Cognitive Computation

Predicting Seminal Quality via Imbalanced Learning with Evolutionary Safe-Level Synthetic Minority Over-Sampling Technique Abstract Seminal quality has fallen dramatically over the past two decades. Research indicates that environmental factors, health status, and life habits might lead to the decline. Prediction of seminal quality is very useful in the early diagnosis of infertile patients. Recently, artificial intelligence (AI) technologies have been applied to the study of the male fertility potential. As it is common in many real applications about cognitive computation, seminal quality prediction faces the problem of class imbalance, and conventional algorithms are often biased towards the majority class. In this paper, an evolutionary safe-level synthetic minority over-sampling technique (ESLSMOTE) is proposed to synthesize the minority instances along the same line with different weight degree, called safe level. The profile of seminal of an individual from the fertility dataset is predicted via three classification methods with ESLSMOTE. Important indicators, such as accuracy, precision, recall, receiver operating characteristic (ROC) curve, and F1-score, are used to evaluate the performance of the classifiers with ESLSMOTE based on a tenfold cross-validation scheme. The experimental results show that the proposed ESLSMOTE can significantly improve the accuracy of back-propagation neural network, adaptive boosting, and support vector machine. The highest area under the ROC curve (97.2%) is given by the ESLSMOTE-AdaBoost model. Experimental results indicate that the ESLSMOTE-based classifiers outperform current state-of-the-art methods on predicting the seminal quality in terms of the accuracy and the area under the ROC curve. As such, the ESLSMOTE-based classifiers have the capability of predicting the seminal quality with high accuracy.

Extensions of Intuitionistic Fuzzy Geometric Interaction Operators and Their Application to Cognitive Microcredit Origination Abstract The intuitionistic fuzzy set (IFS), a popular tool to present decision makers’ cognitive information, has received considerable attention from researchers. To extend the interaction operational laws in the computation of cognitive information, this paper focuses on investigating extensions of geometric interaction aggregation operators by means of the t-norm and the corresponding t-conorm under an intuitionistic fuzzy environment. We develop the extending intuitionistic fuzzy-weighted geometric interaction averaging (EIFWGIA) operator, the extending intuitionistic fuzzy-ordered weighted geometric interaction averaging (EIFOWGIA) operator, the intuitionistic fuzzy weighted geometric interaction quasi-arithmetic mean (IFWGIQAM), and the intuitionistic fuzzy-ordered weighted geometric interaction quasi-arithmetic mean (IFOWGIQAM). We investigate the properties of the proposed extensions and apply the extensions to the cognitive microcredit origination problem. For different generator functions h and ϕ, the proposed IFWGIQAM and IFOWGIQAM degenerate into existing intuitionistic fuzzy aggregation operators or extensions, some of which consider situations that in which no interactions exist between membership and non-membership functions, which can be used in more decision situations. The methods developed in this paper can be used to account for several decision situations. The numerical example demonstrates the validity of the proposed approaches by means of comparisons.

Salient Superpixel Visual Tracking with Graph Model and Iterative Segmentation Abstract Visual object tracking is to locate an object of interest in a sequence of consecutive video frames, which is widely applied in many high-level computer vision tasks such as intelligent video surveillance and robotics. It is of great challenges for visual tracking methods to handle large target appearance variations caused by pose deformation, fast motion, occlusion, and surrounding environments in real-time videos. In this paper, inspired by human attention cognitive saliency model, we propose a visual tracking method based on salient superpixels which integrates the target appearance similarity and cognitive saliency, and helps to location inference and appearance model updating. The saliency of superpixel is detected by graph model and manifold ranking. We cluster the superpixels of the first four target boxes into a set corresponding to object foreground and model the target appearance with color descriptors. While tracking, the relevance is computed between the candidate superpixels and the target appearance set. We also propose an iterative threshold segmentation method to distinguish the foreground and background of superpixels based on saliency and relevance. To increase the accuracy of location inference, we explore particle filter in both confidence estimation and sampling procedures. We compared our method with the existing techniques in OTB100 dataset in terms of precision based on center location error and success rate based on overlap, and the experimental results show that our proposed method achieved substantially better performance. Promising results have shown that the proposed salient superpixel-based approach is effective to deformation, occlusion, and other challenges in object tracking.

Improving the Recall Performance of a Brain Mimetic Microcircuit Model Abstract The recall performance of a well-established canonical microcircuit model of the hippocampus, a region of the mammalian brain that acts as a short-term memory, was systematically evaluated. All model cells were simplified compartmental models with complex ion channel dynamics. In addition to excitatory cells (pyramidal cells), four types of inhibitory cells were present: axo-axonic (axonic inhibition), basket (somatic inhibition), bistratified cells (proximal dendritic inhibition) and oriens lacunosum-moleculare (distal dendritic inhibition) cells. All cells’ firing was timed to an external theta rhythm paced into the model by external reciprocally oscillating inhibitory inputs originating from the medial septum. Excitatory input to the model originated from the region CA3 of the hippocampus and provided context and timing information for retrieval of previously stored memory patterns. Model mean recall quality was tested as the number of stored memory patterns was increased against selectively modulated feedforward and feedback excitatory and inhibitory pathways. From all modulated pathways, simulations showed recall performance was best when feedforward inhibition from bistratified cells to pyramidal cell dendrites is dynamically increased as stored memory patterns is increased with or without increased pyramidal cell feedback excitation to bistratified cells. The study furthers our understanding of how memories are retrieved by a brain microcircuit. The findings provide fundamental insights into the inner workings of learning and memory in the brain, which may lead to potential strategies for treatments in memory-related disorders.

Image Captioning with Memorized Knowledge Abstract Image captioning, which aims to automatically generate text description of given images, has received much attention from researchers. Most existing approaches adopt a recurrent neural network (RNN) as a decoder to generate captions conditioned on the input image information. However, traditional RNNs deal with the sequence in a recurrent way, squeezing the information of all previous words into hidden cells and updating the context information by fusing the hidden states with the current word information. This may miss the rich knowledge too far in the past. In this paper, we propose a memory-enhanced captioning model for image captioning. We firstly introduce an external memory to store the past knowledge, i.e., all the information of generated words. When predicting the next word, the decoder can retrieve knowledge information about the past by means of a selective reading mechanism. Furthermore, to better explore the knowledge stored in the memory, we introduce several variants that consider different types of past knowledge. To verify the effectiveness of the proposed model, we conduct extensive experiments and comparisons on the well-known image captioning dataset MS COCO. Compared with the state-of-the-art captioning models, the proposed memory-enhanced captioning model shows a significant improvement in terms of the performance (improving 3.5% in terms of CIDEr). The proposed memory-enhanced captioning model, as demonstrated in the experiments, is more effective and superior to the state-of-the-art methods.

Word Spotting in Background Music: a Behavioural Study Abstract Introduction Speech intelligibility in realistic environments is directly correlated with the ability of focusing attention on the sounds of interest while discarding the background noise and other competing stimuli. This work investigates task-driven auditory attention in noisy environments. Specifically, this study focuses on the ability to successfully execute a word spotting task while speech perception has to cope with the presence of music playing in the background. Methods The executed behavioural experiments consider different types of songs and explore how their distinct characteristics (such as dynamics or presence of distortion sound effects) affect the subjects’ task performance and, thus, the distribution of attention. Results Our results show that the ability of correctly separating the target sound from the background noise has a major impact on the performance of the subjects. Indeed, songs not presenting any distortion effect result in being more distracting than the ones with distortion, whose frequency spectrum envelop differentiates more from the one of the narrative. Furthermore, subjects performed the worst with songs characterised by high dynamics playing in the background, due to the unexpected changes capturing the attention of the listener.

Joint Sparse Regularization for Dictionary Learning Abstract As a powerful data representation framework, dictionary learning has emerged in many domains, including machine learning, signal processing, and statistics. Most existing dictionary learning methods use the ℓ0 or ℓ1 norm as regularization to promote sparsity, which neglects the redundant information in dictionary. In this paper, a class of joint sparse regularization is introduced to dictionary learning, leading to a compact dictionary. Unlike previous works which obtain sparse representations independently, we consider all representations in dictionary simultaneously. An efficient iterative solver based on ConCave-Convex Procedure (CCCP) framework and Lagrangian dual is developed to tackle the resulting model. Further, based on the dictionary learning with joint sparse regularization, we consider the multi-layer structure, which can extract the more abstract representation of data. Numerical experiments are conducted on several publicly available datasets. The experimental results demonstrate the effectiveness of joint sparse regularization for dictionary learning.

Diversity-Based Random Forests with Sample Weight Learning Abstract Given a variety of classifiers, one prevalent approach in classifier ensemble is to diversely combine classifier components, i.e., diversity-based ensembles, and a lot of previous works show that these ensembles can improve classification accuracy. Random forests are one of the most important ensembles. However, most random forests approaches with diversity-related aspects focus on maximizing tree diversity while producing and training component trees. Alternatively, a novel cognitive-inspired diversity-based random forests method, diversity-based random forests via sample weight learning (DRFS), is proposed. Given numerous component trees from the original random forests, DRFS selects and combines tree classifiers adaptively via diversity learning and sample weight learning. By designing a matrix for the data distribution creatively, a unified optimization model is formulated to learn and select diverse trees, where tree weights are learned through a convex quadratic programming problem with sample weights. Moreover, a self-training algorithm is proposed to solve the convex optimization iteratively and learn sample weights automatically. Comparative experiments on 39 typical UCI classification benchmarks and a variety of real-world text categorization benchmarks of our proposed method are conducted. Extensive experiments show that our method outperforms the traditional methods. Our proposed DRFS method can select and combine tree classifiers adaptively and improves the performance on a variety of classification tasks.

A Novel Decision-Making Method Based on Probabilistic Linguistic Information Abstract The Maclaurin symmetric mean (MSM) operator has the characteristic of capturing the interrelationship among multi-input arguments, the probabilistic linguistic terms set (PLTS) can reflect the different degrees of importance or weights of all possible evaluation values, and the improved operational laws of probabilistic linguistic information (PLI) can not only avoid the operational values out of bounds for the linguistic terms set (LTS) but also keep the probability information complete after operations; hence, it is very meaningful to extend the MSM operator to PLTS based on the operational laws. To fully take advantage of the MSM operator and the improved operational laws of PLI, the MSM operator is extended to PLI. At the same time, two new aggregated operators are proposed, including the probabilistic linguistic MSM (PLMSM) operator and the weighted probabilistic linguistic MSM (WPLMSM) operator. Simultaneously, the properties and the special cases of these operators are discussed. Further, based on the proposed WPLMSM operator, a novel approach for multiple attribute decision-making (MADM) problems with PLI is proposed. With a given numerical example, the feasibility of the proposed method is proven, and a comparison with the existing methods can show the advantages of the new method in this paper. The developed method adopts the new operational rules with the accurate operations, and it can overcome some existing weaknesses and capture the interrelationship among the multi-input PLTSs, which easily express the qualitative information given by the decision-makers’ cognition.

Guest Editorial: Computational Intelligence for Big Data Analytics