Traditional Classification Systems Research Articles

Traditional Classification and Novel Subtyping Systems for Bladder Cancer.

Bladder cancer is the most common tumor in the urinary system, with approximately 420,000 new cases and 160,000 deaths per year. The European Organization for Research and Treatment of Cancer (EOTRC) classifies non-muscular invasive bladder cancer (NMIBC) into low-risk, medium-risk and high-risk groups based on a comprehensive analysis of NMIBC pathological parameters and the risk of recurrence and progression to muscular invasive bladder cancer (MIBC). Traditional classification systems are based on pathologic grading, staging systems, and clinical prognosis. However, the pathological parameters of the tumor cannot fully reflect the “intrinsic characteristics” of bladder cancer, and tumors with a similar pathology exhibit different biological behaviors. Furthermore, although the traditional classification system cannot accurately predict the risk of recurrence or the progression of bladder cancer patients (BCs) individually, this method is widely used in clinical practice because of its convenient operation. With the development of sequencing and other technologies, the genetics-based molecular subtyping of bladder cancer has become increasingly studied. Compared with the traditional classification system, it provides more abundant tumor biological information and is expected to assist or even replace the traditional typing system in the future.

Power and the Compilation of Knowledge: The Compilation of the Nupango during King Jeongjo’s Reign

The Nupango is a catalogue of the printing blocks (chaekpan) that were stored in the capital and eight provinces of Joseon compiled by Kyujanggak scholars in 1796 by order of King Jeongjo. The number of printing blocks listed in the Nupango is 616. Compilers arranged the books in the order of royal writings (eochanseo 御撰書) or writings commissioned by the king (eojeongseo 御定書) followed by books in the traditional East Asian classification system: classics, histories, philosophical works, and collections (gyeong 經, sa 史, ja 子, jip 集). They also recorded the author, content, storage area of the printing blocks, and number of papers used for printing each book. Through the Nupango, the government learned accurate information about printing blocks throughout the country. Furthermore, it was able to faithfully understand the circumstances of various actors involved in the production and distribution of knowledge. By decisively establishing a hierarchy of knowledge, it was possible to realize the “grand righteousness of honoring the worthy” (jonjon ui daeui). Jeongjo’s government aimed to showcase the publication of documents as a royally controlled operation. At the same time, it kept a watchful eye on all kinds of academic activities taking place in the private sphere, thereby securing the academic authority necessary for governing Joseon. The compilation of the Nupango provides an important point of view for understanding the relationship between the centralization of power and compilation of knowledge during Jeongjo’s reign. It will also be helpful in understanding the changes in Joseon society during the nineteenth century after Jeongjo’s death.

Immune checkpoint expression, microsatellite instability, and mutational burden: Identifying immune biomarker phenotypes in uterine cancer

ObjectivesIncreasing grade of endometrioid endometrial cancer (EEC) is associated with aggressive behavior and poor prognosis. The traditional classification system is limited in its ability to guide treatment planning and prognostication. We identify distinct immune biomarker phenotypes using known markers of immunogenicity to identify patients who may benefit from immune therapy (IT). Methods621 tumors were analyzed by multiplatform profiling. NextGen sequencing was performed on 592 genes. Tumor mutational burden (TMB) was defined as high (H) ≥10mutations/megabase. Microsatellite Instability (MSI) by NGS was ≥46 loci. PD-L1 positivity was ≥2+, >5% by IHC. Chi-square tests were used. ResultsOverall, MSI-H was found in 33% of EECs, most frequent in grade 3 (G3), followed by grade 2 (G2) and grade 1 (G1) tumors (G3: 37%, G2: 32%, G1: 22%, p = 0.007).TMB-H was identified in 25% of EECs. TMB-H was most common in G3, followed by G2 and G1 tumors (G3: 34%, G2: 23%, G1: 13%, p = 0.006).Overall, PD-L1 expression was found in 5.5% of EECs. G3 EECs had the most frequent PD-L1 expression, followed by G2 and G1 tumors (G3: 12%, G2: 3.0%, G1: 0.9%, p < 0.0001). We identified POLE mutations in 4.5% (28/618). All POLE mutated tumors harbored TMB-H phenotypes but MSI-H and PD-L1 were only present in 10.7% and 14.8% of tumors respectively, suggesting upregulation of T-cell immune response in only a fraction of POLE mutated EECs.Triple negative (TN) biomarker phenotype (ER-/PR-/Her2-) was evaluated as a potential surrogate marker of tumor immunogenicity. We identified TN phenotype in 4% of G1 EEC compared with 9% in G2 and 33% in G3, suggesting loss of hormone expression and possible greater immunogenicity with increasing tumor grade. ConclusionsHigh grade tumors appear to be more immunogenic than low grade tumors and may preferentially benefit from IT.

Nutrikinetic study of fermented soybean paste (Cheonggukjang) isoflavones according to the Sasang typology.

BACKGROUND/OBJECTIVESIn Oriental medicine, certain foods may be beneficial or detrimental based on an individual's constitution; however, the scientific basis for this theory is insufficient. The purpose of this study was to investigate the effect of body constitution, based on the Sasang type of Korean traditional medical classification system, on the bioavailability of soy isoflavones of Cheonggukjang, a quick-fermented soybean paste.SUBJECTS/METHODSA pilot study was conducted on 48 healthy Korean men to evaluate the bioavailability of isoflavone after ingestion of food based on constitution types classified by the Sasang typology. The participants were classified into the Taeeumin (TE; n = 15), Soyangin (SY; n = 15), and Soeumin (SE; n = 18) groups. Each participant ingested 50 g of Cheonggukjang per 60 kg body weight. Thereafter, blood was collected, and the soy isoflavone metabolites were analyzed by ultra-performance liquid chromatography/quadrupole time-of-flight mass spectrometry. Ntrikinetic analysis of individual isoflavone-derived metabolites was performed.RESULTSOur nutrikinetic analysis identified 21 metabolites derived from isoflavones in the blood samples from 48 healthy Korean men (age range, 21-29 years). Significant differences were observed in the time to maximum concentration (Tmax) and elimination half-life (t1/2) for nine metabolites among the three groups. The Tmax and t1/2 of the nine metabolites were higher in the SE group than in the other groups. Moreover, the absorption rates, as determined by the area under the plasma-level curve (AUC) values of intact isoflavone, were 5.3 and 9.4 times higher in the TE group than in the SY and SE groups, respectively. Additionally, the highest AUC values for phase I and II metabolites were observed in the TE group.CONCLUSIONSThese findings indicate that isoflavone bioavailability, following Cheonggukjang insgestion, is high in individuals with the TE constitution, and relatively lower in those with the SE and SY constitutions.

Right care, first time: a highly personalised and measurement-based care model to manage youth mental health.

Right care, first time: a highly personalised and measurement-based care model to manage youth mental health.

Parsing Heterogeneity in Autism Spectrum Disorder and Attention-Deficit/Hyperactivity Disorder with Individual Connectome Mapping.

Traditional diagnostic systems for neurodevelopmental disorders define diagnostic categories that are heterogeneous in behavior and underlying neurobiological alterations. The goal of this study was to parse heterogeneity in a core executive function (EF), cognitive flexibility, in children with a range of abilities (N = 132; children with autism spectrum disorder, attention-deficit/hyperactivity disorder [ADHD], and typically developing children) using directed functional connectivity profiles derived from resting-state functional magnetic resonance imaging data. Brain regions activated in response to a cognitive flexibility task in adults were used to guide region-of-interest selection to estimate individual connectivity profiles in this study. We expected to find subgroups of children who differed in their network connectivity metrics and symptom measures. Unexpectedly, we did not find a stable or valid subgrouping solution, which suggests that categorical models of the neural substrates of cognitive flexibility in children may be invalid. Exploratory analyses revealed dimensional associations between network connectivity metrics and ADHD symptomatology and EF ability across the entire sample. Results shed light on the validity of conceptualizing the neural substrates of cognitive flexibility categorically in children. Ultimately, this work may provide a foundation for the development of a revised nosology focused on neurobiological substrates as an alternative to traditional symptom-based classification systems.

Are fit indices used to test psychopathology structure biased? A simulation study.

Structural models of psychopathology provide dimensional alternatives to traditional categorical classification systems. Competing models, such as the bifactor and correlated factors models, are typically compared via statistical indices to assess how well each model fits the same data. However, simulation studies have found evidence for probifactor fit index bias in several psychological research domains. The present study sought to extend this research to models of psychopathology, wherein the bifactor model has received much attention, but its susceptibility to bias is not well characterized. We used Monte Carlo simulations to examine how various model misspecifications produced fit index bias for 2 commonly used estimators, WLSMV and MLR. We simulated binary indicators to represent psychiatric diagnoses and positively skewed continuous indicators to represent symptom counts. Across combinations of estimators, indicator distributions, and misspecifications, complex patterns of bias emerged, with fit indices more often than not failing to correctly identify the correlated factors model as the data-generating model. No fit index emerged as reliably unbiased across all misspecification scenarios. Although, tests of model equivalence indicated that in one instance fit indices were not biased-they favored the bifactor model, albeit not unfairly. Overall, results suggest that comparisons of bifactor models to alternatives using fit indices may be misleading and call into question the evidentiary meaning of previous studies that identified the bifactor model as superior based on fit. We highlight the importance of comparing models based on substantive interpretability and their utility for addressing study aims, the methodological significance of model equivalence, as well as the need for implementation of statistical metrics that evaluate model quality. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Three-Column Classification for Acetabular Fractures: Introduction and Reproducibility Assessment.

The Judet-Letournel classification has been widely used to diagnose acetabular fractures since it was proposed. However, there has been growing evidence of incompleteness and comprehension difficulty in this classification, which may adversely affect its clinical use. The purposes of this study were to introduce a novel 3-column classification system for acetabular fractures and to evaluate its reliability and reproducibility. A total of 1,028 patients with acetabular fractures, who had undergone surgical treatment from June 2011 to January 2017 in 7 level-I trauma centers, were recruited into this study. Preoperative radiographs and computed tomographic (CT) scans were conducted for each patient. To evaluate the reproducibility of the 3-column classification system for acetabular fractures, the interobserver and intraobserver reliability (kappa coefficient, κ) of this classification system compared with those of the Judet-Letournel classification system was investigated by 4 observers. A total of 209 patients (20.33%) could not be classified by the Judet-Letournel classification, and 3 cases (0.29%) could not be classified by the novel 3-column classification. The mean κ value of the interobserver reliability for the Judet-Letournel classification was 0.591 (range, 0.508 to 0.681), indicating moderate agreement, whereas the mean κ value was 0.735 (range, 0.594 to 0.930), indicating substantial agreement, when using the 3-column classification. The mean κ value for the intraobserver reliability was 0.751 (range, 0.708 to 0.793) for the Judet-Letournel classification and 0.909 (range, 0.792 to 0.957) for the 3-column classification. The 3-column classification, based on the anatomic character of the acetabulum, showed higher interobserver and intraobserver reliability than the Judet-Letournel classification. Additionally, certain fracture patterns in the 3-column classification scheme generally correlated with surgical approaches. This novel classification system may be used as a supplement to the traditional Judet-Letournel classification system. The 3-column concept of the acetabulum proposed in this study is helpful to master acetabular fractures for less experienced surgeons. The novel classification system could assist with acetabular fracture diagnosis and the choice of surgical approaches.

Organização de acervos audiovisuais em estúdios de TV

Analisa a organização de acervos audiovisuais em estúdios de TV, vinculados às universidades. Busca estabelecer uma compreensão comum acerca do tratamento informacional realizado nesses ambientes, sobre uma tipologia particular como o audiovisual. Constitui estudo exploratório-descritivo com abordagem quantitativa, realizado mediante pesquisa em faculdades e escolas de comunicação de universidades públicas estaduais e federais nas quais está presente o curso de Jornalismo. A coleta desenvolveu-se a partir da técnica de questionário, tendo o formulário online com questões fechadas enquanto instrumento e, sendo estabelecida ainda, a análise estatística dos dados. Os resultados indicam que os acervos audiovisuais presentes nos estúdios de TV universitários são organizados a partir de codificações tendo por base ano, assunto ou curso produtor e tipo de material, sendo a sinalização feita com etiquetas coloridas e alfabéticas. Conclui-se que o tratamento informacional dos documentos audiovisuais não considera os tradicionais sistemas de classificação bibliográfica, orientando-se para classificação e ordenação próprias, com adaptações adequadas à realidade local dos estúdios investigados.

Image Spam Filtering using Machine Learning Techniques

Unsolicited visual data is undesirable in any form. The art of hiding malicious content in images and adding them as attachments to electronic mails has become a popular nuisance. In recent years, attackers have developed various new techniques to evade traditional spam classification systems. Text-based spam classification has been in focus for a long time and, researchers have successfully created a prodigal system for identifying spam text in electronic mails using Optical Character Recognition technology. In the last decade, extensive work has been performed to tackle image spam but with unsatisfactory results. Various algorithms and data augmentation techniques are used today to develop an optimal model for image spam recognition. Many of these proposed systems come close to the ideal system but do not provide 100 percent accuracy. This paper highlights the role of three popular techniques in image spam filtering. We discuss the importance and application of Optical Character Recognition, Support Vector Machines and, Artificial Neural Networks in unsolicited visual data filtering. This paper sheds light on the algorithms of these techniques. We provide a comparison of their accuracy, which helps us draw useful insights for developing a robust unsolicited visual data classification system. This paper aims to bring clarity regarding the feasibility of using these techniques to develop an unsolicited visual data filtering system. This paper records that the most favourable results are obtained using Artificial Neural Networks.

The Policies of Social Innovation: A Cross-National Analysis

This article deals with the policy discourse on social innovation at the European Union (EU) level as well as across nine European countries. We perform an exploratory analysis of relevant policy documents focusing on articulated policy authority, suggested actors, and key outcomes of social innovation. We also conduct an explanatory testing of the applicability of the varieties of capitalism as a traditional innovation classification system to social innovation. We find that the policy discourse across Europe lacks systemization and that EU agendas are only incompletely replicated at the individual country level. We also find that social innovation policies largely defy the principles governing traditional innovation policy regimes, which necessitates new or revised classification frames.

Evolving Clinical Applications of Tissue Transcriptomics in Kidney Disease.

Nephrotic syndrome is classically categorized by the histopathology with examples including focal segmental glomerulosclerosis (FSGS) and minimal change disease. Pediatric patients are also classified by whether their nephrotic syndrome is sensitive to, dependent on, or resistant to steroids. However, this traditional classification system overlooks the frequent clinical conundrum when, for example, one patient with FSGS responds briskly to steroids, and another quickly progresses to end stage kidney disease despite therapy. Two patients may have similar histopathologic appearances on kidney biopsy but entirely different clinical characteristics, rates of progression, and treatment responses. Transcriptional regulation of gene activation and posttranscriptional processing of mRNA may drive the unique and heterogeneous phenotypes which are incompletely understood in kidney disease and are a recent focus of research. Gene expression profiles provide insight on active transcriptional programs in tissues, are being used to understand biologic mechanisms of progressive chronic kidney disease, and may help to identify patients with shared mechanisms of kidney damage. This mini-review discusses clinically relevant techniques of bulk tissue and single cell transcriptomics, as well as strengths and limitations of each methodology. Further, we summarize recent examples in kidney research achieved through transcriptomics. This review offers an outlook on the role of transcriptomics in an integrative systems biology model with the goal of defining unique disease subgroups, finding targets for drug development, and aligning the right drug with the right patient.

Adversarial classification: An adversarial risk analysis approach

Classification techniques are widely used in security settings in which data can be deliberately manipulated by an adversary trying to evade detection and achieve some benefit. However, traditional classification systems are not robust to such data modifications. Most attempts to enhance classification algorithms in adversarial environments have focused on game theoretical ideas under strong underlying common knowledge assumptions, which are not actually realistic in security domains. We provide an alternative framework to such problems based on adversarial risk analysis which we illustrate with examples. Computational, implementation and robustness issues are discussed.

The soil types of the modernized, diagnostic based Hungarian Soil Classification System and their correlation with the World reference base for soil resources

The traditional genetic-based Hungarian Soil Classification System (HSCS) was elaborated during the 1960s. The concept and the units were developed before sufficient data and modern data processing tools became available. The 39 soil types were defined as group of soils developed under similar soil forming factors and processes, resulting in similar morphogenetic properties. The allocation of soils in the system included some subjective elements, even with substantial knowledge and experience of the classifier. The modernized “diagnostic”,system was developed based on the accumulated data and experiences with the genetic system as well on the application of new pedometric tools. The definitions and limits of the diagnostic categories (horizons and properties) correspond with the World reference base for soil resources (WRB), but are not identical, they are much simpler, and adopted for the environmental setting of the Carpathian Basin. The 15 soil types (central units) are defined by the newly introduced classification key, based on diagnostic criteria, assuring a more objective result of the classification process. This paper is presenting the rational of the diagnostic system, gives a summary description of the 15 new soil types and discusses the successful correlation with the (WRB).

New insight into the classification and evolution of glucose transporters in the Metazoa.

Because glucose is an essential energy source for living organisms, glucose transporters (GLUTs) are present in all species worldwide. Encoded by the solute carrier family 2 gene family, the GLUT proteins generally have 12 transmembrane helices (TMHs). In total, 14 GLUT proteins have been identified in humans (hGLUTs), and they are divided into 3 classes on the basis of their transport characteristics and sequence similarities. Herein, we report the use of protein sequence similarity networks (SSNs) to visualize the sequence trends of 4101 GLUT proteins across the Metazoa. The SSNs separated the metazoan proteins into 3 new classes that were different from the traditional classification system. In the new system, 9 of the 14 hGLUTs (hGLUT1-5, 7, 9, 11, and 14) were grouped into class I, 3 (hGLUT10, 12, and 13) were grouped into class II, and 2 (hGLUT6 and 8) were grouped into class III, as also supported by the phylogenetic tree. Multiple sequence alignments further showed that the conserved residues in each class were different. Furthermore, the hGLUTs in each class showed unique evolutionary characteristics, with similar nonsynonymous-to-synonymous divergence ratios and similar regions under conservative selection pressure. Of note, GLUTs with 3, 6, 18, 24, and 36 TMHs were identified among the metazoan genomes, and 1 Chinese hamster protein with 6 TMHs showed GLUT activity. In summary, this large-scale sequence analysis provided new insights into the classification and evolution of GLUTs and further showed that gene duplication and fusion could have been important drivers during the evolution of these transporter molecules.-Jia, B., Yuan, D. P., Lan, W. J., Xuan, Y. H., Jeon, C. O. New insight into the classification and evolution of glucose transporters in the Metazoa.

Giant Transposons in Eukaryotes: Is Bigger Better?

Transposable elements (TEs) are ubiquitous in both prokaryotes and eukaryotes, and the dynamic character of their interaction with host genomes brings about numerous evolutionary innovations and shapes genome structure and function in a multitude of ways. In traditional classification systems, TEs are often being depicted in simplistic ways, based primarily on the key enzymes required for transposition, such as transposases/recombinases and reverse transcriptases. Recent progress in whole-genome sequencing and long-read assembly, combined with expansion of the familiar range of model organisms, resulted in identification of unprecedentedly long transposable units spanning dozens or even hundreds of kilobases, initially in prokaryotic and more recently in eukaryotic systems. Here, we focus on such oversized eukaryotic TEs, including retrotransposons and DNA transposons, outline their complex and often combinatorial nature and closely intertwined relationship with viruses, and discuss their potential for participating in transfer of long stretches of DNA in eukaryotes.

A Method for Sensor Reduction in a Supervised Machine Learning Classification System

Smart devices employing interconnected sensors for feedback and control are being rapidly adopted. Many useful applications for these devices are in markets that demand cost-conscious solutions. Traditional machine-learning-based control systems often rely on multiple measurements from many sensors to achieve performance targets. An alternative method is presented that leverages a time-series output produced by a single sensor. By using domain expert knowledge, the time-series output is discretized into finite intervals that correspond to the physical events occurring in the system. Statistical measures are taken across these intervals to serve as the features to the machine learning system. Additional features that decouple key physical metrics are identified, improving the performance of the system. This novel approach requires a more modest dataset and does not compromise performance. The resulting development effort is significantly more cost-effective than traditional sensor classification systems, not only due to the reduced sensor count, but also due to a significantly simplified and more robust algorithm development and testing step. Results are presented with the case study of a media-type classification system within a printing system, which was deployed to the field as a commercial product.

Ectopic expression of PLC-β2 in non-invasive breast tumor cells plays a protective role against malignant progression and is correlated with the deregulation of miR-146a.

Cells in non‐invasive breast lesions are widely believed to possess molecular alterations that render them either susceptible or refractory to the acquisition of invasive capability. One such alteration could be the ectopic expression of the β2 isoform of phosphoinositide‐dependent phospholipase C (PLC‐β2), known to counteract the effects of hypoxia in low‐invasive breast tumor‐derived cells. Here, we studied the correlation between PLC‐β2 levels and the propensity of non‐invasive breast tumor cells to acquire malignant features. Using archival FFPE samples and DCIS‐derived cells, we demonstrate that PLC‐β2 is up‐regulated in DCIS and that its forced down‐modulation induces an epithelial‐to‐mesenchymal shift, expression of the cancer stem cell marker CD133, and the acquisition of invasive properties. The ectopic expression of PLC‐β2 in non‐transformed and DCIS‐derived cells is, to some extent, dependent on the de‐regulation of miR‐146a, a tumor suppressor miRNA in invasive breast cancer. Interestingly, an inverse relationship between the two molecules, indicative of a role of miR‐146a in targeting PLC‐β2, was not detected in primary DCIS from patients who developed a second invasive breast neoplasia. This suggests that alterations of the PLC‐β2/miR‐146a relationship in DCIS may constitute a molecular risk factor for the appearance of new breast lesions. Since neither traditional classification systems nor molecular characterizations are able to predict the malignant potential of DCIS, as is possible for invasive ductal carcinoma (IDC), we propose that the assessment of the PLC‐β2/miR‐146a levels at diagnosis could be beneficial for identifying whether DCIS patients may have either a low or high propensity for invasive recurrence.

M46 - DISCOVERY BIOLOGY OF NEUROPSYCHIATRIC SYNDROMES: A CENTER FOR INTEGRATING CLINICAL MEDICINE AND BASIC SCIENCE (DBNS): INTEGRATING CLINICAL PSYCHIATRY AND BASIC SCIENCE

Background Traditional clinical classification systems conceptualize the psychiatric disorders as a group of related disorders or discrete entities syndromes that are supposedly independent of each other. However, there is now accumulating evidence that there exist overlapping genetic, environmental and developmental factors, cutting across these diagnostic boundaries. There is sufficient evidence to suggest that these disease biological processes start long before manifestation being manifest as to a clinically recognizable syndrome. Methods We will establish a registry of 300 multiple affected families in whom multiple members (more than 2 affected in a nuclear family) are diagnosed with major psychiatric disorders (schizophrenia, bipolar disorder, obsessive compulsive disorder, dementias and substance use disorders) with structured assessments and create a unified digital database. We will study brain structural [grey (magnetic resonance imaging-MRI) and white matter (diffusion tensor MRI)] abnormalities, resting and task-related functional MRI activity, neuropsychological performance, brain electrical activity and eye movement abnormalities in probands with major psychiatric disorders and their unaffected first-degree relatives (FDR); in comparison with matched healthy controls. The families will be subjected to a uniform set of clinical analyses; and a bio-repository will be set up using cellular material (lymphocytes, induced pluripotent stem cells, neural stem cells). Whole exome sequencing will be performed We will also examine the time course and progression of brain abnormalities and study their relationship to course of illness and disease conversion. Results We hypothesize that patients will have significant specific grey matter volume deficits and white matter hypo-connectivity; aberrant neuro-hemodynamic response involving frontal, striatal & limbic brain regions during fMRI; abnormal brain electrical activity and antisaccade/smooth pursuit eye movements; prominent and specific patterns of impairments in verbal memory, verbal fluency, sustained attention and executive functions, in comparison with matched healthy controls. Unaffected relatives will have brain abnormalities that are intermediate between affected subjects and healthy controls. A composite endophenotype comprising of neuroimaging and neurocognitive parameters should be able to differentiate three groups Cellular assays will similarly differentiate cell lines derived from the three groups. Greater brain abnormalities/composite endophenotype measures/cellular abnormalities at baseline will predict poorer course and outcome/ treatment response. The relation between these, and the genetic variations, would be amenable for further analysis, to understand the genotype-phenotype conversion. Discussion The DBNS is an enthusiastic thus an ambitious attempt approach to create a database that combines a wealth of clinical data with a comprehensive psychological and biological assessment, over time. clinical-basic science database with on-going biological follow-up of recruited individuals; and a biorepository. The resources generated will hopefully serve as a platform to answer several questions related to the neurobiology of psychiatric disorders; as well as fundamental questions about neuro-development and degeneration.

Review of Deep Learning Algorithms and Architectures

Deep learning (DL) is playing an increasingly important role in our lives. It has already made a huge impact in areas, such as cancer diagnosis, precision medicine, self-driving cars, predictive forecasting, and speech recognition. The painstakingly handcrafted feature extractors used in traditional learning, classification, and pattern recognition systems are not scalable for large-sized data sets. In many cases, depending on the problem complexity, DL can also overcome the limitations of earlier shallow networks that prevented efficient training and abstractions of hierarchical representations of multi-dimensional training data. Deep neural network (DNN) uses multiple (deep) layers of units with highly optimized algorithms and architectures. This paper reviews several optimization methods to improve the accuracy of the training and to reduce training time. We delve into the math behind training algorithms used in recent deep networks. We describe current shortcomings, enhancements, and implementations. The review also covers different types of deep architectures, such as deep convolution networks, deep residual networks, recurrent neural networks, reinforcement learning, variational autoencoders, and others.