming to make the features somewhat equivariant to scale and translation [18].

The output of the 2nd Pooling Layer acts as an input to the Fully Connected Layer, which we will discuss in the next section.

Fully Connected Layer

The Fully Connected layer is a traditional Multi Layer Perceptron that uses a softmax activation function in the output layer (other classifiers like SVM can also be used, but will stick to softmax in this post). The term “Fully Connected” implies that every neuron in the previous layer is connected to every neuron on the next layer. I recommend reading this post if you are unfamiliar with Multi Layer Perceptrons.

The output from the convolutional and pooling layers represent high-level features of the input image. The purpose of the Fully Connected layer is to use these features for classifying the input image into various classes based on the training dataset. For example, the image classification task we set out to perform has four possible outputs as shown in Figure 14 below (note that Figure 14 does not show connections between the nodes in the fully connected layer)

Screen Shot 2016-08-06 at 12.34.02 AM.png

Figure 14: Fully Connected Layer -each node is connected to every other node in the adjacent layer

Apart from classification, adding a fully-connected layer is also a (usually) cheap way of learning non-linear combinations of these features. Most of the features from convolutional and pooling layers may be good for the classification task, but combinations of those features might be even better [11].

The sum of output probabilities from the Fully Connected Layer is 1. This is ensured by using the Softmax as the activation function in the output layer of the Fully Connected Layer. The Softmax function takes a vector of arbitrary real-valued scores and squashes it to a vector of values between zero and one that sum to one.

Putting it all together – Training using Backpropagation

As discussed above, the Convolution + Pooling layers act as Feature Extractors from the input image while Fully Connected layer acts as a classifier.

Note that in Figure 15 below, since the input image is a boat, the target probability is 1 for Boat class and 0 for other three classes, i.e.

  • Input Image = Boat
  • Target Vector = [0, 0, 1, 0]

Screen Shot 2016-08-07 at 9.15.21 PM.png

Figure 15: Training the ConvNet

The overall training process of the Convolution Network may be summarized as below:

  • Step1: We initialize all filters and parameters / weights with random values
  • Step2: The network takes a training image as input, goes through the forward propagation step (convolution, ReLU and pooling operations along with forward propagation in the Fully Connected layer) and finds the output probabilities for each class.
    • Lets say the output probabilities for the boat image above are [0.2, 0.4, 0.1, 0.3]
    • Since weights are randomly assigned for the first training example, output probabilities are also random.
  • Step3: Calculate the total error at the output layer (summation over all 4 classes)
    •  Total Error = ∑  ½ (target probability – output probability) ²
  • Step4: Use Backpropagation to calculate the gradients of the error with respect to all weights in the network and use gradient descent to update all filter values / weights and parameter values to minimize the output error.
    • The weights are adjusted in proportion to their contribution to the total error.
    • When the same image is input again, output probabilities might now be [0.1, 0.1, 0.7, 0.1], which is closer to the target vector [0, 0, 1, 0].
    • This means that the network has learnt to classify this particular image correctly by adjusting its weights / filters such that the output error is reduced.
    • Parameters like number of filters, filter sizes, architecture of the network etc. have all been fixed before Step 1 and do not change during training process – only the values of the filter matrix and connection weights get updated.
  • Step5: Repeat steps 2-4 with all images in the training set.

The above steps train the ConvNet – this essentially means that all the weights and parameters of the ConvNet have now been optimized to correctly classify images from the training set.

When a new (unseen) image is input into the ConvNet, the network would go through the forward propagation step and output a probability for each class (for a new image, the output probabilities are calculated using the weights which have been optimized to correctly classify all the previous training examples). If our training set is large enough, the network will (hopefully) generalize well to new images and classify them into correct categories.

Note 1: The steps above have been oversimplified and mathematical details have been avoided to provide intuition into the training process. See [4] and [12] for a mathematical formulation and thorough understanding.

Note 2: In the example above we used two sets of alternating Convolution and Pooling layers. Please note however, that these operations can be repeated any number of times in a single ConvNet. In fact, some of the best performing ConvNets today have tens of Convolution and Pooling layers! Also, it is not necessary to have a Pooling layer after every Convolutional Layer. As can be seen in the Figure 16 below, we can have multiple Convolution + ReLU operations in succession before having a Pooling operation. Also notice how each layer of the ConvNet is visualized in the Figure 16 below.


Figure 16: Source [4]

Visualizing Convolutional Neural Networks

In general, the more convolution steps we have, the more complicated features our network will be able to learn to recognize. For example, in Image Classification a ConvNet may learn to detect edges from raw pixels in the first layer, then use the edges to detect simple shapes in the second layer, and then use these shapes to deter higher-level features, such as facial shapes in higher layers [14]. This is demonstrated in Figure 17 below – these features were learnt using a Convolutional Deep Belief Network and the figure is included here just for demonstrating the idea (this is only an example: real life convolution filters may detect objects that have no meaning to humans).

Screen Shot 2016-08-10 at 12.58.30 PM.png

Figure 17: Learned features from a Convolutional Deep Belief Network

Adam Harley created amazing visualizations of a Convolutional Neural Network trained on the MNIST Database of handwritten digits [13]. I highly recommend playing around with it to understand details of how a CNN works.

We will see below how the network works for an input ‘8’. Note that the visualization in Figure 18 does not show the ReLU operation separately.


Figure 18: Visualizing a ConvNet trained on handwritten digits

The input image contains 1024 pixels (32 x 32 image) and the first Convolution layer (Convolution Layer 1) is formed by convolution of six unique 5 × 5 (stride 1) filters with the input image. As seen, using six different filters produces a feature map of depth six.

Convolutional Layer 1 is followed by Pooling Layer 1 that does 2 × 2 max pooling (with stride 2) separately over the six feature maps in Convolution Layer 1. You can move your mouse pointer over any pixel in the Pooling Layer and observe the 4 x 4 grid it forms in the previous Convolution Layer (demonstrated in Figure 19). You’ll notice that the pixel having the maximum value (the brightest one) in the 4 x 4 grid makes it to the Pooling layer.

Screen Shot 2016-08-06 at 12.45.35 PM.png

Figure 19: Visualizing the Pooling Operation

Pooling Layer 1 is followed by sixteen 5 × 5 (stride 1) convolutional filters that perform the convolution operation. This is followed by Pooling Layer 2 that does 2 × 2 max pooling (with stride 2). These two layers use the same concepts as described above.

We then have three fully-connected (FC) layers. There are:

  • 120 neurons in the first FC layer
  • 100 neurons in the second FC layer
  • 10 neurons in the third FC layer corresponding to the 10 digits – also called the Output layer

Notice how in Figure 20, each of the 10 nodes in the output layer are connected to all 100 nodes in the 2nd Fully Connected layer (hence the name Fully Connected).

Also, note how the only bright node in the Output Layer corresponds to ‘8’ – this means that the network correctly classifies our handwritten digit (brighter node denotes that the output from it is higher, i.e. 8 has the highest probability among all other digits).


Figure 20: Visualizing the Filly Connected Layers

The 3d version of the same visualization is available here.

Other ConvNet Architectures

Convolutional Neural Networks have been around since early 1990s. We discussed the LeNet above which was one of the very first convolutional neural networks. Some other influential architectures are listed below [3] [4].

  • LeNet (1990s): Already covered in this article.
  • 1990s to 2012: In the years from late 1990s to early 2010s convolutional neural network were in incubation. As more and more data and computing power became available, tasks that convolutional neural networks could tackle became more and more interesting.
  • AlexNet (2012) – In 2012, Alex Krizhevsky (and others) released AlexNet which was a deeper and much wider version of the LeNet and won by a large margin the difficult ImageNet Large Scale Visual Recognition Challenge (ILSVRC) in 2012. It was a significant breakthrough with respect to the previous approaches and the current widespread application of CNNs can be attributed to this work.
  • ZF Net (2013) – The ILSVRC 2013 winner was a Convolutional Network from Matthew Zeiler and Rob Fergus. It became known as the ZFNet (short for Zeiler & Fergus Net). It was an improvement on AlexNet by tweaking the architecture hyperparameters.
  • GoogLeNet (2014) – The ILSVRC 2014 winner was a Convolutional Network from Szegedy et al. from Google. Its main contribution was the development of an Inception Module that dramatically reduced the number of parameters in the network (4M, compared to AlexNet with 60M).
  • VGGNet (2014) – The runner-up in ILSVRC 2014 was the network that became known as the VGGNet. Its main contribution was in showing that the depth of the network (number of layers) is a critical component for good performance.
  • ResNets (2015) – Residual Network developed by Kaiming He (and others) was the winner of ILSVRC 2015. ResNets are currently by far state of the art Convolutional Neural Network models and are the default choice for using ConvNets in practice (as of May 2016).
  • DenseNet (August 2016) – Recently published by Gao Huang (and others), the Densely Connected Convolutional Network has each layer directly connected to every other layer in a feed-forward fashion. The DenseNet has been shown to obtain significant improvements over previous state-of-the-art architectures on five highly competitive object recognition benchmark tasks. Check out the Torch implementation here.


In this post, I have tried to explain the main concepts behind Convolutional Neural Networks in simple terms. There are several details I have oversimplified / skipped, but hopefully this post gave you some intuition around how they work.

This post was originally inspired from Understanding Convolutional Neural Networks for NLP by Denny Britz (which I would recommend reading) and a number of explanations here are based on that post. For a more thorough understanding of some of these concepts, I would encourage you to go through the notes from Stanford’s course on ConvNets as well as other excellent resources mentioned under References below. If you face any issues understanding any of the above concepts or have questions / suggestions, feel free to leave a comment below.

All images and animations used in this post belong to their respective authors as listed in References section below.


  1. Clarifai Home Page
  2. Shaoqing Ren, et al, “Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks”, 2015, arXiv:1506.01497 
  3. Neural Network Architectures, Eugenio Culurciello’s blog
  4. CS231n Convolutional Neural Networks for Visual Recognition, Stanford
  5. Clarifai / Technology
  6. Machine Learning is Fun! Part 3: Deep Learning and Convolutional Neural Networks
  7. Feature extraction using convolution, Stanford
  8. Wikipedia article on Kernel (image processing) 
  9. Deep Learning Methods for Vision, CVPR 2012 Tutorial 
  10. Neural Networks by Rob Fergus, Machine Learning Summer School 2015
  11. What do the fully connected layers do in CNNs? 
  12. Convolutional Neural Networks, Andrew Gibiansky 
  13. A. W. Harley, “An Interactive Node-Link Visualization of Convolutional Neural Networks,” in ISVC, pages 867-877, 2015 (link)
  14. Understanding Convolutional Neural Networks for NLP
  15. Backpropagation in Convolutional Neural Networks
  16. A Beginner’s Guide To Understanding Convolutional Neural Networks
  17. Vincent Dumoulin, et al, “A guide to convolution arithmetic for deep learning”, 2015, arXiv:1603.07285
  18. What is the difference between deep learning and usual machine learning?
  19. How is a convolutional neural network able to learn invariant features?
  20. A Taxonomy of Deep Convolutional Neural Nets for Computer Vision

Posted by steloflute

Generate bitcoin for me

What's this?