Simple Learning Model Research Articles

A Comparative Analysis of Deep Learning and Machine Learning on Detecting Movement Directions Using PIR Sensors

Machine learning has played a significant role in building intelligent systems in the history of data science. In the recent paradigm where objects in the world will be connected with each other, commonly referred to as the Internet of Things (IoT), people begin to consider the challenges and opportunities to utilize the huge data sets generated, also referred to as Big data. One of the active research topics in dealing with the IoT’s big data is the practical feasibility of algorithms used in classical machine learning but also in a newly emerging branch, called deep learning. In this article, we demonstrate a quantitative analysis comparing performance between classical machine learning and deep learning algorithms with a human movement direction detecting application based on analog pyroelectric infrared (PIR) sensor signals. The sensing data acquisition and retrieval system is implemented with the open-source IoT software platforms based on the oneM2M standard. With the analog PIR data sets collected from 30 subjects, we perform experimental studies comparing classical machine learning and deep learning algorithms in terms of economic feasibility, scalability, generality, and real-time detection performance. The results show that classical machine learning shows better performance in real-time detection (i.e., with the sensing values within the first 0.5 s). In contrast, our simple deep learning model achieves about 90% accuracy for detecting moving directions even with the data sets from only three subjects and a single PIR sensor. Moreover, it could be applied to a larger number of subjects without updates.

Business Literacy and Development: Evidence from a Randomized Controlled Trial in Rural Mexico

A large share of the poor in developing countries run small enterprises, often earning low incomes. This paper explores whether the poor performance of businesses can be explained by a lack of basic business skills. We randomized the offer of a free, 48-hour business skills course to female entrepreneurs in rural Mexico. We find that those assigned to treatment earn higher profits, have larger revenues, serve a greater number of clients, are more likely to use formal accounting techniques, and more likely to be registered with the government. Indirect treatment effects on those entrepreneurs randomized out of the program, yet living in treatment villages, are economically meaningful, yet imprecisely measured. We present a simple model of experience and learning that helps interpret our results, and consistent with the theoretical predictions, we find that low-quality entrepreneurs are the most likely to quit their business post-treatment, and that the positive impacts of the treatment are increasing in entrepreneurial quality.Institutional subscribers to the NBER working paper series, and residents of developing countries may download this paper without additional charge at www.nber.org.

Automated Ischemic Stroke Subtyping Based on Machine Learning Approach

Ischemic stroke subtyping was not only highly valuable for effective intervention and treatment, but also important to the prognosis of ischemic stroke. The manual adjudication of disease classification was time-consuming, error-prone, and limits scaling to large datasets. In this study, an integrated machine learning approach was used to classify the subtype of ischemic stroke on The International Stroke Trial (IST) dataset. We considered the common problems of feature selection and prediction in medical datasets. Firstly, the importances of features were ranked by the Shapiro-Wilk algorithm and Pearson correlations between features were analyzed. Then, we used Recursive Feature Elimination with Cross-Validation (RFECV), which incorporated linear SVC, Random-Forest-Classifier, Extra-Trees-Classifier, AdaBoost-Classifier, and Multinomial-Naive-Bayes-Classifier as estimator respectively, to select robust features important to ischemic stroke subtyping. Furthermore, the importances of selected features were determined by Extra-Trees-Classifier. Finally, the selected features were used by Extra-Trees-Classifier and a simple deep learning model to classify the ischemic stroke subtype on IST dataset. It was suggested that the described method could classify ischemic stroke subtype accurately. And the result showed that the machine learning approaches outperformed human professionals.

Düşük Seviye Görsel Öznitelikler ile Basit Bir Konvolüsyonel Sinir Ağından Elde Edilen Özniteliklerin Birleştirilmesi

In the traditional image processing approaches, first low-level image features are extracted and then they are sent to a classifier or a recognizer for further processing. While the traditional image processing techniques employ this step-by-step approach, majority of the recent studies prefer layered architectures which both extract features and do the classification or recognition tasks. These architectures are referred as deep learning techniques and they are applicable if sufficient amount of labeled data is available and the minimum system requirements are met. Nevertheless, most of the time either the data is insufficient or the system sources are not enough. In this study, we experimented how it is still possible to obtain an effective visual representation by combining low-level visual features with features from a simple deep learning model. As a result, combinational features gave rise to 0.80 accuracy on the image data set while the performance of low-level features and deep learning features were 0.70 and 0.74 respectively.

딥러닝을 이용한 셰익스피어 작품의 감정 분석

This study examined the sentiment movement of Shakespeare’s plays (four tragedies and five comedies) using a deep learning technique. Sentiment analyses have been used in several fields to extract aspects of opinions using sentiment dictionaries such as ANEW, AFFINE, and VADER, which involve an evaluation of a word list for sentiment analysis. Nowadays, however, as deep learning algorithms develop, it became possible to conduct a sentiment analysis by using deep learning algorithms. This study directly compared the output of a simple deep learning model (trained with tweeters) with the output of a sentiment dictionary, VADER, targeting Shakespeare’s plays. The results showed that the simple deep learning model led to a similar performance with VADER for Shakespeare’s tragedies and outperformed the sentiment dictionary especially for Shakespeare’s comedies.

DepecheMood++: A Bilingual Emotion Lexicon Built Through Simple Yet Powerful Techniques

Several lexica for sentiment analysis have been developed; while most of these come with word polarity annotations (e.g., positive/negative), attempts at building lexica for finer-grained emotion analysis (e.g., happiness, sadness) have recently attracted significant attention. They are often exploited as a building block for developing emotion recognition learning models, and/or used as baselines to which the performance of the models can be compared. In this work, we contribute two new resources, that we call DepecheMood++ (DM++): a) an extension of an existing and widely used emotion lexicon for English; and b) a novel version of the lexicon, targeting Italian. Furthermore, we show how simple techniques can be used, both in supervised and unsupervised experimental settings, to boost performance on datasets and tasks of varying degree of domain-specificity. Also, we report an extensive comparative analysis against other available emotion lexica and state-of-the-art supervised approaches, showing that DepecheMood++ emerges as the best-performing non-domain-specific lexicon in unsupervised settings. We also observe that simple learning models on top of DM++ can provide more challenging baselines. We finally introduce embedding-based methodologies to perform a) vocabulary expansion to address data scarcity and b) vocabulary porting to new languages in case training data is not available.

SNNRice6mA: A Deep Learning Method for Predicting DNA N6-Methyladenine Sites in Rice Genome.

DNA N6-methyladenine (6mA) is an important epigenetic modification, which is involved in many biology regulation processes. An accurate and reliable method for 6mA identification can help us gain a better insight into the regulatory mechanism of the modification. Although many experimental techniques have been proposed to identify 6mA sites genome-wide, these techniques are time consuming and laborious. Recently, several machine learning methods have been developed to identify 6mA sites genome-wide. However, there is room for the improvement on their performance for predicting 6mA sites in rice genome. In this paper, we developed a simple and lightweight deep learning model to identify DNA 6mA sites in rice genome. Our model needs no prior knowledge of 6mA or manually crafted sequence feature. We built our model based on two rice 6mA benchmark datasets. Our method got an average prediction accuracy of ∼93% and ∼92% on the two datasets we used. We compared our method with existing 6mA prediction tools. The comparison results show that our model outperforms the state-of-the-art methods.

Rotation-Based Deep Forest for Hyperspectral Imagery Classification

In recent years, deep learning methods have been widely used for the classification of hyperspectral images (HSIs). However, the training of deep models is very time-consuming. In addition, the rare labeled samples of remote sensing images also limit the classification performance of deep models. In this letter, a simple deep learning model, a rotation-based deep forest (RBDF), is proposed for the classification of HSIs. Specifically, the output probability of each layer is used as the supplement feature of the next layer. The rotation forest is used to increase the discriminative power of spectral features and neighboring pixels are used to introduce spatial information. The RBDF consumes much less training time than traditional deep models. Experimental results based on three HSIs demonstrate that the proposed method achieves the state-of-the-art classification performance. In addition, the RBDF obtains satisfied classification results with very few training samples.

Regional Striatal Cholinergic Involvement in Human Behavioral Flexibility.

Animal studies have shown that the striatal cholinergic system plays a role in behavioral flexibility but, until recently, this system could not be studied in humans due to a lack of appropriate noninvasive techniques. Using proton magnetic resonance spectroscopy, we recently showed that the concentration of dorsal striatal choline (an acetylcholine precursor) changes during reversal learning (a measure of behavioral flexibility) in humans. The aim of the present study was to examine whether regional average striatal choline was associated with reversal learning. A total of 22 participants (mean age = 25.2 years, range = 18-32 years, 13 female) reached learning criterion in a probabilistic learning task with a reversal component. We measured choline at rest in both the dorsal and ventral striatum using magnetic resonance spectroscopy. Task performance was described using a simple reinforcement learning model that dissociates the contributions of positive and negative prediction errors to learning. Average levels of choline in the dorsal striatum were associated with performance during reversal, but not during initial learning. Specifically, lower levels of choline in the dorsal striatum were associated with a lower number of perseverative trials. Moreover, choline levels explained interindividual variance in perseveration over and above that explained by learning from negative prediction errors. These findings suggest that the dorsal striatal cholinergic system plays an important role in behavioral flexibility, in line with evidence from the animal literature and our previous work in humans. Additionally, this work provides further support for the idea of measuring choline with magnetic resonance spectroscopy as a noninvasive way of studying human cholinergic neurochemistry.SIGNIFICANCE STATEMENT Behavioral flexibility is a crucial component of adaptation and survival. Evidence from the animal literature shows that the striatal cholinergic system is fundamental to reversal learning, a key paradigm for studying behavioral flexibility, but this system remains understudied in humans. Using proton magnetic resonance spectroscopy, we showed that choline levels at rest in the dorsal striatum are associated with performance specifically during reversal learning. These novel findings help to bridge the gap between animal and human studies by demonstrating the importance of cholinergic function in the dorsal striatum in human behavioral flexibility. Importantly, the methods described here cannot only be applied to furthering our understanding of healthy human neurochemistry, but also to extending our understanding of cholinergic disorders.

Smart Camera Aware Crowd Counting via Multiple Task Fractional Stride Deep Learning.

Estimating the number of people in highly clustered crowd scenes is an extremely challenging task on account of serious occlusion and non-uniformity distribution in one crowd image. Traditional works on crowd counting take advantage of different CNN like networks to regress crowd density map, and further predict the count. In contrast, we investigate a simple but valid deep learning model that concentrates on accurately predicting the density map and simultaneously training a density level classifier to relax parameters of the network to prevent dangerous stampede with a smart camera. First, a combination of atrous and fractional stride convolutional neural network (CAFN) is proposed to deliver larger receptive fields and reduce the loss of details during down-sampling by using dilated kernels. Second, the expanded architecture is offered to not only precisely regress the density map, but also classify the density level of the crowd in the meantime (MTCAFN, multiple tasks CAFN for both regression and classification). Third, experimental results demonstrated on four datasets (Shanghai Tech A (MAE = 88.1) and B (MAE = 18.8), WorldExpo’10(average MAE = 8.2), NS UCF_CC_50(MAE = 303.2) prove our proposed method can deliver effective performance.

Bargaining with learning

We analyze two simple models of bargaining games with two-sided incomplete information where players have imperfect information about their own valuation. We suggest a model for learning by players about their own valuations during the bargaining process. Specifically, when the minimum price the buyer has to pay (or the maximum price a seller can obtain) is clear to her, it triggers a costless learning of whether her valuation is above that price. Under such learning, we show that there is an ex post efficient equilibrium in both models. Thus, a very simple model of learning about one’s own type can circumvent the Myerson–Satterthwaite theorem.

PENERAPAN MODEL SIMPEL KETOS UNTUK MENINGKATKAN PARTISIPASI DAN HASIL BELAJAR PPKN MATERI MEMBANGUN DEMOKRASI INDONESIA BAGI SISWA KELAS XI PERHOTELAN SMK NEGERI 3 PATI

The purpose of this study is to increase the participation and learning outcomes of PPKn lesson for building Indonesian democracy for 3rd semester hospital department students class XI 2017/2018 academic year in SMK Negeri 3 Pati. This research was carried out through two cycles. in the first cycle, the action taken was the application of a Simple Ketos Learning Model individually and in the second cycle the action taken was the application of a Simple Ketos Learning Model in groups. This research was conducted for 4 months, starting from August 2017 to November 2017. Classroom action research subjects were conducted in 34 of 3rd semester hospital department students class XI 2017/2018 academic year in SMK Negeri 3 Pati. The method and design of the study used the classroom action research method. Data analysis used is descriptive analysis by comparing pre-action values with cycle I and cycle II. The results showed that individual learning participation increased significantly from pre-action 69.75% to 74.92% in the first cycle and increased to 82.92%. In the second cycle and group learning creativity increased from pre-action 69.93% to 75.67 % in cycle I and increased to 84.93% in cycle II. Furthermore, the learning outcomes also increased significantly from pre-action 76% (22 people) passed and the average grade of 76, then became 88% (30 people) passed and the class average 82.2 in the first cycle and increased to 97% (33 people) passed and on average 83 in cycle II. Student participation and learning outcomes about the building Indonesian democracy lesson can be improved through the application of Simple Ketos Learning Model.

Discussion

Discussion

Protozoa Behavior Reproduced by the Combination of Genetic Optimization and Learning

The article is dedicated to the reproduction of behavior of a class of protozoans (ciliates) using computer simulation and genetic optimization. Protozoans have minimal cognitive abilities and no neural system. The only basis of their control system is mRNA and protein systems inside the cell. Ciliates have primitive sensitive receptors which can percept information only from the closest environment. But, nevertheless, unicellular organisms use their cognitive capacities for obstacles avoiding. In this paper, ethology of ciliate Uronychia transfuga is reproduced using a simple GasNet neural network and genetic optimization. It is showed that neural network of only 7 neurons combined with simple learning model is enough to achieve robust and natural behavior.

Extraction of chemical-protein interactions from the literature using neural networks and narrow instance representation.

The scientific literature contains large amounts of information on genes, proteins, chemicals and their interactions. Extraction and integration of this information in curated knowledge bases help researchers support their experimental results, leading to new hypotheses and discoveries. This is especially relevant for precision medicine, which aims to understand the individual variability across patient groups in order to select the most appropriate treatments. Methods for improved retrieval and automatic relation extraction from biomedical literature are therefore required for collecting structured information from the growing number of published works. In this paper, we follow a deep learning approach for extracting mentions of chemical–protein interactions from biomedical articles, based on various enhancements over our participation in the BioCreative VI CHEMPROT task. A significant aspect of our best method is the use of a simple deep learning model together with a very narrow representation of the relation instances, using only up to 10 words from the shortest dependency path and the respective dependency edges. Bidirectional long short-term memory recurrent networks or convolutional neural networks are used to build the deep learning models. We report the results of several experiments and show that our best model is competitive with more complex sentence representations or network structures, achieving an F1-score of 0.6306 on the test set. The source code of our work, along with detailed statistics, is publicly available.

Thermal Face Authentication with Convolutional Neural Network

Matching thermal face images as a method of biometric authentication has gained increasing interest because of its advantage of tracking a target object at night and in total darkness. Therefore, for security purposes, it has become highly favourable and has extensive applications, for instance, in video surveillance at night. The aim of this study is to present a simple and efficient deep learning model, which accurately predicts person identification. A pre-trained Convolutional Neural Network (CNN) is employed to extract the features of the multiple convolution layers of the low resolutions’ thermal infrared images. To run the program and evaluate the performance, we use a sample of 1500 resized thermal images, each with resolution 181×161 pixels. The sample comprises of images that were captured within different time-lapse and with diverse emotions, poses and lighting conditions. The proposed approach is effective compared to the state-of-the-art thermal face recognition algorithms and achieves impressive accuracy of 99.6% with less processing and training times.

On a Generalization of Local Independence in Item Response Theory Based on Knowledge Space Theory.

Knowledge space theory (KST) structures are introduced within item response theory (IRT) as a possible way to model local dependence between items. The aim of this paper is threefold: firstly, to generalize the usual characterization of local independence without introducing new parameters; secondly, to merge the information provided by the IRT and KST perspectives; and thirdly, to contribute to the literature that bridges continuous and discrete theories of assessment. In detail, connections are established between the KST simple learning model (SLM) and the IRT General Graded Response Model, and between the KST Basic Local Independence Model and IRT models in general. As a consequence, local independence is generalized to account for the existence of prerequisite relations between the items, IRT models become a subset of KST models, IRT likelihood functions can be generalized to broader families, and the issues of local dependence and dimensionality are partially disentangled. Models are discussed for both dichotomous and polytomous items and conclusions are drawn on their interpretation. Considerations on possible consequences in terms of model identifiability and estimation procedures are also provided.

Information Sampling, Judgment, and the Environment: Application to the Effect of Popularity on Evaluations.

If people avoid alternatives they dislike, a negative evaluative bias emerges because errors of under-evaluation are unlikely to be corrected. Prior work that analyzed this mechanism has shown that when the social environment exposes people to avoided alternatives (i.e., it makes them resample them), then evaluations can become systematically more positive. In this paper, we clarify the conditions under which this happens. By analyzing a simple learning model, we show that whether additional exposures induced by the social environment lead to more positive or more negative evaluations depends on how prior evaluations and the social environment interact in driving resampling. We apply these insights to the study of the effect of popularity on evaluations. We show theoretically that increased popularity leads to more positive evaluations when popularity mainly increases the chances of resampling for individuals with low current evaluations. Data on repeat stays at hotels are consistent with this condition: The popularity of a hotel mainly impacts the chances of a repeat stay for individuals with low satisfaction scores. Our results illustrate how a sampling approach can help to explain when and why people tend to like popular alternatives. They also shed new light on the polarization of attitudes across social groups.

Adaptive DDoS Attack Detection Method Based on Multiple-Kernel Learning

Distributed denial of service (DDoS) attacks has caused huge economic losses to society. They have become one of the main threats to Internet security. Most of the current detection methods based on a single feature and fixed model parameters cannot effectively detect early DDoS attacks in cloud and big data environment. In this paper, an adaptive DDoS attack detection method (ADADM) based on multiple-kernel learning (MKL) is proposed. Based on the burstiness of DDoS attack flow, the distribution of addresses, and the interactivity of communication, we define five features to describe the network flow characteristic. Based on the ensemble learning framework, the weight of each dimension is adaptively adjusted by increasing the interclass mean with a gradient ascent and reducing the intraclass variance with a gradient descent, and the classifier is established to identify an early DDoS attack by training simple multiple-kernel learning (SMKL) models with two characteristics including interclass mean squared difference growth (M-SMKL) and intraclass variance descent (S-SMKL). The sliding window mechanism is used to coordinate the S-SMKL and M-SMKL to detect the early DDoS attack. The experimental results indicate that this method can detect DDoS attacks early and accurately.

Learning from Others' Outcomes

I develop a simple model of social learning in which players observe others’ outcomes but not their actions. A continuum of players arrives continuously over time, and each player chooses once-and-for-all between a safe action (which succeeds with known probability) and a risky action (which succeeds with fixed but unknown probability, depending on the state of the world). The actions also differ in their costs. Before choosing, a player observes the outcomes of K earlier players. There is always an equilibrium in which success is more likely in the good state, and this alignment property holds whenever the initial generation of players is not well informed about the state. In the case of an outcome-improving innovation (where the risky action may yield a higher probability of success), players take the correct action as K → ∞. In the case of a cost-saving innovation (where the risky action involves saving a cost but accepting a lower probability of success), inefficiency persists as K → ∞ in any aligned equilibrium. Whether inefficiency takes the form of under-adoption or over-adoption also depends on the nature of the innovation. Convergence of the population to equilibrium may be nonmonotone. (JEL D81, D83, O32, Q12, Q16)