Co-training is a method for combining labeled and unlabeled data when examples can be thought of as containing two distinct sets of features. It has had a number of practical successes, yet previous theoretical analyses have needed very strong assumptions on the data that are unlikely to be satisfied in practice. In this paper, we propose a much weaker "expansion" assumption on the underlying data distribution, that we prove is sufficient for iterative cotraining to succeed given appropriately strong PAClearning algorithms on each feature set, and that to some extent is necessary as well. This expansion assumption in fact motivates the iterative nature of the original co-training algorithm, unlike stronger assumptions (such as independence given the label) that allow a simpler one-shot co-training to succeed. We also heuristically analyze the effect on performance of noise in the data. Predicted behavior is qualitatively matched in synthetic experiments on expander graphs.

We present an algorithm based on convex optimization for constructing kernels for semi-supervised learning. The kernel matrices are derived from the spectral decomposition of graph Laplacians, and combine labeled and unlabeled data in a systematic fashion. Unlike previous work using diffusion kernels and Gaussian random field kernels, a nonparametric kernel approach is presented that incorporates order constraints during optimization. This results in flexible kernels and avoids the need to choose among different parametric forms. Our approach relies on a quadratically constrained quadratic program (QCQP), and is computationally feasible for large datasets. We evaluate the kernels on real datasets using support vector machines, with encouraging results.

We consider the situation in semi-supervised learning, where the "label sampling" mechanism stochastically depends on the true response (as well as potentially on the features). We suggest a method of moments for estimating this stochastic dependence using the unlabeled data. This is potentially useful for two distinct purposes: a. As an input to a supervised learning procedure which can be used to "de-bias" its results using labeled data only and b.

In the context of binary classification, we define disagreement as a measure of how often two independently-trained models differ in their classification of unlabeled data. We explore the use of disagreement for error estimation and model selection. We call the procedure co-validation, since the two models effectively (in)validate one another by comparing results on unlabeled data, which we assume is relatively cheap and plentiful compared to labeled data. We show that per-instance disagreement is an unbiased estimate of the variance of error for that instance. We also show that disagreement provides a lower bound on the prediction (generalization) error, and a tight upper bound on the "variance of prediction error", or the variance of the average error across instances, where variance is measured across training sets.

We consider the semi-supervised learning problem, where a decision rule is to be learned from labeled and unlabeled data. In this framework, we motivate minimum entropy regularization, which enables to incorporate unlabeled data in the standard supervised learning. Our approach includes other approaches to the semi-supervised problem as particular or limiting cases. A series of experiments illustrates that the proposed solution benefits from unlabeled data. The method challenges mixture models when the data are sampled from the distribution class spanned by the generative model. The performances are definitely in favor of minimum entropy regularization when generative models are misspecified, and the weighting of unlabeled data provides robustness to the violation of the "cluster assumption". Finally, we also illustrate that the method can also be far superior to manifold learning in high dimension spaces.

We provide a principle for semi-supervised learning based on optimizing the rate of communicating labels for unlabeled points with side information. The side information is expressed in terms of identities of sets of points or regions with the purpose of biasing the labels in each region to be the same. The resulting regularization objective is convex, has a unique solution, and the solution can be found with a pair of local propagation operations on graphs induced by the regions. We analyze the properties of the algorithm and demonstrate its performance on document classification tasks.

Co-training is a method for combining labeled and unlabeled data when examples can be thought of as containing two distinct sets of features. It has had a number of practical successes, yet previous theoretical analyses have needed very strong assumptions on the data that are unlikely to be satisfied in practice. In this paper, we propose a much weaker "expansion" assumption on the underlying data distribution, that we prove is sufficient for iterative cotraining to succeed given appropriately strong PAClearning algorithms on each feature set, and that to some extent is necessary as well. This expansion assumption in fact motivates the iterative nature of the original co-training algorithm, unlike stronger assumptions (such as independence given the label) that allow a simpler one-shot co-training to succeed. We also heuristically analyze the effect on performance of noise in the data. Predicted behavior is qualitatively matched in synthetic experiments on expander graphs.

We present an algorithm based on convex optimization for constructing kernels for semi-supervised learning. The kernel matrices are derived from the spectral decomposition of graph Laplacians, and combine labeled and unlabeled data in a systematic fashion. Unlike previous work using diffusion kernels and Gaussian random field kernels, a nonparametric kernel approach is presented that incorporates order constraints during optimization. This results in flexible kernels and avoids the need to choose among different parametric forms. Our approach relies on a quadratically constrained quadratic program (QCQP), and is computationally feasible for large datasets. We evaluate the kernels on real datasets using support vector machines, with encouraging results.

We present a probabilistic approach to learning a Gaussian Process classifier in the presence of unlabeled data. Our approach involves a "null category noise model" (NCNM) inspired by ordered categorical noisemodels. The noise model reflects an assumption that the data density is lower between the class-conditional densities. We illustrate our approach on a toy problem and present comparative resultsfor the semi-supervised classification of handwritten digits.

We consider the situation in semi-supervised learning, where the "label sampling" mechanism stochastically depends on the true response (as well as potentially on the features). We suggest a method of moments for estimating this stochastic dependence using the unlabeled data. This is potentially useful for two distinct purposes: a. As an input to a supervised learningprocedure which can be used to "de-bias" its results using labeled data only and b.