I am currently looking through Michael Nielsen's ebook Neural Networks and Deep Learning and have run the code found at the end of chapter 1 which trains a neural network to recognize hand-written digits (with a slight modification to make the backpropagation algorithm over a mini-batch matrix-based).
However, having run this code and achieving a classification accuracy of just under 94%, I decided to remove the use of biases from the network. After re-training the modified network, I found no difference in classification accuracy!
NB: The output layer of this network contains ten neurons; if the ith of these neurons has the highest activation then the input is classified as being the digit i.
This got me wondering why it is necessary to use biases in a neural network, rather than just weights, and what differentiates between a task where biases will improve the performance of a network and a task where they will not?
My code can be found here: https://github.com/pipthagoras/neural-network-1
Biases are used to account for the fact that your underlying data might not be centered. It is clearer to see in the case of a linear regression.
If you do a regression without an intercept (or bias), you are forcing the underlying model to pass through the origin, which will result in a poor model if the underlying data is not centered (for example if the true generating process is Y=3000). If, on the other hand, your data is centered or close to centered, then eliminating bias is good, since you won't introduce a term that is, in fact, independent to your predictive variable (it's like selecting a simpler model, which will tend to generalize better PROVIDED that it actually reflects the underlying data).
Related
I understand all the computational steps of training a neural network with gradient descent using forwardprop and backprop, but I'm trying to wrap my head around why they work so much better than logistic regression.
For now all I can think of is:
A) the neural network can learn it's own parameters
B) there are many more weights than simple logistic regression thus allowing for more complex hypotheses
Can someone explain why a neural network works so well in general? I am a relative beginner.
Neural Networks can have a large number of free parameters (the weights and biases between interconnected units) and this gives them the flexibility to fit highly complex data (when trained correctly) that other models are too simple to fit. This model complexity brings with it the problems of training such a complex network and ensuring the resultant model generalises to the examples it’s trained on (typically neural networks require large volumes of training data, that other models don't).
Classically logistic regression has been limited to binary classification using a linear classifier (although multi-class classification can easily be achieved with one-vs-all, one-vs-one approaches etc. and there are kernalised variants of logistic regression that allow for non-linear classification tasks). In general therefore, logistic regression is typically applied to more simple, linearly-separable classification tasks, where small amounts of training data are available.
Models such as logistic regression and linear regression can be thought of as simple multi-layer perceptrons (check out this site for one explanation of how).
To conclude, it’s the model complexity that allows neural nets to solve more complex classification tasks, and to have a broader application (particularly when applied to raw data such as image pixel intensities etc.), but their complexity means that large volumes of training data are required and training them can be a difficult task.
Recently Dr. Naftali Tishby's idea of Information Bottleneck to explain the effectiveness of deep neural networks is making the rounds in the academic circles.
His video explaining the idea (link below) can be rather dense so I'll try to give the distilled/general form of the core idea to help build intuition
https://www.youtube.com/watch?v=XL07WEc2TRI
To ground your thinking, vizualize the MNIST task of classifying the digit in the image. For this, I am only talking about simple fully-connected neural networks (not Convolutional NN as is typically used for MNIST)
The input to a NN contains information about the output hidden inside of it. Some function is needed to transform the input to the output form. Pretty obvious.
The key difference in thinking needed to build better intuition is to think of the input as a signal with "information" in it (I won't go into information theory here). Some of this information is relevant for the task at hand (predicting the output). Think of the output as also a signal with a certain amount of "information". The neural network tries to "successively refine" and compress the input signal's information to match the desired output signal. Think of each layer as cutting away at the unneccessary parts of the input information, and
keeping and/or transforming the output information along the way through the network.
The fully-connected neural network will transform the input information into a form in the final hidden layer, such that it is linearly separable by the output layer.
This is a very high-level and fundamental interpretation of the NN, and I hope it will help you see it clearer. If there are parts you'd like me to clarify, let me know.
There are other essential pieces in Dr.Tishby's work, such as how minibatch noise helps training, and how the weights of a neural network layer can be seen as doing a random walk within the constraints of the problem.
These parts are a little more detailed, and I'd recommend first toying with neural networks and taking a course on Information Theory to help build your understanding.
Consider you have a large dataset and you want to build a binary classification model for that, Now you have two options that you have pointed out
Logistic Regression
Neural Networks ( Consider FFN for now )
Each node in a neural network will be associated with an activation function for example let's choose Sigmoid since Logistic regression also uses sigmoid internally to make decision.
Let's see how the decision of logistic regression looks when applied on the data
See some of the green spots present in the red boundary?
Now let's see the decision boundary of neural network (Forgive me for using a different color)
Why this happens? Why does the decision boundary of neural network is so flexible which gives more accurate results than Logistic regression?
or the question you asked is "Why neural networks works so well ?" is because of it's hidden units or hidden layers and their representation power.
Let me put it this way.
You have a logistic regression model and a Neural network which has say 100 neurons each of Sigmoid activation. Now each neuron will be equivalent to one logistic regression.
Now assume a hundred logistic units trained together to solve one problem versus one logistic regression model. Because of these hidden layers the decision boundary expands and yields better results.
While you are experimenting you can add more number of neurons and see how the decision boundary is changing. A logistic regression is same as a neural network with single neuron.
The above given is just an example. Neural networks can be trained to get very complex decision boundaries
Neural networks allow the person training them to algorithmically discover features, as you pointed out. However, they also allow for very general nonlinearity. If you wish, you can use polynomial terms in logistic regression to achieve some degree of nonlinearity, however, you must decide which terms you will use. That is you must decide a priori which model will work. Neural networks can discover the nonlinear model that is needed.
'Work so well' depends on the concrete scenario. Both of them do essentially the same thing: predicting.
The main difference here is neural network can have hidden nodes for concepts, if it's propperly set up (not easy), using these inputs to make the final decission.
Whereas linear regression is based on more obvious facts, and not side effects. A neural network should de able to make more accurate predictions than linear regression.
Neural networks excel at a variety of tasks, but to get an understanding of exactly why, it may be easier to take a particular task like classification and dive deeper.
In simple terms, machine learning techniques learn a function to predict which class a particular input belongs to, depending on past examples. What sets neural nets apart is their ability to construct these functions that can explain even complex patterns in the data. The heart of a neural network is an activation function like Relu, which allows it to draw some basic classification boundaries like:
Example classification boundaries of Relus
By composing hundreds of such Relus together, neural networks can create arbitrarily complex classification boundaries, for example:
Composing classification boundaries
The following article tries to explain the intuition behind how neural networks work: https://medium.com/machine-intelligence-report/how-do-neural-networks-work-57d1ab5337ce
Before you step into neural network see if you have assessed all aspects of normal regression.
Use this as a guide
and even before you discard normal regression - for curved type of dependencies - you should strongly consider kernels with SVM
Neural networks are defined with an objective and loss function. The only process that happens within a neural net is to optimize for the objective function by reducing the loss function or error. The back propagation helps in finding the optimized objective function and reach our output with an output condition.
Many of the papers I have read so far have this mentioned "pre-training network could improve computational efficiency in terms of back-propagating errors", and could be achieved using RBMs or Autoencoders.
If I have understood correctly, AutoEncoders work by learning the
identity function, and if it has hidden units less than the size of
input data, then it also does compression, BUT what does this even have
anything to do with improving computational efficiency in propagating
error signal backwards? Is it because the weights of the pre
trained hidden units does not diverge much from its initial values?
Assuming data scientists who are reading this would by theirselves
know already that AutoEncoders take inputs as target values since
they are learning identity function, which is regarded as
unsupervised learning, but can such method be applied to
Convolutional Neural Networks for which the first hidden layer is
feature map? Each feature map is created by convolving a learned
kernel with a receptive field in the image. This learned kernel, how
could this be obtained by pre-training (unsupervised fashion)?
One thing to note is that autoencoders try to learn the non-trivial identify function, not the identify function itself. Otherwise they wouldn't have been useful at all. Well the pre-training helps moving the weight vectors towards a good starting point on the error surface. Then the backpropagation algorithm, which is basically doing gradient descent, is used improve upon those weights. Note that gradient descent gets stuck in the closes local minima.
[Ignore the term Global Minima in the image posted and think of it as another, better, local minima]
Intuitively speaking, suppose you are looking for an optimal path to get from origin A to destination B. Having a map with no routes shown on it (the errors you obtain at the last layer of the neural network model) kind of tells you where to to go. But you may put yourself in a route which has a lot of obstacles, up hills and down hills. Then suppose someone tells you about a route a a direction he has gone through before (the pre-training) and hands you a new map (the pre=training phase's starting point).
This could be an intuitive reason on why starting with random weights and immediately start to optimize the model with backpropagation may not necessarily help you achieve the performance you obtain with a pre-trained model. However, note that many models achieving state-of-the-art results do not use pre-training necessarily and they may use the backpropagation in combination with other optimization methods (e.g. adagrad, RMSProp, Momentum and ...) to hopefully avoid getting stuck in a bad local minima.
Here's the source for the second image.
I don't know a lot about autoencoder theory, but I've done a bit of work with RBMs. What RBMs do is they predict what the probability is of seeing the specific type of data in order to get the weights initialized to the right ball park- it is considered an (unsupervised) probabilistic model, so you don't correct using the known labels. Basically, the idea here is that having a learning rate that is too big will never lead to convergence but having one that is too small will take forever to train. Thus, by "pretraining" in this way you find out the ball park of the weights and then can set the learning rate to be small in order to get them down to the optimal values.
As for the second question, no, you don't generally prelearn kernels, at least not in an unsupervised fashion. I suspect that what is meant by pretraining here is a bit different than in your first question- this is to say, that what is happening is that they are taking a pretrained model (say from model zoo) and fine tuning it with a new set of data.
Which model you use generally depends on the type of data you have and the task at hand. Convnets I've found to train faster and efficiently, but not all data has meaning when convolved, in which case dbns may be the way to go. Unless say, you have a small amount of data then I'd use something other than neural networks entirely.
Anyways, I hope this helps clear some of your questions.
I've been studying neural networks for a bit and recently learned about the dropout training algorithm. There are excellent papers out there to understand how it works, including the ones from the authors.
So I built a neural network with dropout training (it was fairly easy) but I'm a bit confused about how to perform model selection. From what I understand, looks like dropout is a method to be used when training the final model obtained through model selection.
As for the test part, papers always talk about using the complete network with halved weights, but they do not mention how to use it in the training/validation part (at least the ones I read).
I was thinking about using the network without dropout for the model selection part. Say that makes me find that the net performs well with N neurons. Then, for the final training (the one I use to train the network for the test part) I use 2N neurons with dropout probability p=0.5. That assures me to have exactly N neurons active on average, thus using the network at the right capacity most of the time.
Is this a correct approach?
By the way, I'm aware of the fact that dropout might not be the best choice with small datasets. The project I'm working on has academic purposes, so it's not really needed that I use the best model for the data, as long as I stick with machine learning good practices.
First of all, model selection and the training of a particular model are completely different issues. For model selection, you would usually need a data set that is completely independent of both training set used to build the model and test set used to estimate its performance. So if you're doing for example a cross-validation, you would need an inner cross-validation (to train the models and estimate the performance in general) and an outer cross-validation to do the model selection.
To see why, consider the following thought experiment (shamelessly stolen from this paper). You have a model that makes a completely random prediction. It has a number of parameters that you can set, but have no effect. If you're trying different parameter settings long enough, you'll eventually get a model that has a better performance than all the others simply because you're sampling from a random distribution. If you're using the same data for all of these models, this is the model you will choose. If you have a separate test set, it will quickly tell you that there is no real effect because the performance of this parameter setting that achieves good results during the model-building phase is not better on the separate set.
Now, back to neural networks with dropout. You didn't refer to any particular paper; I'm assuming that you mean Srivastava et. al. "Dropout: A Simple Way to Prevent Neural Networks from Overfitting". I'm not an expert on the subject, but the method to me seems to be similar to what's used in random forests or bagging to mitigate the flaws an individual learner may exhibit by applying it repeatedly in slightly different contexts. If I understood the method correctly, essentially what you end up with is an average over several possible models, very similar to random forests.
This is a way to make an individual model better, but not for model selection. The dropout is a way of adjusting the learned weights for a single neural network model.
To do model selection on this, you would need to train and test neural networks with different parameters and then evaluate those on completely different sets of data, as described in the paper I've referenced above.
I Wrote a simple recurrent neural network (7 neurons, each one is initially connected to all the neurons) and trained it using a genetic algorithm to learn "complicated", non-linear functions like 1/(1+x^2). As the training set, I used 20 values within the range [-5,5] (I tried to use more than 20 but the results were not changed dramatically).
The network can learn this range pretty well, and when given examples of other points within this range, it can predict the value of the function. However, it can not extrapolate correctly and predicting the values of the function outside the range [-5,5]. What are the reasons for that and what can I do to improve its extrapolation abilities?
Thanks!
Neural networks are not extrapolation methods (no matter - recurrent or not), this is completely out of their capabilities. They are used to fit a function on the provided data, they are completely free to build model outside the subspace populated with training points. So in non very strict sense one should think about them as an interpolation method.
To make things clear, neural network should be capable of generalizing the function inside subspace spanned by the training samples, but not outside of it
Neural network is trained only in the sense of consistency with training samples, while extrapolation is something completely different. Simple example from "H.Lohninger: Teach/Me Data Analysis, Springer-Verlag, Berlin-New York-Tokyo, 1999. ISBN 3-540-14743-8" shows how NN behave in this context
All of these networks are consistent with training data, but can do anything outside of this subspace.
You should rather reconsider your problem's formulation, and if it can be expressed as a regression or classification problem then you can use NN, otherwise you should think about some completely different approach.
The only thing, which can be done to somehow "correct" what is happening outside the training set is to:
add artificial training points in the desired subspace (but this simply grows the training set, and again - outside of this new set, network's behavious is "random")
add strong regularization, which will force network to create very simple model, but model's complexity will not guarantee any extrapolation strength, as two model's of exactly the same complexity can have for example completely different limits in -/+ infinity.
Combining above two steps can help building model which to some extent "extrapolates", but this, as stated before, is not a purpose of a neural network.
As far as I know this is only possible with networks which do have the echo property. See Echo State Networks on scholarpedia.org.
These networks are designed for arbitrary signal learning and are capable to remember their behavior.
You can also take a look at this tutorial.
The nature of your post(s) suggests that what you're referring to as "extrapolation" would be more accurately defined as "sequence recognition and reproduction." Training networks to recognize a data sequence with or without time-series (dt) is pretty much the purpose of Recurrent Neural Network (RNN).
The training function shown in your post has output limits governed by 0 and 1 (or -1, since x is effectively abs(x) in the context of that function). So, first things first, be certain your input layer can easily distinguish between negative and positive inputs (if it must).
Next, the number of neurons is not nearly as important as how they're layered and interconnected. How many of the 7 were used for the sequence inputs? What type of network was used and how was it configured? Network feedback will reveal the ratios, proportions, relationships, etc. and aid in the adjustment of network weight adjustments to match the sequence. Feedback can also take the form of a forward-feed depending on the type of network used to create the RNN.
Producing an 'observable' network for the exponential-decay function: 1/(1+x^2), should be a decent exercise to cut your teeth on RNNs. 'Observable', meaning the network is capable of producing results for any input value(s) even though its training data is (far) smaller than all possible inputs. I can only assume that this was your actual objective as opposed to "extrapolation."
Is anyone here who is familiar with echo state networks? I created an echo state network in c#. The aim was just to classify inputs into GOOD and NOT GOOD ones. The input is an array of double numbers. I know that maybe for this classification echo state network isn't the best choice, but i have to do it with this method.
My problem is, that after training the network, it cannot generalize. When i run the network with foreign data (not the teaching input), i get only around 50-60% good result.
More details: My echo state network must work like a function approximator. The input of the function is an array of 17 double values, and the output is 0 or 1 (i have to classify the input into bad or good input).
So i have created a network. It contains an input layer with 17 neurons, a reservoir layer, which neron number is adjustable, and output layer containing 1 neuron for the output needed 0 or 1. In a simpler example, no output feedback is used (i tried to use output feedback as well, but nothing changed).
The inner matrix of the reservoir layer is adjustable too. I generate weights between two double values (min, max) with an adjustable sparseness ratio. IF the values are too big, it normlites the matrix to have a spectral radius lower then 1. The reservoir layer can have sigmoid and tanh activaton functions.
The input layer is fully connected to the reservoir layer with random values. So in the training state i run calculate the inner X(n) reservor activations with training data, collecting them into a matrix rowvise. Using the desired output data matrix (which is now a vector with 1 ot 0 values), i calculate the output weigths (from reservoir to output). Reservoir is fully connected to the output. If someone used echo state networks nows what im talking about. I ise pseudo inverse method for this.
The question is, how can i adjust the network so it would generalize better? To hit more than 50-60% of the desired outputs with a foreign dataset (not the training one). If i run the network again with the training dataset, it gives very good reults, 80-90%, but that i want is to generalize better.
I hope someone had this issue too with echo state networks.
If I understand correctly, you have a set of known, classified data that you train on, then you have some unknown data which you subsequently classify. You find that after training, you can reclassify your known data well, but can't do well on the unknown data. This is, I believe, called overfitting - you might want to think about being less stringent with your network, reducing node number, and/or training based on a hidden dataset.
The way people do it is, they have a training set A, a validation set B, and a test set C. You know the correct classification of A and B but not C (because you split up your known data into A and B, and C are the values you want the network to find for you). When training, you only show the network A, but at each iteration, to calculate success you use both A and B. So while training, the network tries to understand a relationship present in both A and B, by looking only at A. Because it can't see the actual input and output values in B, but only knows if its current state describes B accurately or not, this helps reduce overfitting.
Usually people seem to split 4/5 of data into A and 1/5 of it into B, but of course you can try different ratios.
In the end, you finish training, and see what the network will say about your unknown set C.
Sorry for the very general and basic answer, but perhaps it will help describe the problem better.
If your network doesn't generalize that means it's overfitting.
To reduce overfitting on a neural network, there are two ways:
get more training data
decrease the number of neurons
You also might think about the features you are feeding the network. For example, if it is a time series that repeats every week, then one feature is something like the 'day of the week' or the 'hour of the week' or the 'minute of the week'.
Neural networks need lots of data. Lots and lots of examples. Thousands. If you don't have thousands, you should choose a network with just a handful of neurons, or else use something else, like regression, that has fewer parameters, and is therefore less prone to overfitting.
Like the other answers here have suggested, this is a classic case of overfitting: your model performs well on your training data, but it does not generalize well to new test data.
Hugh's answer has a good suggestion, which is to reduce the number of parameters in your model (i.e., by shrinking the size of the reservoir), but I'm not sure whether it would be effective for an ESN, because the problem complexity that an ESN can solve grows proportional to the logarithm of the size of the reservoir. Reducing the size of your model might actually make the model not work as well, though this might be necessary to avoid overfitting for this type of model.
Superbest's solution is to use a validation set to stop training as soon as performance on the validation set stops improving, a technique called early stopping. But, as you noted, because you use offline regression to compute the output weights of your ESN, you cannot use a validation set to determine when to stop updating your model parameters---early stopping only works for online training algorithms.
However, you can use a validation set in another way: to regularize the coefficients of your regression! Here's how it works:
Split your training data into a "training" part (usually 80-90% of the data you have available) and a "validation" part (the remaining 10-20%).
When you compute your regression, instead of using vanilla linear regression, use a regularized technique like ridge regression, lasso regression, or elastic net regression. Use only the "training" part of your dataset for computing the regression.
All of these regularized regression techniques have one or more "hyperparameters" that balance the model fit against its complexity. The "validation" dataset is used to set these parameter values: you can do this using grid search, evolutionary methods, or any other hyperparameter optimization technique. Generally speaking, these methods work by choosing values for the hyperparameters, fitting the model using the "training" dataset, and measuring the fitted model's performance on the "validation" dataset. Repeat N times and choose the model that performs best on the "validation" set.
You can learn more about regularization and regression at http://en.wikipedia.org/wiki/Least_squares#Regularized_versions, or by looking it up in a machine learning or statistics textbook.
Also, read more about cross-validation techniques at http://en.wikipedia.org/wiki/Cross-validation_(statistics).