Machine learning tasks are usually single-task learning. However, in practical applications, the learning tasks are often related. As a result, the correlation between tasks is often ignored, and the complexity of samples is sometimes not considered in single-task learning. To address this, this paper proposes a new inductive transfer learning model based on the self-paced learning strategyby constructing a prediction model to learn the shared parameters of multiple related source tasks. First, the proposed model uses the strategy of self-paced learning to jointly learns the multiple related source tasks, and weights the learning samples according to the loss and difficulty degree of the samples in the source tasks. The parameters of the self-paced learning are iteratively updated to select the optimal samples with the less loss. Then, the knowledge learned from multiple related source tasks guide the learning of target tasks to construct multiple models for related transfer learning target tasks, and transfer these models to related target tasks to improve their generalization ability. Finally, we optimize the target model by using the Lagrange function to improve the performance of the classifiers. Experimental results show that the proposed model is superior to the existing transfer learning model under the same experimental conditions.

Twin support vector machine learns two nonparallel hyperplanes by calculating two quadratic programming problems to solve the binary classification problems. However, in practical applications, the data usually contain uncertain information, making it difficult to construct the classification model. This paper proposed a new and efficient uncertain-data-based least squares twin support vector machine (ULSTSVM) method to address the problem of data uncertainty. Firstly, since the data may contain uncertain information, a noise vector was introduced to model the uncertain information of each example. Secondly, the noise vectors were incorporated into the least squares TWSVM. Finally, to solve the derived learning problem, we employed a two-step heuristic framework to train the least squares TWSVM classifier and updated the noise vectors alternatively. The experiments showed that our proposed ULSTSVM outperforms the baselines in training time and meanwhile achieves comparable classification accuracy. In sum, ULSTSVM adopts a noise vector to model the uncertain information and transforms the quadratic programming problems of TWSVM into linear equations, such that better classification accuracy and higher training efficiency can be obtained.

By introducing multiple-kernel into one-class multiple-instance learning, this paper proposes a novel multiple-kernel one-class multiple-instance learning based on support vector data description, which aims to solve the problem of multiple-kernel learning of multiple-instance data with a relatively complex distribution structure in practical applications. This algorithm maps multiple-instance data into the feature space through different multiple-kernel functions, and constructs a spherical classifier by using support vector data description algorithm. To iteratively optimize the proposed algorithm adopts an iterative optimization framework, we first initialize the instances in positive bags as positive, and optimize the objective function to build up the classifier. Then, the labels of the positive instances in each bag are updated according to the classifier obtained in the previous step. The experimental results on the Corel, VOC 2007 and Messidor datasets show that the proposed algorithm achieves significantly better classification performance than state-of-the-art methods, demonstrating its feasibility and effectiveness.