I don't understand how adding the product of lambda and the sum of squared thetas to the cost function would decrease the amount of over fitting in a data set. Can someone please explain?
Imagine two extreme cases:
You do not need to learn anything ==> You need 0 parameters to learn (an extreme case of underfitting).
You want to memorize everything you see (in the training set) ==> You need a huge amount of parameters to remember everything (an extreme case of overfitting)
The real training should happen between these two cases, to lead to a good generalizations. A good generalization helps to get more realistic predictions on the unseen test data.
When you try to minimize a cost function, you are penalizing the machine for each of the wrong predictions on the training set. To overcome this penalization, and most of the time easier than getting to a real generalization, machine would prefer to memorize everything in training, so that it gets to a lower loss, and get less penalized. This easily happens when you provide a complex network( with big amount of training parameters, AKA when W is big)
To prevent this trick from the machine, we force the machine to reduce the cost, but also we put the condition to use not a very large set of parameters. That is one way to do the regularization.
Related
I'm fairly new to machine learning, and working on optimizing hyperparameters for my model. I'm doing this via a randomized search. My question is: should I be searching over # of epochs and batch size along with my other hyperparameters (e.g. loss function, number of layers, etc.)? If not, should I fix a these values first, find the other parameters, then return to tune these?
My concern is a) that searching over many epochs will be extremely time-consuming, so leaving it at one low value for the initial scan would be useful and b) that these parameters, esp. # of epochs, will disproportionately affect the results when the model is behaving well, and won't really give me much information about the rest of my architecture, as there should be a regime where more epochs, up to a point, are better. I know this isn't totally accurate, i.e. # of epochs is a real hyperparameter and too many can lead to overfitting issues, for example. Currently, my model is not clearly improving with # of epochs, though it was suggested by someone working on a similar problem within my area of research that this may be mitigated by implementing batch normalization, which is another parameter I am testing. Finally, I am worried that batch size will be quite affected by the fact that I am scaling my data down to 60% to allow my code to run reasonably (and I think the final model will be trained on vastly more data than the simulated data currently available to me).
I agree with your intuition on epochs. It is common to keep this value as low as possible in order to complete more training "experiments" in the same number of working hours. I don't have a great reference here, but I would welcome one in the comments.
For almost everything else, there is a paper by Leslie N. Smith that I can't recommend enough, A disciplined approach to neural network hyper-parameters: Part 1 -- learning rate, batch size, momentum, and weight decay.
As you can see, batch size is included but epochs are not. You will also notice that the model architecture is not included (number of layers, layer size, etc). Neural Architecture Search is a huge research field in its own right, separate from hyper-parameter tuning.
As for the loss function, I can't think of any reason to "tune" that except in the context of an Auxiliary Loss for training only, which I suspect is not what you are talking about.
The loss function that will be applied to your validation or test set is part of the problem statement. That, along with the data, defines the problem you are solving. You don't changing it by tuning, you change it by convincing a product manager that your alternative is better for the business need.
I'm using bayesian optimization with gaussian processes to optimize my recurrent network parameters.
I'm getting pretty good results. I'm returning the validation loss of the last epoch as a feedback for the optimization algorithm.
I really would like to include some other parameters inside the feedback of each iteration, e.g. like training duration and the amount of modell parameter, because I want to find a small model with a good performance without changing the values of the parameter to a smaller range.
Problem is that I don't know the scale of the loss. If a just scale the other parameters between 0 and 1, the loss has probably to less influence if it's always in the range 0.02-0.01. In this case the smallest model would be picked, but it would have a bad performance.
Thanks for your ideas. Cheers!
I am reading the a deep learning with python book.
After reading chapter 4, Fighting Overfitting, I have two questions.
Why might increasing the number of epochs cause overfitting?
I know increasing increasing the number of epochs will involve more attempts at gradient descent, will this cause overfitting?
During the process of fighting overfitting, will the accuracy be reduced ?
I'm not sure which book you are reading, so some background information may help before I answer the questions specifically.
Firstly, increasing the number of epochs won't necessarily cause overfitting, but it certainly can do. If the learning rate and model parameters are small, it may take many epochs to cause measurable overfitting. That said, it is common for more training to do so.
To keep the question in perspective, it's important to remember that we most commonly use neural networks to build models we can use for prediction (e.g. predicting whether an image contains a particular object or what the value of a variable will be in the next time step).
We build the model by iteratively adjusting weights and biases so that the network can act as a function to translate between input data and predicted outputs. We turn to such models for a number of reasons, often because we just don't know what the function is/should be or the function is too complex to develop analytically. In order for the network to be able to model such complex functions, it must be capable of being highly-complex itself. Whilst this complexity is powerful, it is dangerous! The model can become so complex that it can effectively remember the training data very precisely but then fail to act as an effective, general function that works for data outside of the training set. I.e. it can overfit.
You can think of it as being a bit like someone (the model) who learns to bake by only baking fruit cake (training data) over and over again – soon they'll be able to bake an excellent fruit cake without using a recipe (training), but they probably won't be able to bake a sponge cake (unseen data) very well.
Back to neural networks! Because the risk of overfitting is high with a neural network there are many tools and tricks available to the deep learning engineer to prevent overfitting, such as the use of dropout. These tools and tricks are collectively known as 'regularisation'.
This is why we use development and training strategies involving test datasets – we pretend that the test data is unseen and monitor it during training. You can see an example of this in the plot below (image credit). After about 50 epochs the test error begins to increase as the model has started to 'memorise the training set', despite the training error remaining at its minimum value (often training error will continue to improve).
So, to answer your questions:
Allowing the model to continue training (i.e. more epochs) increases the risk of the weights and biases being tuned to such an extent that the model performs poorly on unseen (or test/validation) data. The model is now just 'memorising the training set'.
Continued epochs may well increase training accuracy, but this doesn't necessarily mean the model's predictions from new data will be accurate – often it actually gets worse. To prevent this, we use a test data set and monitor the test accuracy during training. This allows us to make a more informed decision on whether the model is becoming more accurate for unseen data.
We can use a technique called early stopping, whereby we stop training the model once test accuracy has stopped improving after a small number of epochs. Early stopping can be thought of as another regularisation technique.
More attempts of decent(large number of epochs) can take you very close to the global minima of the loss function ideally, Now since we don't know anything about the test data, fitting the model so precisely to predict the class labels of the train data may cause the model to lose it generalization capabilities(error over unseen data). In a way, no doubt we want to learn the input-output relationship from the train data, but we must not forget that the end goal is for the model to perform well over the unseen data. So, it is a good idea to stay close but not very close to the global minima.
But still, we can ask what if I reach the global minima, what can be the problem with that, why would it cause the model to perform badly on unseen data?
The answer to this can be that in order to reach the global minima we would be trying to fit the maximum amount of train data, this will result in a very complex model(since it is less probable to have a simpler spatial distribution of the selected number of train data that is fortunately available with us). But what we can assume is that a large amount of unseen data(say for facial recognition) will have a simpler spatial distribution and will need a simpler Model for better classification(I mean the entire world of unseen data, will definitely have a pattern that we can't observe just because we have an access small fraction of it in the form of training data)
If you incrementally observe points from a distribution(say 50,100,500, 1000 ...), we will definitely find the structure of the data complex until we have observed a sufficiently large number of points (max: the entire distribution), but once we have observed enough points we can expect to observe the simpler pattern present in the data that can be easily classified.
In short, a small fraction of train data should have a complex structure as compared to the entire dataset. And overfitting to the train data may cause our model to perform worse on the test data.
One analogous example to emphasize the above phenomenon from day to day life is as follows:-
Say we meet N number of people till date in our lifetime, while meeting them we naturally learn from them(we become what we are surrounded with). Now if we are heavily influenced by each individual and try to tune to the behaviour of all the people very closely, we develop a personality that closely resembles the people we have met but on the other hand we start judging every individual who is unlike me -> unlike the people we have already met. Becoming judgemental takes a toll on our capability to tune in with new groups since we trained very hard to minimize the differences with the people we have already met(the training data). This according to me is an excellent example of overfitting and loss in genralazition capabilities.
I'm working with an extremelly unbalanced and heterogeneous multiclass {K = 16} database for research, with a small N ~= 250. For some labels the database has a sufficient amount of examples for supervised machine learning, but for others I have almost none. I'm also not in a position to expand my database for a number of reasons.
As a first approach I divided my database into training (80%) and test (20%) sets in a stratified way. On top of that, I applied several classification algorithms that provide some results. I applied this procedure over 500 stratified train/test sets (as each stratified sampling takes individuals randomly within each stratum), hoping to select an algorithm (model) that performed acceptably.
Because of my database, depending on the specific examples that are part of the train set, the performance on the test set varies greatly. I'm dealing with runs that have as high (for my application) as 82% accuracy and runs that have as low as 40%. The median over all runs is around 67% accuracy.
When facing this situation, I'm unsure on what is the standard procedure (if there is any) when selecting the best performing model. My rationale is that the 90% model may generalize better because the specific examples selected in the training set are be richer so that the test set is better classified. However, I'm fully aware of the possibility of the test set being composed of "simpler" cases that are easier to classify or the train set comprising all hard-to-classify cases.
Is there any standard procedure to select the best performing model considering that the distribution of examples in my train/test sets cause the results to vary greatly? Am I making a conceptual mistake somewhere? Do practitioners usually select the best performing model without any further exploration?
I don't like the idea of using the mean/median accuracy, as obviously some models generalize better than others, but I'm by no means an expert in the field.
Confusion matrix of the predicted label on the test set of one of the best cases:
Confusion matrix of the predicted label on the test set of one of the worst cases:
They both use the same algorithm and parameters.
Good Accuracy =/= Good Model
I want to firstly point out that a good accuracy on your test set need not equal a good model in general! This has (in your case) mainly to do with your extremely skewed distribution of samples.
Especially when doing a stratified split, and having one class dominatingly represented, you will likely get good results by simply predicting this one class over and over again.
A good way to see if this is happening is to look at a confusion matrix (better picture here) of your predictions.
If there is one class that seems to confuse other classes as well, that is an indicator for a bad model. I would argue that in your case it would be generally very hard to find a good model unless you do actively try to balance your classes more during training.
Use the power of Ensembles
Another idea is indeed to use ensembling over multiple models (in your case resulting from different splits), since it is assumed to generalize better.
Even if you might sacrifice a lot of accuracy on paper, I would bet that a confusion matrix of an ensemble is likely to look much better than the one of a single "high accuracy" model. Especially if you disregard the models that perform extremely poor (make sure that, again, the "poor" performance comes from an actual bad performance, and not just an unlucky split), I can see a very good generalization.
Try k-fold Cross-Validation
Another common technique is k-fold cross-validation. Instead of performing your evaluation on a single 80/20 split, you essentially divide your data in k equally large sets, and then always train on k-1 sets, while evaluating on the other set. You then not only get a feeling whether your split was reasonable (you usually get all the results for different splits in k-fold CV implementations, like the one from sklearn), but you also get an overall score that tells you the average of all folds.
Note that 5-fold CV would equal a split into 5 20% sets, so essentially what you are doing now, plus the "shuffling part".
CV is also a good way to deal with little training data, in settings where you have imbalanced classes, or where you generally want to make sure your model actually performs well.
Based on data that our business department supplied to us, I used the sklearn decision tree algorithm to determine the ROC_AUC for a binary classification problem.
The data consists of 450 rows and there are 30 features in the data.
I used 10 times StratifiedKFold repetition/split of training and test data. As a result, I got the following ROC_AUC values:
0.624
0.594
0.522
0.623
0.585
0.656
0.629
0.719
0.589
0.589
0.592
As I am new in machine learning, I am unsure whether such a variation in the ROC_AUC values can be expected (with minimum values of 0.522 and maximum values of 0.719).
My questions are:
Is such a big variation to be expected?
Could it be reduced with more data (=rows)?
Will the ROC_AUC variance get smaller, if the ROC_AUC gets better ("closer to 1")?
Well, you do k-fold splits to actually evaluate how well your model generalizes.
Therefore, from your current results I would assume the following:
This is a difficult problem, the AUCs are usually low.
0.71 is an outlier, you were just lucky there (probably).
Important questions that will help us help you:
What is the proportion of the binary classes? Are they balanced?
What are the features? Are they all continuous? If categorical, are they ordinal or nominal?
Why Decision Tree? Have you tried other methods? Logistic Regression for instance is a good start before you move on to more advanced ML methods.
You should run more iterations, instead of k fold use the ShuffleSplit function and run at least 100 iterations, compute the Average AUC with 95% Confidence Intervals. That will give you a better idea of how well the models perform.
Hope this helps!
Is such a big variation to be expected?
This is a textbook case of high variance.
Depending on the difficulty of your problem, 405 training samples may not be enough for it to generalize properly, and the random forest may be too powerful.
Try adding some regularization, by limiting the number of splits that the trees are allowed to make. This should reduce the variance in your model, though you might expect a potentially lower average performance.
Could it be reduced with more data (=rows)?
Yes, adding data is the other popular way of lowering the variance of your model. If you're familiar with deep learning, you'll know that deep models usually need LOTS of samples to learn properly. That's because they are very powerful models with an intrinsically high variance, and therefore a lot of data is needed for them to generalize.
Will the ROC_AUC variance get smaller, if the ROC_AUC gets better ("closer to 1")?
Variance will decrease with regularization and adding data, it has no relation to the actual performance "number" that you get.
Cheers