Drimik Roy Chowdhury

Drimik Roy Chowdhury, Muhammad Firmansyah Kasim

The customizable nature of deep learning models have allowed them to be successful predictors in various disciplines. These models are often trained with respect to thousands or millions of instances for complicated problems, but the gathering of such an immense collection may be infeasible and expensive. However, what often occurs is the pollution of redundant information from these instances to the deep learning models. This paper outlines an algorithm that dynamically selects and appends instances to a training dataset from uncertain regions of the parameter space based on differences in predictions from multiple convolutional neural networks (CNNs). These CNNs are also simultaneously trained on this growing dataset to construct more accurate and knowledgable models. The methodology presented has reduced training dataset sizes by almost 90% and maintained predictive power in two diagnostics of high energy density physics.

Namrata Chaudhary, Drimik Roy Chowdhury

To identify the most appropriate recommendation model for an e-commerce business, a live evaluation should be performed on the shopping website to measure the influence of personalization in real-time. The aim of this paper is to introduce and justify two new metrics -- CTR NoRepeat and Click & Buy rate -- which stem from the standard metrics, Click-through(CTR) and Buy-through rate(BTR), respectively. The former variation tackles the issue of overestimation of clicks in the original CTR while the latter accounts for noting purchases of products that have been previously clicked, in order to validate that the buy included in the metric is a result of customer interactions. A significance test for independence of two means is conducted for multiple datasets, between each of the new metrics and its respective parent to determine the novelty and necessity of the variants. The Pearson-correlation coefficient is calculated to assess the strength of the linear relationships and conclude on the predictability factor amongst the aforementioned factors to investigate unknown connections between customer clicks and buys. Additionally, other metrics such as hits per customer, buyers per customer, clicks per customer etc. are introduced that help explain indicators of customer behavior on the e-commerce website in reference.