I have a supervised learning problem that I am solving with the Keras functional API.
As this model is predicting the state of a physical system, I know the supervised model should follow additional constraints.
I would like to add that as an additional loss term that penalizes the model for making predictions that do not follow those constraints. Unfortunately, the number of training examples for the supervised learning problem >> the number of constraint examples.
Basically, I am trying to do this:
Minimizing both the supervised learning error, and the constraint error as an auxiliary loss.
I do not believe that alternating training batches on each dataset will be successful, because the gradient will only capture the error of one problem at a time, when I really want the physical constraint to act as regularization on the supervised learning task. (If I am incorrect in my interpretation, please let me know).
I know this could be implemented in pure Tensorflow or Theano, but I am hesitant to leave the Keras ecosystem that makes everything else so convenient. If anybody knows how to train a model with batch sizes that vary across inputs, I'd really appreciate the help.
Related
Consider the training process of deep FF neural network using mini-batch gradient descent. As far as I understand, at each epoch of the training we have different random set of mini-batches. Then iterating over all mini batches and computing the gradients of the NN parameters we will get random gradients at each iteration and, therefore, random directions for the model parameters to minimize the cost function. Let's imagine we fixed the hyperparameters of the training algorithm and started the training process again and again, then we would end up with models, which completely differs from each other, because in those trainings the changes of model parameters were different.
1) Is it always the case when we use such random based training algorithms?
2) If it is so, where is the guaranty that training the NN one more time with the best hyperparameters found during the previous trainings and validations will yield us the best model again?
3) Is it possible to find such hyperparameters, which will always yield the best models?
Neural Network are solving a optimization problem, As long as it is computing a gradient in right direction but can be random, it doesn't hurt its objective to generalize over data. It can stuck in some local optima. But there are many good methods like Adam, RMSProp, momentum based etc, by which it can accomplish its objective.
Another reason, when you say mini-batch, there is at least some sample by which it can generalize over those sample, there can be fluctuation in the error rate, and but at least it can give us a local solution.
Even, at each random sampling, these mini-batch have different-2 sample, which helps in generalize well over the complete distribution.
For hyperparameter selection, you need to do tuning and validate result on unseen data, there is no straight forward method to choose these.
I have used the neural network software for predicting the continous data. Obviously the prediction was better than the results obtained through regression analysis. Now i would like to derive a model expression from the trained weights obtained from the training of the continous data through the software, as suggested by many researchers on how to interpret the trained weights and biases for deriving the model equation i tried to derive one from the similar lines.
After deriving the equation i found that the equation was not able to replicate the same results as given by the neural network software. so i am exploring the new methods to derive the equation. I want to know where i am going wrong and if any one can provide me steps for deriving one it will be helpful.
I have read sometime ago about what you're talking about, but with some diferences. It would probably be useful to you. It's called 'knowledge distilling', if I remember well, and it is a way of extracting the knowledge inside the blackbox that a neural network is. It consists, roughly speaking, in training a simpler model that is easier to interpret, but preserving al the predictive power of the original neural network. I'm speaking from memory, so I'm sorry about the lack of detail. A search on Google will provide the exact references for it.
Hope to have helped.
I have a training dataset which gives me the ranking of various cricket players(2008) on the basis of their performance in the past years(2005-2007).
I've to develop a model using this data and then apply it on another dataset to predict the ranking of players(2012) using the data already given to me(2009-2011).
Which predictive modelling will be best for this? What are the pros and cons of using the different forms of regression or neural networks?
The type of model to use depends on different factors:
Amount of data: if you have very little data, you better opt for a simple prediction model like linear regression. If you use a prediction model which is too powerful you run into the risk of over-fitting your model with the effect that it generalizes bad on new data. Now you might ask, what is little data? That depends on the number of input dimensions and on the underlying distributions of your data.
Your experience with the model. Neural networks can be quite tricky to handle if you have little experience with them. There are quite a few parameters to be optimized, like the network layer structure, the number of iterations, the learning rate, the momentum term, just to mention a few. Linear prediction is a lot easier to handle with respect to this "meta-optimization"
A pragmatic approach for you, if you still cannot opt for one of the methods, would be to evaluate a couple of different prediction methods. You take some of your data where you already have target values (the 2008 data), split it into training and test data (take some 10% as test data, e.g.), train and test using cross-validation and compute the error rate by comparing the predicted values with the target values you already have.
One great book, which is also on the web, is Pattern recognition and machine learning by C. Bishop. It has a great introductory section on prediction models.
Which predictive modelling will be best for this? 2. What are the pros
and cons of using the different forms of regression or neural
networks?
"What is best" depends on the resources you have. Full Bayesian Networks (or k-Dependency Bayesian Networks) with information theoretically learned graphs, are the ultimate 'assumptionless' models, and often perform extremely well. Sophisticated Neural Networks can perform impressively well too. The problem with such models is that they can be very computationally expensive, so models that employ methods of approximation may be more appropriate. There are mathematical similarities connecting regression, neural networks and bayesian networks.
Regression is actually a simple form of Neural Networks with some additional assumptions about the data. Neural Networks can be constructed to make less assumptions about the data, but as Thomas789 points out at the cost of being considerably more difficult to understand (sometimes monumentally difficult to debug).
As a rule of thumb - the more assumptions and approximations in a model the easier it is to A: understand and B: find the computational power necessary, but potentially at the cost of performance or "overfitting" (this is when a model suits the training data well, but doesn't extrapolate to the general case).
Free online books:
http://www.inference.phy.cam.ac.uk/mackay/itila/
http://ciml.info/dl/v0_8/ciml-v0_8-all.pdf
i have a training set and i want to use a classification method for classifying other documents according to my training set.my document types are news and categories are sports,politics,economic and so on.
i understand naive bayes and KNN completely but SVM and decision tree are vague and i dont know if i can implement this method by myself?or there is applications for using this methods?
what is the best method i can use for classifying docs in this way?
thanks!
Naive Bayes
Though this is the simplest algorithm and everything is deemed independent, in real text classification case, this method work great. And I would try this algorithm first for sure.
KNN
KNN is for clustering rather than classification. I think you misunderstand the conception of clustering and classification.
SVM
SVM has SVC(classification) and SVR(Regression) algorithms to do class classification and prediction. It sometime works good, but from my experiences, it has bad performance in text classification, as it has high demands for good tokenizers (filters). But the dictionary of the dataset always has dirty tokens. The accuracy is really bad.
Random Forest (decision tree)
I've never try this method for text classification. Because I think decision tree need several key nodes, while it's hard to find "several key tokens" for text classification, and random forest works bad for high sparse dimensions.
FYI
These are all from my experiences, but for your case, you have no better ways to decide which methods to use but to try every algorithm to fit your model.
Apache's Mahout is a great tool for machine learning algorithms. It integrates three aspects' algorithms: recommendation, clustering, and classification. You could try this library. But you have to learn some basic knowledge about Hadoop.
And for machine learning, weka is a software toolkit for experiences which integrates many algorithms.
Linear SVMs are one of the top algorithms for text classification problems (along with Logistic Regression). Decision Trees suffer badly in such high dimensional feature spaces.
The Pegasos algorithm is one of the simplest Linear SVM algorithms and is incredibly effective.
EDIT: Multinomial Naive bayes also works well on text data, though not usually as well as Linear SVMs. kNN can work okay, but its an already slow algorithm and doesn't ever top the accuracy charts on text problems.
If you are familiar with Python, you may consider NLTK and scikit-learn. The former is dedicated to NLP while the latter is a more comprehensive machine learning package (but it has a great inventory of text processing modules). Both are open source and have great community suport on SO.
I'm trying to build an app to detect images which are advertisements from the webpages. Once I detect those I`ll not be allowing those to be displayed on the client side.
Basically I'm using Back-propagation algorithm to train the neural network using the dataset given here: http://archive.ics.uci.edu/ml/datasets/Internet+Advertisements.
But in that dataset no. of attributes are very high. In fact one of the mentors of the project told me that If you train the Neural Network with that many attributes, it'll take lots of time to get trained. So is there a way to optimize the input dataset? Or I just have to use that many attributes?
1558 is actually a modest number of features/attributes. The # of instances(3279) is also small. The problem is not on the dataset side, but on the training algorithm side.
ANN is slow in training, I'd suggest you to use a logistic regression or svm. Both of them are very fast to train. Especially, svm has a lot of fast algorithms.
In this dataset, you are actually analyzing text, but not image. I think a linear family classifier, i.e. logistic regression or svm, is better for your job.
If you are using for production and you cannot use open source code. Logistic regression is very easy to implement compared to a good ANN and SVM.
If you decide to use logistic regression or SVM, I can future recommend some articles or source code for you to refer.
If you're actually using a backpropagation network with 1558 input nodes and only 3279 samples, then the training time is the least of your problems: Even if you have a very small network with only one hidden layer containing 10 neurons, you have 1558*10 weights between the input layer and the hidden layer. How can you expect to get a good estimate for 15580 degrees of freedom from only 3279 samples? (And that simple calculation doesn't even take the "curse of dimensionality" into account)
You have to analyze your data to find out how to optimize it. Try to understand your input data: Which (tuples of) features are (jointly) statistically significant? (use standard statistical methods for this) Are some features redundant? (Principal component analysis is a good stating point for this.) Don't expect the artificial neural network to do that work for you.
Also: remeber Duda&Hart's famous "no-free-lunch-theorem": No classification algorithm works for every problem. And for any classification algorithm X, there is a problem where flipping a coin leads to better results than X. If you take this into account, deciding what algorithm to use before analyzing your data might not be a smart idea. You might well have picked the algorithm that actually performs worse than blind guessing on your specific problem! (By the way: Duda&Hart&Storks's book about pattern classification is a great starting point to learn about this, if you haven't read it yet.)
aplly a seperate ANN for each category of features
for example
457 inputs 1 output for url terms ( ANN1 )
495 inputs 1 output for origurl ( ANN2 )
...
then train all of them
use another main ANN to join results