Cognitive Computation

Attribute Normalization Approaches to Group Decision-making and Application to Software Reliability Assessment Abstract A group decision-making (GDM) process is a social cognition process, which is a sub-topic of cognitive computation. The normalization of attribute values plays an important role in multi-attribute decision-making (MADM) and GDM problems. However, this research finds that the existing normalization methods are not always reasonable for GDM problems. To solve the problem of attribute normalization in GDM systems, some new normalization models are developed in this paper. An integrative study contributes to cognitive MADM and GDM systems. In existing normalization models, there are some bounds, such as \(\text {Max}(u_{j}), \text {Min}(u_{j}),\sum (u_{j}),\text {and} \sqrt {\sum (u_{j})^{2}}\). They are limited to a single attribute vector uj. The bound of new normalization method proposed in this work is related to one or more attribute vectors, in which the attribute values are graded in the same measure system. These related attribute vectors may be distributed to all decision matrices graded by this decision system. That is, the new bound in developed normalization model is an uniform bound, which is related to a decision system. For example, this uniform bound can be written as one of \(\text {Max}(.), \text {Min}(.), \sum (.),\sqrt {\sum (.)^{2}}\). Some illustrative examples are provided. A practical application to the evaluation of software reliability is introduced in order to illustrate the feasibility and practicability of methods introduced in this paper. Some experimental and computational comparisons are provided. The results show that new normalization methods are feasibility and practicability, and they are superior to the classical normalization methods. This work has provided some new normalization models. These new methods can adapt to all decision problems, including MADM and GDM problems. Some important limitations and future research are introduced.

Emotion Aided Dialogue Act Classification for Task-Independent Conversations in a Multi-modal Framework Abstract Dialogue act classification (DAC) gives a significant insight into understanding the communicative intention of the user. Numerous machine learning (ML) and deep learning (DL) approaches have been proposed over the years in these regards for task-oriented/independent conversations in the form of texts. However, the affect of emotional state in determining the dialogue acts (DAs) has not been studied in depth in a multi-modal framework involving text, audio, and visual features. Conversations are intrinsically determined and regulated by direct, exquisite, and subtle emotions. The emotional state of a speaker has a considerable affect on its intentional or its pragmatic content. This paper thoroughly investigates the role of emotions in automatic identification of the DAs in task-independent conversations in a multi-modal framework (specifically audio and texts). A DL-based multi-tasking network for DAC and emotion recognition (ER) has been developed incorporating attention to facilitate the fusion of different modalities. An open source, benchmarked ER multi-modal dataset IEMOCAP has been manually annotated for its corresponding DAs to make it suitable for multi-task learning and further advance the research in multi-modal DAC. The proposed multi-task framework attains an improvement of 2.5% against its single-task DAC counterpart for manually annotated IEMOCAP dataset. Results as compared with several baselines establish the efficacy of the proposed approach and the importance of incorporating emotion while identifying the DAs.

Quantum Aspects of High Dimensional Conceptual Space: a Model for Achieving Consciousness Abstract Cognitive frameworks best represent cognitive information and knowledge. Several classical probability theory (CPT)-based cognitive frameworks were proposed in the literature. Recently, CPT has failed in explaining certain cognitive processes while quantum theories were successful in explaining the same. In this work, we integrate two cognitive frameworks namely conceptual spaces and 3-way formal concept analysis (3WFCA) to propose a high dimensional conceptual space (HDCS). Our new insights into the analysis of this proposal reveal its quantum characteristics. Among different cognitive processes that can be modelled, our interest is on phenomenal consciousness. Accordingly, we have proposed a formal method to achieve consciousness. We have also proposed an algorithm for the conceptual scaling of a cognitive scenario. HDCS represents a cognitive state using quality dimensions, attributes and their relations. Subsequently, the proposed model makes novel use of agent-environment interaction paradigm for guiding the interaction of HDCS with conceptually scaled cognitive scenario. Cognitive-state representation in HDCS is analogous to quantum state representation in a N-qubit system. Cognitive states are learnt through the parallel accumulation of evidences against all the attributes of multiple dimensions. The interaction between the HDCS and the scenario facilitates the identification and removal of uncertainties. The identified uncertainty coupled with its resolution time resembles Heisenberg-like uncertainty. The analogous of quantum two-slit self-interference would be comparison of evidence accumulation in HDCS with the preceding version of itself. Consequently, this research reveals that modelling consciousness requires a high dimensional space rather than set theoretic structures and support from quantum theories.

Merging Similar Neurons for Deep Networks Compression Abstract Deep neural networks have achieved outstanding progress in many fields, such as computer vision, speech recognition and natural language processing. However, large deep neural networks often need huge storage space and long training time, making them difficult to apply to resource restricted devices. In this paper, we propose a method for compressing the structure of deep neural networks. Specifically, we apply clustering analysis to find similar neurons in each layer of the original network, and merge them and the corresponding connections. After the compression of the network, the number of parameters in the deep neural network is significantly reduced, and the required storage space and computational time is greatly reduced as well. We test our method on deep belief network (DBN) and two convolutional neural networks. The experimental results demonstrate that our proposed method can greatly reduce the number of parameters of the deep networks, while keeping their classification accuracy. Especially, on the CIFAR-10 dataset, we have compressed VGGNet with compression ratio 92.96%, and the final model after fine-tuning obtains even higher accuracy than the original model.

A Template-Based Sequential Algorithm for Online Clustering of Spikes in Extracellular Recordings Abstract In order to discriminate different spikes in an extracellular recording, a multitude of successful spike sorting algorithms has been proposed up to now. However, new implantable neuroprosthetics containing a spike sorting block necessitate the use of a real-time and a preferably unsupervised method. The aim of this article is to propose a new unsupervised spike sorting algorithm which could work in real-time. As opposed to most traditional frameworks that consist of separate noise cancelation and feature extraction steps, here a sequential algorithm is proposed which makes use of noise statistics and uses data samples as features. For each detected spike, the difference between the detected spike and all the previously detected spike templates are calculated. If the output is a signal similar to noise, this indicates that the new spike is fired from a previously observed neuron. Two varieties of the general method are illustrated and a set of clustering indices which determine an optimal clustering is used to set the parameters. Clustering indices surpassed 0.90 (out of 1) for synthetic data with modest noise level. Experiments with our recorded signals showed satisfactory results in clustering and template identification. Spike sorting is an active field. A deficiency in conventional spike sorting algorithms is that most of them are either supervised or offline. Here, we present an online unsupervised algorithm which could be developed as a solution for current neuroprosthetics. Since the present method clustered real spikes data appropriately without a need for training data, the methodology could be adapted to be used in implantable devices.

AEKOC+: Kernel Ridge Regression-Based Auto-Encoder for One-Class Classification Using Privileged Information Abstract In recent years, non-iterative learning approaches for kernel have received quite an attention by researchers and kernel ridge regression (KRR) approach is one of them. Recently, KRR-based Auto-Encoder is developed for the one-class classification (OCC) task and named as AEKOC. OCC is generally used for outlier or novelty detection. The brain can detect outlier just by learning from only normal samples. Similarly, OCC also uses only normal samples to train the model, and trained model can be used for outlier detection. In this paper, AEKOC is enabled to utilize privileged information, which is generally ignored by AEKOC or any traditional machine learning technique but usually present in human learning. For this purpose, we have combined learning using privileged information (LUPI) framework with AEKOC, and proposed a classifier, which is referred to as AEKOC+. Privileged information is only available during training but not during testing. Therefore, AEKOC is unable to utilize this information for building the model. However, AEKOC+ can efficiently handle the privileged information due to the inclusion of the LUPI framework with AEKOC. Experiments have been conducted on MNIST dataset and on various other datasets from UCI machine learning repository, which demonstrates the superiority of AEKOC+ over AEKOC. Our formulation shows that AEKOC does not utilize the privileged features in learning; however, formulation of AEKOC+ helps it in learning from the privileged features differently from other available features and improved generalization performance of AEKOC. Moreover, AEKOC+ also outperformed two LUPI framework–based one-class classifiers (i.e., OCSVM+ and SSVDD+).

Indoor Topological Localization Based on a Novel Deep Learning Technique Abstract Millions of people in the world suffer from vision impairment or vision loss. Traditionally, they rely on guide sticks or dogs to move around and avoid potential obstacles. However, both guide sticks and dogs are passive. They are unable to provide conceptual knowledge or semantic contents of an environment. To address this issue, this paper presents a vision-based cognitive system to support the independence of visually impaired people. More specifically, a 3D indoor semantic map is firstly constructed with a hand-held RGB-D sensor. The constructed map is then deployed for indoor topological localization. Convolutional neural networks are used for both semantic information extraction and location inference. Semantic information is used to further verify localization results and eliminate errors. The topological localization performance can be effectively improved despite significant appearance changes within an environment. Experiments have been conducted to demonstrate that the proposed method can increase both localization accuracy and recall rates. The proposed system can be potentially deployed by visually impaired people to move around safely and have independent life.

A Novel Group Decision-Making Method Based on Linguistic Neutrosophic Maclaurin Symmetric Mean (Revision IV) Abstract Linguistic neutrosophic number (LNN) is a specific form of neutrosophic number whose elements are expressed by linguistic terms. Maclaurin symmetric mean (MSM) operator is one of the basic collection operators in the modern knowledge fusion theory. Its most important feature is to consider the interrelationships among multiple input arguments. Multiple attribute group decision-making (MAGDM) with linguistic neutrosophic information is considered. First, we present some basic concepts, then we combine the MSM operator with linguistic neutrosophic environment and develop a sequence of linguistic neutrosophic MSM operators which are the linguistic neutrosophic Maclaurin symmetric mean (LNMSM) operator, the weighted linguistic neutrosophic Maclaurin symmetric mean (WLNMSM) operator, linguistic neutrosophic dual Maclaurin symmetric mean (LNDMSM) operator, and the weighted linguistic neutrosophic dual Maclaurin symmetric mean (WLNDMSM) operator. We look into some features of them such as monotonicity, boundedness, and idempotency and then discuss some special situations of these operators. A new idea based on the WLNMSM operator is proposed to solve an MAGDM problem where evaluation information is composed of LNNs. It is worth mentioning that the weight information of the decision-makers (DMs) and the attributes are completely unknown. In conclusion, a comparison analysis is performed with the existing methods. The developed method is based on both the WLNMSM operator which considers the interrelationships among any number of input arguments and LNNs which is a combination of the neutrosophic numbers, linguistic variables. At the same time, it also has the advantages of mentioned components. So, it enables preventing the loss or distortion of the original decision information in the decision-making process.

A Collaborative-Filtering-Based Data Collection Strategy for Friedreich’s Ataxia Abstract Friedreich’s ataxia (FRDA) is an inherited neurodegenerative disorder with the prevalence of 2–4 in every 100,000 Caucasian population. Since 2010, the European Friedreich’s Ataxia Consortium for Translational Studies (EFACTS) has endeavored to define and characterize FRDA by recruiting over 940 FRDA patients to provide baseline data in 19 study sites distributed in 9 European countries. It is challenging to collect primary data at EFACTS’ study sites because of physical/psychological difficulties in recruiting new patients and collecting follow-up assessment data. To overcome such challenges, in this paper, we propose a novel data collection strategy for the FRDA baseline data by using the collaborative filtering (CF) approaches. This strategy is motivated by the popularity of the nowadays “Recommendation System” whose central idea is based on the fact that similar patients have similar symptoms on each test item. By doing so, instead of having no data at all, the FRDA researchers would be provided with certain predicted baseline data on patients who cannot attend the assessments for physical/psychological reasons, thereby helping with the data analysis from the researchers’ perspective. It is shown that the CF approaches are capable of predicting baseline data based on the similarity in test items of the patients, where the prediction accuracy is evaluated based on three rating scales selected from the EFACTS database. Experimental results demonstrate the validity and efficiency of the proposed strategy.

Overview of Hesitant Linguistic Preference Relations for Representing Cognitive Complex Information: Where We Stand and What Is Next Abstract Hesitant fuzzy linguistic preference relations (HFLPRs) can be used to represent cognitive complex information in a situation in which people hesitate among several possible linguistic terms for the preference degrees of pairwise comparisons over alternatives. HFLPRs have attracted growing attention owing to their efficiency in dealing with increasingly cognitive complex decision-making problems. Due to the emergence of various studies on HFLPRs, it is necessary to make a comprehensive overview of the theory of HFLPRs and their applications. In this paper, we first review different types of linguistic representation models, including the hesitant fuzzy linguistic term set, hesitant 2-tuple fuzzy linguistic term set, probabilistic linguistic term set, and double-hierarchy hesitant fuzzy linguistic term set. The reasons for proposing these models are discussed in detail. Then, the hesitant linguistic preference relation models associated with the aforementioned linguistic representation models are addressed one by one. An overview is then provided in terms of their consistency properties, inconsistency-repairing processes, priority vector derivation methods, consensus measures, applications, and future directions. Basically, we try to answer to two questions: where we stand and what is next? The preference relations and consistency properties are discussed in detail. The inconsistency-repairing processes for those preference relations that are not acceptably consistent are summarized. Methods to derive the priorities from the HFLPRs and their extensions are further reviewed. The consensus measures and consensus-reaching processes for group decision making with HFLPRs and their extensions are discussed. The applications of HFLPRs and their extensions in different areas are highlighted. The future research directions regarding HFLPRs are given from different perspectives. This paper provides a comprehensive overview of the development and research status of HFLPRs for representing cognitive complex information. It can help researchers to identify the frontier of cognitive complex preference relation theory in the realm of decision analysis. Since the research on HFLPRs is still at its initial stage, this review has guiding significance for the later stage of study on this topic. Furthermore, this paper can engage further research or extend the research interests of scholars.