I would like to know a bit more about Neural Network, I'm developing a C++ program to make a NN but I'm stuck with the BackPropagation algorithm, sorry for not offering some working code.
I know that there are so many libraries for creating a NN in many languages, but I prefer to make one from my self. The point is that I don't know how many layers and how many neurons should be necessary for achieving a particular goal such as pattern recognition, or functions approximations, or whatever.
My questions are: if I'd like to recognize some particulars patterns, like in image detection, how many layers and neurons-per-layer should be necessary? Let's say my images are all 8x8 pixels, I would start naturally with an input layer of 64 neurons, but I don't have any idea of how many neurons I have to put in hidden layers, and also in output layer. Let's say I have to distinguish from cats and dogs, or whatever you may think, how could be the output layer? I can imagine an output layer with only-one neuron outputting a value between 0 and 1 with the classical logistic function (1/(1+exp(-x)) and when it is near 0 the input was a cat and when approaches 1 it was a dog, but ... is it correct? What if I add a new pattern like a fish? and what if the input contains a dog and a cat ( ..and a fish)? This make me thinking that the logistic function in the output layer is not very suitable for pattern recognition like this, only because 1/(1+exp(-x)) has a range in (0,1). Do I have to change the activation function or maybe add some other neurons to the output layer? Are there some other activations function more accurate to do this? Do every neurons in every layers have the same activation function, or it is different from layer to layer?
Sorry for all of this questions, but this topic is not very clear to me.
I read a lot around internet, and I found libraries all-yet-implemented and hard to read from, and many explanations to what a NN can do, but not how it can do.
I read a lot from https://mattmazur.com/2015/03/17/a-step-by-step-backpropagation-example/ and http://neuralnetworksanddeeplearning.com/chap1.html, and here I understood how to approximate a function (because every neurons in a layer can be thought as a step-function with a particular step for weights and bias) and how back-propagation algorithm works, but other tutorials and similars were more focused on preexisting libraries. I also read this question Determining the proper amount of Neurons for a Neural Network but I would like to involve also the activation functions of a NN, which is the best and for what is the best.
Thanks in advance for your answers!
Your questions are quite general, so I can only give some general recommendations:
The number of layers you need depends on the complexity of the problem you want to solve. The more calculation is required to obtain an output from a given input, the more layers you need.
Only very simple problems can be solved with a single layer network. These are called linearly separable and are usually trivial. With two layers it gets better and with three layers, at least in theory, all kinds of classification tasks can be performed if you have enough cells within the layers. In practice, however it is often better to add a 4th or 5th layer to the network while reducing the number of cells within a single layer.
Be aware that the standard backpropagation algorithm performs badly with more than 4 or 5 layers. If you need more layers, have a look at Deep Learning.
The numbers of cells within each layer mainly depends on the number of inputs and, if you solve a classification task, the number of classes you want to detect. In practice it is quite common to reduce the number of cells from layer to layer, but there are exceptions.
Concerning your question about the output function: In most cases you should stick with one type of sigmoid function. The case you describe is not really an issue because you could add another output cell for your "fish" class. The choice of a specific activation function is not that critical. Basically you use one whose values and derivative can be calculated efficiently.
#Frank Puffer has already provided some nice information, but let me add my two cents. First off, much of what you're asking is in the area of hyperparameter optimization. Although there are various "rules of thumb", the reality is that determining the optimal architecture (number/size of layers, connectivity structure, etc.) and other parameters like the learning rate typically requires extensive experimentation. The good news is that the parameterization of these hyperparameters is among the simplest aspects of the implementation of a neural network. So I would recommend focusing on building your software such that the number of layers, size of layers, learning rate, etc., are all easily configurable.
Now you specifically asked about detecting patterns in an image. It's worth mentioning that using standard multi-layer perceptrons (MLPs) to perform classification on raw image data can be computationally expensive, especially for larger images. It's common to use architectures that are designed to extract useful, spacially-local features (i.e.: Convolutional Neural Networks or CNNs).
You could still use standard MLPs for this, but the computational complexity can make it an untenable solution. The sparse connectivity of CNNs for example dramatically reduce the number of parameters requiring optimization and simultaneously build a conceptual hierarchy of representations better suited for classification of images.
Regardless, I would recommend implementing backpropagation using stochastic gradient descent for optimization. This is still the approach typically used for training neural nets, CNNs, RNNs, etc.
Regarding the number of output neurons, this is one question that does have a simple answer: use "one-hot" encoding. For each class you want to recognize, you have an output neuron. In your example of the dog, cat, and fish classes, you have three neurons. For an input image representing a dog, you would expect a value of 1 for the "dog" neuron, and 0 for all the others. Then, during inference, you can interpret the output as a probability distribution reflecting the confidence of the NN. For example, if you get output dog:0.70, cat:0.25, fish:0.05, then you have a 70% confidence that the image is a dog, and so on.
For activation functions, the most recent research I've seen seems to indicate that Rectified Linear Units are generally a good choice since they're easy to differentiate and compute, and they avoid a problem that plagues deeper networks called the "vanishing gradient problem".
Best of luck!
Related
I am working on an implementation of the back propagation algorithm. What I have implemented so far seems working but I can't be sure that the algorithm is well implemented, here is what I have noticed during training test of my network :
Specification of the implementation :
A data set containing almost 100000 raw containing (3 variable as input, the sinus of the sum of those three variables as expected output).
The network does have 7 layers all the layers use the Sigmoid activation function
When I run the back propagation training process:
The minimum of costs of the error is found at the fourth iteration (The minimum cost of error is 140, is it normal? I was expecting much less than that)
After the fourth Iteration the costs of the error start increasing (I don't know if it is normal or not?)
The short answer would be "no, very likely your implementation is incorrect". Your network is not training as can be observed by the very high cost of error. As discussed in comments, your network suffers very heavily from vanishing gradient problem, which is inevitable in deep networks. In essence, the first layers of you network learn much slower than the later. All neurons get some random weights at the beginning, right? Since the first layer almost doesn't learn anything, the large initial error propagates through the whole network!
How to fix it? From the description of your problem it seems that a feedforward network with just a single hidden layer in should be able to do the trick (as proven in universal approximation theorem).
Check e.g. free online book by Michael Nielsen if you'd like to learn more.
so I do understand from that the back propagation can't deal with deep neural networks? or is there some method to prevent this problem?
It can, but it's by no mean a trivial challenge. Deep neural networks have been used since 60', but only in 90' researchers came up with methods how to deal with them efficiently. I recommend reading "Efficient BackProp" chapter (by Y.A. LeCun et al.) of "Neural Networks: Tricks of the Trade".
Here is the summary:
Shuffle the examples
Center the input variables by subtracting the mean
Normalize the input variable to a standard deviation of 1
If possible, decorrelate the input variables.
Pick a network with the sigmoid function f(x)=1.7159*(tanh(2/3x): it won't saturate at +1 / -1, but instead will have highest gain at these points (second derivative is at max.)
Set the target values within the range of the sigmoid, typically +1 and -1.
The weights should be randomly drawn from a distribution with mean zero and a standard deviation given by m^(-1/2), where m is the number of inputs to the unit
The preferred method for training the network should be picked as follows:
If the training set is large (more than a few hundred samples) and redundant, and if the task is classification, use stochastic gradient with careful tuning, or use the stochastic diagonal Levenberg Marquardt method.
If the training set is not too large, or if the task is regression, use conjugate gradient.
Also, some my general remarks:
Watch for numerical stability if you implement it yourself. It's easy to get into troubles.
Think of the architecture. Fully-connected multi-layer networks are rarely a smart idea. Unfortunately ANN are poorly understood from theoretical point of view and one of the best things you can do is just check what worked for others and learn useful patterns (with regularization, pooling and dropout layers and such).
I am using the Caffe framework for CNN training. My aim is to perform simple object recognition for a few basic object categories. Since pretrained networks are not an alternative for my proposed usage I prepared an own training- and testset with about 1000 images for each of 2 classes (say chairs and cars).
The results are quite good. If I present an yet unseen image of a chair it is likely classified as such, same for a car image. My problem is that the results on miscellaneous images that do not show any of these classes often shows a very high confidence (=1) for one random class (which is not surprising regarding the onesided training data but a problem for my application). I thought about different solutions:
1) Adding a third class with also about 1000 negative examples that shows any objects except a chair and a car.
2) Adding more object categories in general, just to let the network classify other objects as such and not any more as a chair or car (of course this would require much effort). Maybe also the broader prediction results would show a more uniform distribution at negative images, allowing to evaluate the target objects presence based on a threshold?
Because it was not much time-consuming to grab random images as negative examples from the internet, I already tested my first solution with about 1200 negative examples. It helped, but the problem remains, perhaps because it were just too few? My concern is that if I increment the number of negative examples, the imbalance of the number of examples for each class leads to less accurate detection of the original classes.
After some research I found one person with a similar problem, but there was no solution:
Convolutional Neural Networks with Caffe and NEGATIVE IMAGES
My question is: Has anyone had the same problem and knows how to deal with it? What way would you recommend, adding more negative examples or more object categories or do you have any other recommendation?
The problem is not unique to Caffe or ConvNets. Any Machine Learning technique runs this risk. In the end, all classifiers take a vector in some input space (usually very high-dimensional), which means they partition that input space. You've given examples of two partitions, which helps to estimate the boundary between the two, but only that boundary. Both partitions have very, very large boundaries, precisely because the input space is so high-dimensional.
ConvNets do try to tackle the high-dimensionality of image data by having fairly small convolution kernels. Realistic negative data helps in training those, and the label wouldn't really matter. You could even use the input image as goal (i.e. train it as an autoencoder) when training the convolution kernels.
One general reason why you don't want to lump all counterexamples is because they may be too varied. If you have a class A with some feature value from the range [-1,+1] on some scale, with counterexamples B [-2,-1] and C [+1,+2], lumping B and C together creates a range [-2,+2] for counterexamples which overlaps the real real range. Given enough data and powerful enough classifiers, this is not fatal, but for instance an SVM can fail badly on this.
I have a pairs of images (input-output) but I don't know the transformation to going from A (input) to B (output). I want to record image A and get image B. Physically I can change the setup to get A or B, but I want to do it by software.
If I understood well, a trained Artificial Neural Network is able to do that, having an input can give the corresponding output, is it right?
Is there any software/ANN that just "training" it with entering a number of input-output pairs will be able to provide the correct output if the input is a new (but similar to the others) image?
Thanks
If you have some relevant amount of image pairs (input/output pair) and you don't know transformation between input and output you could train ANN on that training set to imitate that unknown transformation. You will be able to well train your ANN only if you have sufficient amount of training image pairs, but it could be pretty impossible when that unknown transformation is complicated.
For example if that transformation simply increases intensity values of pixels at input image by given value, ANN will very fast learn to imitate that behavior, but if that unknown transformation is some complicated convolution or few serial convolutions or something more complicated it will be very hard, near impossible to train ANN to imitate that transformation. So, more complex transformation will need bigger training set and more complex ANN design.
There are plenty of free opensource ANN libraries implemented in many languages. You could start for example with that tutorial: http://www.codeproject.com/Articles/13091/Artificial-Neural-Networks-made-easy-with-the-FANN
What you are asking is possible in principle -- in theory, an ANN with sufficiently many hidden units can learn an arbitrary function to map inputs to outputs. However, as the comments and other answers have mentioned, there may be many technical issues with your particular problem that could make it impractical. I would classify these problems as (a) mapping complexity, (b) model complexity, (c) scaling complexity, and (d) implementation complexity. They are all somewhat related, but hopefully this is a useful way to break things down.
Mapping complexity
As mentioned by Springfield762, there are many possible functions that map from one image to another image. If the relationship between your input images and your output images is relatively simple -- like increasing the intensity of each pixel by a constant amount -- then an ANN would be able to learn this mapping without much difficulty. There are probably many more transformations that would be similarly easy to learn, such as skewing, flipping, rotating, or translating an image -- basically any affine transformation would be easy to learn. Other, nonlinear transformations could also be feasible, such as squaring the intensity of each pixel.
As a general rule, the more complicated the relationship between your input and output images, the more difficult it will be to get a model to learn this mapping for you.
Model complexity
The more complex the mapping from inputs to outputs, the more complex your ANN model will be to be able to capture this mapping. Models with many hidden layers have been shown in the past 10 years to perform quite well on tasks that people had previously thought impossible, but often these state-of-the-art models have millions or even billions of parameters and take weeks to train on GPU hardware. A simple model can capture many simple mappings, but if you have a complex input-output map to learn, you'll need a large, complex model.
Scaling complexity
Yves mentioned in the comments that it can be difficult to scale models up to typical image sizes. If your images are relatively small (currently the state of the art is to model images on the order of 100x100 pixels), then you can probably just throw a bunch of raw pixel data at an ANN model and see what happens. But if you're using 6000x4000 images from your shiny Nikon DSLR, it's going to be quite difficult to process those in a reasonable amount of time. You'd be better off compressing your image data somehow (PCA is a common technique) and then trying to learn the mapping in the compressed space.
In addition, larger images will have a larger space of possible mappings between them, so you'll need more of your larger images as training data than you would if you had small images.
Springfield762 also mentioned this: If the mapping between your input and output images is simple, then you'll only need a few examples to learn the mapping successfully. But if you have a complicated mapping, then you'll need much more training data to have a chance at learning the mapping properly.
Implementation complexity
It's unlikely that a tool already exists that would let you just throw image data into an ANN model and have a mapping appear. Most likely you'll need, at a minimum, to implement some code that will pre-process your image data. In addition, if you have lots of large images you'll probably need to write code to handle loading data from disk, etc. (There are a lot of "big data" tools for things like this, but they all require some amount of work to get set up.)
There are many, many open source ANN toolkits out there nowadays. FANN (already mentioned) is a popular one in C++ with bindings in other languages. Caffe is quite popular, and is also implemented in C++ with bindings. There seem to be many toolkits that use Python and Theano or some other GPU acceleration library -- Keras, Lasagne, Hebel, Pylearn2, neon, and Theanets (I wrote this one). Many people use Torch, written in Lua. Matlab has at least one neural network toolbox. I'm less familiar with other ecosystems, but Java seems to have Deeplearning4j, C# has Accord, and even R has darch.
But with any of these neural network toolkits, you're going to have to write some code to load the data, process it into the appropriate input format, construct (or load) a network model, train the model, etc.
The problem you're trying to solve is a canonical classification problem that neural networks can help you solve. You treat the B images as a set of labels that you match to A, and once trained, the neural network will be able to match the B images to new input based on where the network locates new input in a high-dimensional vector space. I assume you'd use some combination of convolutional networks to create your features, and softmax for multinomial classification on the output layer. More here: http://deeplearning4j.org/convolutionalnets.html
Since this has been written there has been a lot of work in the realm of cgans ( conditional generative adversarial networks ) please refer to:
https://arxiv.org/pdf/1611.07004.pdf
If I have a large set of data that describes physical 'things', how could I go about measuring how well that data fits the 'things' that it is supposed to represent?
An example would be if I have a crate holding 12 widgets, and I know each widget weighs 1 lb, there should be some data quality 'check' making sure the case weighs 13 lbs maybe.
Another example would be that if I have a lamp and an image representing that lamp, it should look like a lamp. Perhaps the image dimensions should have the same ratio of the lamp dimensions.
With the exception of images, my data is 99% text (which includes height, width, color...).
I've studied AI in school, but have done very little outside of that.
Are standard AI techniques the way to go? If so, how do I map a problem to an algorithm?
Are some languages easier at this than others? Do they have better libraries?
thanks.
Your question is somewhat open-ended, but it sounds like you want is what is known as a "classifier" in the field of machine learning.
In general, a classifier takes a piece of input and "classifies" it, ie: determines a category for the object. Many classifiers provide a probability with this determination, and some may even return multiple categories with probabilities on each.
Some examples of classifiers are bayes nets, neural nets, decision lists, and decision trees. Bayes nets are often used for spam classification. Emails are classified as either "spam" or "not spam" with a probability.
For you question you'd want to classify your objects as "high quality" or "not high quality".
The first thing you'll need is a bunch of training data. That is, a set of objects where you already know the correct classification. One way to obtain this could be to get a bunch of objects and classify them by hand. If there are too many objects for one person to classify you could feed them to Mechanical Turk.
Once you have your training data you'd then build your classifier. You'll need to figure out what attributes are important to your classification. You'll probably need to do some experimentation to see what works well. You then have your classifier learn from your training data.
One approach that's often used for testing is to split your training data into two sets. Train your classifier using one of the subsets, and then see how well it classifies the other (usually smaller) subset.
AI is one path, natural intelligence is another.
Your challenge is a perfect match to Amazon's Mechanical Turk. Divvy your data space up into extremely small verifiable atoms and assign them as HITs on Mechanical Turk. Have some overlap to give yourself a sense of HIT answer consistency.
There was a shop with a boatload of component CAD drawings that needed to be grouped by similarity. They broke it up and set it loose on Mechanical Turk to very satisfying results. I could google for hours and not find that link again.
See here for a related forum post.
This is a tough answer. For example, what defines a lamp? I could google images a picture of some crazy looking lamps. Or even, look up the definition of a lamp (http://dictionary.reference.com/dic?q=lamp). Theres no physical requirements of what a lamp must look like. Thats the crux of the AI problem.
As for data, you could setup Unit testing on the project to ensure that 12 widget() weighs less than 13 lbs in the widetBox(). Regardless, you need to have the data at hand to be able to test things like that.
I hope i was able to answer your question somewhat. Its a bit vauge, and my answers are broad, but hopefully it'll at least send you in a good direction.
Generally speaking what do you get out of extending an artificial neural net by adding more nodes to a hidden layer or more hidden layers?
Does it allow for more precision in the mapping, or does it allow for more subtlety in the relationships it can identify, or something else?
There's a very well known result in machine learning that states that a single hidden layer is enough to approximate any smooth, bounded function (the paper was called "Multilayer feedforward networks are universal approximators" and it's now almost 20 years old). There are several things to note, however.
The single hidden layer may need to be arbitrarily wide.
This says nothing about the ease with which an approximation may be found; in general large networks are hard to train properly and fall victim to overfitting quite frequently (the exception are so-called "convolutional neural networks" which really are only meant for vision problems).
This also says nothing about the efficiency of the representation. Some functions require exponential numbers of hidden units if done with one layer but scale much more nicely with more layers (for more discussion of this read Scaling Learning Algorithms Towards AI)
The problem with deep neural networks is that they're even harder to train. You end up with very very small gradients being backpropagated to the earlier hidden layers and the learning not really going anywhere, especially if weights are initialized to be small (if you initialize them to be of larger magnitude you frequently get stuck in bad local minima). There are some techniques for "pre-training" like the ones discussed in this Google tech talk by Geoff Hinton which attempt to get around this.
This is very interesting question but it's not so easy to answer. It depends on the problem you try to resolve and what neural network you try to use. There are several neural network types.
I general it's not so clear that more nodes equals more precision. Research show that you need mostly only one hidden layer. The numer of nodes should be the minimal numer of nodes that are required to resolve a problem. If you don't have enough of them - you will not reach solution.
From the other hand - if you have reached the number of nodes that is good to resolve solution - you can add more and more of them and you will not see any further progress in result estimation.
That's why there are so many types of neural networks. They try to resolve different types of problems. So you have NN to resolve static problems, to resolve time related problems and so one. The number of nodes is not so important like the design of them.
When you have a hidden layer is that you are creating a combined feature of the input. So, is the problem better tackled by more features of the existing input, or through higher-order features that come from combining existing features? This is the trade-off for a standard feed-forward network.
You have a theoretical reassurance that any function can be represented by a neural network with two hidden layers and non-linear activation.
Also, consider using additional resources for boosting, instead of adding more nodes, if you're not certain of the appropriate topology.
Very rough rules of thumb
generally more elements per layer for bigger input vectors.
more layers may let you model more non-linear systems.
If the kind of network you are using has delays in propagation , more layers may allow modelling of time series . Take care to have time jitter in the delays or it wont work very well. If this is just gobbledegook to you, ignore it.
More layers lets you insert recurrent features. This can be very useful for discrimination tasks. You ANN implementation my not permit this.
HTH
The number of units per hidden layer accounts for the ANN's potential to describe an arbitrarily complex function. Some (complicated) functions may require many hidden nodes, or possibly more than one hidden layer.
When a function can be roughly approximated by a certain number of hidden units, any extra nodes will provide more accuracy...but this is only true if the training samples used are enough to justify this addition - otherwise what will happen is "overconvergence". Overconvergence means that your ANN has lost its generalization abilities because it has overemphasized on the particular samples.
In general it is best to use the less hidden units possible, if the resulting network can give good results. The additional training patterns required to justify more hidden nodes can not be found easily in most cases, and accuracy is not the NNs' strong point.