I have a multilayer perceptron with 5 hidden layers and 256 neurons each. When I start training, I get different prediction probabilities for each train sample until epoch 50, but then the number of duplicate predictions increases, on epoch 300 I already have 30% of duplicate predictions which does not make sense since the input data is different for all training samples. Any idea what causes this behavior?
Clarifications:
with "duplicate predictions", I mean items with the exactly same predicted probability to belong to class A (it's a binary classification problem)
I have 4000 training samples with 200 features each and all samples are different, it does not make sense that the number of duplicate predictions increases to 30% while training. So I wonder what can cause this behavior.
One point, you say you are doing a binary prediction, and when you say "duplicate predictions", even with your clarification it's hard to understand your meaning. I am guessing that you have two outputs for your binary classifier, one for class A and one for class B and you are getting roughly the same value for a given sample. If that's the case, then the first thing to do is to use 1 output. A binary classification problem is better modeled with 1 output that ranges between 0 and 1 (sigmoid the output neuron). This way there will be no ambiguity, the network will have to choose one or the other, or when it's confused you'll get ~0.5 and it will be clear.
Second, it is very common for a network to start learning well and then to perform more poorly after overtraining. Especially with small datasets such as what you have. In fact, even with the little knowledge I have of your dataset I would put a small bet on you getting better performance out of an algorithm like XGA Boost than a neural network (I assume you're using a neural net and not literally a perceptron).
But regarding the performance degrading over time. When this happens you want to look into something called "early stopping". At some point the network will start memorizing the input, and may be part of what's happening. Essentially you train until the performance on your held out test data starts to worsen.
To address this you can apply various forms of regularization (L2 regularization, dropout, batch normalization all come to mind). You can also reduce the size of your network. 5 layers of 256 neurons sounds too big for the problem. Try trimming this down and I bet your results will improve. There is a sweet spot for architecture size in neural networks. When your network is too large it can, and often will, over fit. When it's too small it won't be expressive enough for the data. Angrew Ng's coursera class has some helpful practical advice on dealing with this.
Related
I have a 2 layered Neural Network that I'm training on about 10000 features (genomic data) with about 100 samples in my data set. Now I realized that anytime I run my model (i.e. compile & fit) I get varying validation/testing accuracys even if I leave the train/test/validation split untouched. Sometimes its around 70% sometimes around 90%.
Due to the stochastic nature of the NN I anticipate some variation but could these strong fluctuations be a sign of something else?
The reason why you're seeing such a big instability with your validation accuracy is because your neural network is huge in comparison to the data you train it on.
Even with just 12 neurons per layer, you still have 12 * 10000 + 12 = 120012 parameters in your first layer. Now think about what the neural network does under the hood. It takes your 10000 inputs, it multiplies each input by some weight and then sums all these inputs. Now you provide it only 64 training examples on which the training algorithm is supposed to decide what are the correct input weights. Just based on intuition, from a purely combinatorial perspective there is going to be large amount of weight assignments that do well on your 64 training samples. And you have no guarantee that the training algorithm will pick such weight assignment that will also do well on your out-of-sample data.
Given neural network is able to represent a wide variety of functions (it's been proven that under certain assumptions it can approximate any function, that's called general approximation). To select the function you want you provide the training algorithm with data to constrain the space of all possible functions the network can represent to a subspace of functions that fit your data. However, such function is in no way guaranteed to represent the true underlying relationship between the input and the output. And especially if the number of parameters is larger than the number of samples (in this case by a few orders of magnitude), you're nearly guaranteed to see your network simply memorize the samples in your training data, simply because it has the capacity to do so and you haven't constrained it enough.
In other words, what you're seeing is overfitting. In NNs, the general rule of thumb is that you want at least a couple of times more samples than you have parameters (look in to the Hoeffding Inequality for theoretical rationale of this) and in effect the more samples you have, the less you're afraid of overfitting.
So here is a couple of possible solutions:
Use an algorithm that's more suitable for the case where you have high input dimension and low sample count, such as Kernel SVM (Support Vector Machine). With such a low sample count, it's quite possible that a Kernel SVM algorithm will achieve better and more consistent validation accuracy. (You can easily test this, they are available in the scikit-learn package, really easy to use)
If you insist on using NN - use regularization. Given the fact you already have working code, this will be easy, just add kernel_regularizer to all your layers, I would try both L1 and L2 regularization (probably separately). L1 regularization tends to push weights to zero so it might help reduce the number of parameters in your problem. L2 just tries to make all the weights small. Use your validation set to decide the best value for each regularization. You can optimize both for the best mean accuracy and also the lowest variance in accuracy on your validation data (do something like 20 training runs for each parameter value of L1 and L2 regularization, usually just trying different orders of magnitude is sufficient, e.g. 1e-4, 1e-3, 1e-2, 1e-1, 1, 1e1).
If most of your input features are not really predictive or if they are highly correlated, PCA (Principal Component Analysis) can be used to project your inputs into a much lower dimensional space (e.g. from 10000 to 20), where you'd have much smaller neural network (still I'd use L1 or L2 for regularization because even then you'd have more weights than training samples)
On a final note, the point of a testing set is to use it very sparsely (ideally only once). It should be the final reported metric after all your research and model tuning is done. You should not optimize any values on it. You should do all this on your validation set. To avoid overfitting on your validation set, look into k-fold cross validation.
I just trained a CNN with Tensorflow/Keras and saved it as a model. I tried running about 1000 inputs through it multiple times, and each time got a slightly different prediction accuracy. The accuracy was good, and I am not concerned with the performance; however, I thought that CNN models, once trained, should be deterministic. That is, any input will always be classified the same way. Is this not the case? Is there variability in the way a model can predict once trained? If not, hopefully I can assume that I have programmed some variability into my code unawares. Any help would be appreciated.
Once a CNN is trained, should its ouputs be deterministic?
Well, in theory, yes. In practise, as Peter Duniho points out in his excellent explanatory comment, we can see very small deviations because of the way values are calculated, aggregated, etc.
In practice the probability of such small deviations changing the predicted category (and therefore the accuracy) of a classification model are so small that I'd be almost certain something else is at play in your example. Even over a sample size of 1000.
Have you left on some training regularisation like batch normalisation? Are you certain you are evaluating precisely the same 1000 inputs each time? Got to suspect the issue is in the code rather than rounding errors.
Can you determine which specific classification changes?
I created a char-based CNN model for text classification on keras + tensorflow - mainly using Conv1D, mainly based on:
http://www.wildml.com/2015/12/implementing-a-cnn-for-text-classification-in-tensorflow/
The model is performing very good with 80%+ accuracy on test data set. However I'm having problem with false positive. One of the reason could be that the final layer is a Dense layer with softmax activation function.
To give an idea of how the model is performing, I train the model with data set with 31 classes with 1021 samples, the performance is ~85% on 25% test data set
However if you include false negative the performance is pretty bad (I didn't run another test data with false negative since it's pretty obvious just testing by hand) - every input has a corresponding prediction. For example a sentence acasklncasdjsandjas can result in a class ask_promotion.
Are there any best practice on how to deal with false positive in this case?
My idea is to:
Implement a noise class where samples are just a set of totally random text. However this doesn't seem to help since the noise doesn't contain any pattern thus it would be difficult to train the model
Replace softmax with something that doesn't require all output probability to 1 so small values can stay small regardless of other values. I did some research on this but there's not much information on changing the activation function for this specific case
That sounds like the issue of imbalanced data, where two classes have completely different supports (the number of instances in each class). This issue is particularly crucial in the task of hierarchical classification in which some classes with a deep hierarchy tend to have much more instances than the others.
Anyway, let's simply the issue as binary classification, and name the class with much more support Class-A and the other one with less support Class-B. Generally speaking, there are two popular ways to circumvent this issue.
Under-sampling: You fix Class-B as is. Then you sample instances from Class-A for the same amount as Class-B. Combine these instances and train your classifier with them.
Over-sampling: You fix Class-A as is. Then you sample instances from Class-B for the same amount as Class-A. The same goes with Choice 1.
For more information, please refer to this KDNuggets page.
https://www.kdnuggets.com/2017/06/7-techniques-handle-imbalanced-data.html
Hope this helps. :P
Usually, when using Keras, the datasets used to train the neural network are labeled.
For example, if I have a 100,000 rows of patients with 12 field per each row, then the last field will indicate if this patient is diabetic or no (0 or 1).
And then after training is finished I can insert a new record and predict if this person is diabetic or no.
But in the case of unlabeled datasets, where I can not label the data due to some reasons, how can I train the neural network to let him know that those are the normal records and any new record that does not match this network will be malicious or not accepted ?
This is called one-class learning and is usually done by using autoencoders. You train an autoencoder on the training data to reconstruct the data itself. The labels in this case is the input itself. This will give you a reconstruction error. https://en.wikipedia.org/wiki/Autoencoder
Now you can define a threshold where the data is benign or not, depending on the reconstruction error. The hope is that the reconstruction of the good data is better than the reconstruction of the bad data.
Edit to answer the question about the difference in performance between supervised and unsupervised learning.
This cannot be said with any certainty, because I have not tried it and I do not know what the final accuracy is going to be. But for a rough estimate supervised learning will perform better on the trained data, because more information is supplied to the algorithm. However if the actual data is quite different to the training data the network will underperform in practice, while the autoencoder tends to deal better with different data. Additionally, per rule of thumb you should have 5000 examples per class to train a neural network reliably, so labeling could take some time. But you will need some data to test anyways.
It sounds like you need fit two different models:
a model for bad record detection
a model for prediction of a patient's likelihood to be diabetic
For both of these models, you will need to have labels. For the first model your labels would indicate whether the record is good or bad (malicious) and the second would be whether the patient is diabetic or not.
In order to detect bad records, you may find that simple logistic regression or SVM performs adequately.
I have a dataset for classification which is composed of a training of size 8000x(32x32x3 images) and of a test of size 2000x(same size images).
I am doing a very simple task of distinguishing vehicules and background. I am using the cross_entropy as a cost function.
The net I am using is almost the same as the one used in DeepMNIST except the first filter has size 3x... instead of 1x... because it is a colour image and the output has size two because there are only two classes : vehicules or not vehicules.
Seeing the results of this relatively straight forward task has led me to ask myself several interrogations :
-First if I do not use a large enough batch size (>200) I get stuck almost every time at accuracy 62% (in a local optima) over the two sets which is not sufficient for my need
-Secondly whenever I use the right optimizer Adam with the right batch size and learning rate I go up to 92% however the outputs are always very disturbingly good like [0.999999999 0.000000000001].
This should not happen as the task is difficult.
Therefore when I go fully convolutional to create a heatmap I got 1.000001 almost everywhere due to the saturation.
What am I doing wrong ? Do you think whitening would solve the problem ? Batch normalization ? Something else ? What am I facing ?
That's a sign of overfitting. If you train on small dataset long enough with a large enough model, eventually your confidences get saturated to 0's and 1's. Hence, same techniques that prevent overfitting (regularization penalties, dropout, early stopping, data augmentation) will help there.
My first step for a tiny dataset like this, would be to augment the dataset with noise corrupted examples. IE, for your example I would add 800k noise corrupted examples with original labels, and train on those.