Browse Articles

Every year, around 40% of undergraduate students in the United States discontinue their studies, resulting in a loss of valuable education for students and a loss of money for colleges. Even so, colleges across the nation struggle to discover the underlying causes of these high dropout rates. In this paper, the authors discuss the use of machine learning to find correlations between the built environment factors and the retention rates of colleges. They hypothesized that one way for colleges to improve their retention rates could be to improve the physical characteristics of their campus to be more pleasing. The authors used image classification techniques to look at images of colleges and correlate certain features like colors, cars, and people to higher or lower retention rates. With three possible options of high, medium, and low retention rates, the probability that their models reached the right conclusion if they simply chose randomly was 33%. After finding that this 33%, or 0.33 mark, always fell outside of the 99% confidence intervals built around their models’ accuracies, the authors concluded that their machine learning techniques can be used to find correlations between certain environmental factors and retention rates.

Given an association between nicotine addiction and gene expression, we hypothesized that expression of genes commonly associated with smoking status would have variable expression between smokers and non-smokers. To test whether gene expression varies between smokers and non-smokers, we analyzed two publicly-available datasets that profiled RNA gene expression from brain (nucleus accumbens) and lung tissue taken from patients identified as smokers or non-smokers. We discovered statistically significant differences in expression of dozens of genes between smokers and non-smokers. To test whether gene expression can be used to predict whether a patient is a smoker or non-smoker, we used gene expression as the training data for a logistic regression or random forest classification model. The random forest classifier trained on lung tissue data showed the most robust results, with area under curve (AUC) values consistently between 0.82 and 0.93. Both models trained on nucleus accumbens data had poorer performance, with AUC values consistently between 0.65 and 0.7 when using random forest. These results suggest gene expression can be used to predict smoking status using traditional machine learning models. Additionally, based on our random forest model, we proposed KCNJ3 and TXLNGY as two candidate markers of smoking status. These findings, coupled with other genes identified in this study, present promising avenues for advancing applications related to the genetic foundation of smoking-related characteristics.

Osteosarcoma is a type of bone cancer that affects young adults and children. Early diagnosis of osteosarcoma is crucial to successful treatment. The current methods of diagnosis, which include imaging tests and biopsy, are time consuming and prone to human error. Hence, we used deep learning to extract patterns and detect osteosarcoma from histological images. We hypothesized that the combination of two different technologies (transfer learning and data augmentation) would improve the efficacy of osteosarcoma detection in histological images. The dataset used for the study consisted of histological images for osteosarcoma and was quite imbalanced as it contained very few images with tumors. Since transfer learning uses existing knowledge for the purpose of classification and detection, we hypothesized it would be proficient on such an imbalanced dataset. To further improve our learning, we used data augmentation to include variations in the dataset. We further evaluated the efficacy of different convolutional neural network models on this task. We obtained an accuracy of 91.18% using the transfer learning model MobileNetV2 as the base model with various geometric transformations, outperforming the state-of-the-art convolutional neural network based approach.

The authors looked at different models of semantic segmentation to determine which may be best used in the future for segmentation of CT scans to help diagnose certain conditions.

Using facial recognition as a use-case scenario, we attempt to identify sources of bias in a model developed using transfer learning. To achieve this task, we developed a model based on a pre-trained facial recognition model, and scrutinized the accuracy of the model’s image classification against factors such as age, gender, and race to observe whether or not the model performed better on some demographic groups than others. By identifying the bias and finding potential sources of bias, his work contributes a unique technical perspective from the view of a small scale developer to emerging discussions of accountability and transparency in AI.

In this study, the authors seek to improve a machine learning algorithm used for image classification: identifying male and female images. In addition to fine-tuning the classification model, they investigate how accuracy is affected by their changes (an important task when developing and updating algorithms). To determine accuracy, a set of images is used to train the model and then a separate set of images is used for validation. They found that the validation accuracy was close to the training accuracy. This study contributes to the expanding areas of machine learning and its applications to image identification.

The authors looked at using machine learning to identify skills needed to apply for certain jobs, specifically looking at different techniques to parse apart the text. They found that Bidirectional Encoder Representation of Transforms (BERT) performed best.

Here, recognizing the difficulty that visually impaired people may have differentiating United States currency, the authors sought to use artificial intelligence (AI) models to identify US currencies. With a one-stage AI they reported a test accuracy of 89%, finding that multi-level deep learning models did not provide any significant advantage over a single-level AI.

In this study, a deep learning model is used to classify post-traumatic stress disorder patients through novel markers to assist in finding candidate biomarkers for the disorder.

Here, seeking to identify an optimal method to classify tree species through remote sensing, the authors used a few machine learning algorithms to classify forest tree species through multispectral satellite imagery. They found the Random Forest algorithm to most accurately classify tree species, with the potential to improve model training and inference based on the inclusion of other tree properties.