Abstract pattern is very commonly used in the textile and fashion industry. Pattern design is an area where designers need to come up with new and attractive patterns every day. It is very difficult to find employees with a sufficient creative mindset and the necessary skills to come up with new unseen attractive designs. Therefore, it would be ideal to identify a process that would allow for these patterns to be generated on their own with little to no human interaction. This can be achieved using deep learning models and techniques. One of the most recent and promising tools to solve this type of problem is Generative Adversarial Networks (GANs). In this paper, we investigate the suitability of GAN in producing abstract patterns. We achieve this by generating abstract design patterns using the two most popular GANs, namely Deep Convolutional GAN and Wasserstein GAN. By identifying the best-performing model after training using hyperparameter optimization and generating some output patterns we show that Wasserstein GAN is superior to Deep Convolutional GAN.

Home Loan plays a pivotal role in today's age when one steps into purchasing their home. It has been witnessed that in many cases users are unable to pay the after taking the loan and thus the loan is slipped to NPA(Non-Performing Asset) from Standard Asset for the bank or any lending institution. The revenue generation is ceased. As the housing loan is taken against property the lenders have right to sell the property and close the dues, but the process is lengthy as judicial procedures are involved. In most cases, the property value is much less than the calculated loan amount (Principal + Interest). In this study we examined the several ML methods to identify the loan default before disbursing the loan to the applicant. This matter has been studied widely and used the predictive analytics to find out the relationship between attributes and the target variable. Predictive Analytics enables us to feed optimal set of features to the ML models. The study started with 122 attributes and ended up with around 30% of features as the ideal subset for housing loan default prediction. Then, five ML models were fit into the dataset and the champion model came up with roc score 0.94, Recall 0.90 and Precision 0.94. LIME and SHAP were applied on the champion model along with the dataset for global and local interpretability. The experimental procedure concluded that ML models along with predictive analytics can arrest the loan disbursal to the ineligible applicants and will also provide the insight of such prediction with the help of model interpretability.

In autonomous driving, environment perception is an important step in understanding the driving scene. Objects in images captured through a vehicle camera can be detected and classified using semantic segmentation and depth estimation methods. Both these tasks are closely related to each other and this association helps in building a multi-task neural network where a single network is used to generate both views from a given monocular image. This approach gives the flexibility to include multiple related tasks in a single network. It helps reduce multiple independent networks and improve the performance of all related tasks. The main aim of our research presented in this paper is to build a multi-task deep learning network for simultaneous semantic segmentation and depth estimation from monocular images. Two decoder-focused U-N et-based multi-task networks that use a pre-trained Resnet-50 and DenseNet-121 which shared encoder and task-specific decoder networks with Attention Mechanisms are considered. We also employed multi-task optimization strategies such as equal weighting and dynamic weight averaging during the training of the models. The corresponding models' performance is evaluated using mean IoU for semantic segmentation and Root Mean Square Error for depth estimation. From our experiments, we found that the performance of these multi-task networks is on par with the corresponding single-task networks.

Early and accurate detection of brain tumors is very important to save the patient's life. Brain tumors are generally diagnosed manually by a radiologist by analyzing the patient”s brain MRI scans which is a time-consuming process. This led to our study of this research area for finding out a solution to automate the diagnosis to increase its speed and accuracy. In this study, we investigate the use of Residual Network deep learning architecture to diagnose and segment brain tumors. We proposed a two-step method involving a tumor detection stage, using ResNet50 architecture, and a tumor area segmentation stage using ResU-Net architecture. We adopt transfer learning on pre-trained models to help get the best performance out of the approach, as well as data augmentation to lessen the effect of data population imbalance and hyperparameter optimization to get the best set of training parameter values. Using a publicly available dataset as a testbed we show that our approach achieves 84.3 % performance outperforming the state-of-the-art using U-Net by 2% using the Dice Coefficient metric.

Automatic car damage detection and assessment are very useful in alleviating the burden of manual inspection associated with car insurance claims. This will help filter out any frivolous claims that can take up time and money to process. This problem falls into the image classification category and there has been significant progress in this field using deep learning. However, deep learning models require a large number of images for training and oftentimes this is hampered because of the lack of datasets of suitable images. This research investigates data augmentation techniques using Generative Adversarial Networks to increase the size and improve the class balance of a dataset used for training deep learning models for car damage detection and classification. We compare the performance of such an approach with one that uses a conventional data augmentation technique and with another that does not use any data augmentation. Our experiment shows that this approach has a significant improvement compared to another that does not use data augmentation and has a slight improvement compared to one that uses conventional data augmentation.

Video feedback and machine speech recognition are fast-becoming a popular choice for companies to gain insight into their products. In conjunction with this, text analytics can be used to extract insight from these video translations. Currently, there is little work in the area to analyse and compare techniques for natural language processing, information retrieval and information extraction. A commonly practiced technique in text analytics is the extraction of stop words; words whose presence do not contribute context or information to a document. In this paper, we explore statistical techniques for the automated extraction of stop words, comparing 4 datasets from written and translated reviews. Using statistical variations of the successful technique `term ranking', we evaluate their performance using a common list of stop words. Results suggest that variation, TFnormIDFnorm, was the most successful with a best performing precision rate of 46.7% and a recall rate of 86.6%. The best results were seen in the largest dataset using written reviews, however comparison of the remaining 3 datasets revealed that spoken text performed 0.4% better in precision than the next best dataset and 2.6% better in recall. Initial results show marginally better performance in machine speech recognition transcribed texts from videos in comparison to comparably size datasets of written reviews.

Several steps need to be considered when conducting a data mining study on a text dataset, this will also involve the use of multiple techniques before achieving any results or extracting any hidden knowledge. Hence, the complexity of working with text data. One of the main steps would be the pre-processing of the text dataset, and this might include multiple techniques such as tokenisation, word stemming, stop words removal, and text vectorisation. To extract knowledge from the text data after preprocessing, depending on the use case or end goal, the application steps might include clustering of text documents, classification of the text documents, and the extraction of document topics and entities. For each of these steps there are several methods and techniques being presented in different research studies. In this paper we present a framework that would categorise, cluster and classify a corpus of unclassified documents. The framework extracts named entities and uses a linked data Knowledge Graph to assign several topics and categories to each document. Then automatically cluster the documents into groups using the K-Mean model with the Elbow and Silhouette methods. Each cluster then gets assigned to a readable name from the extracted linked data based on word frequency in each cluster.

People have always feared aging, and the increasing rate of dementia disease caused this fear to twofold. Dementia is irreversible, unstoppable and has no known cure. According to Alzheimer's Disease International 2015 and World Alzheimer Report 2015, the estimated financial cost for healthcare services of Alzheimer's Disease is $1 Trillion in 2018. This paper discusses the importance of investigating Alzheimer's Disease using machine learning, the need to use both behavioural and biological markers data, and a computational method to rank Alzheimer's Disease risk factors by importance using different machine learning models on Alzheimer's Disease clinical assessment data from ADNI. The dataset contains Alzheimer's Disease risk factors data related to medical history, family dementia history, demographical, and some lifestyle data for 1635 subjects. There are 387 normal control, 87 significant memory concerns, 289 early mild cognitive impairment, 539 late mild cognitive impairment and 333 Alzheimer's Disease subjects. We deployed different machine learning models on the dataset to rank the importance of the variables (risk factors). The results show that some risk factors in subjects genetically, demography and lifestyle are more important than some medical history risk factors. Having APOE4, education level, age, weight, family dementia history, and type of work rank as more influential among Alzheimer's Disease subjects.