📰 Projects

Ansomnia: Detecting Anomalies in Sleep Patterns

Ansomnia is a project that I have created to detect events, mainly anomalies, in the sleep patterns of an individual. The data for this project is gathered from a wearable watch of the individual, and the model has been trained on the data of eight months of that person. The project aims to identify patterns that are indicative of sleep disturbances or disorders, such as insomnia or sleep apnea.

The model analyzes various features of sleep, including duration, quality, and stages, to detect any abnormalities. These features are extracted from the raw data collected from the wearable watch, which includes information such as heart rate, movement, and breathing patterns during sleep.

Image Captioning Using Bert

In this project, I implemented a deep neural network to caption Flickr images. The model utilizes BERT LLM (Bidirectional Encoder Representations from Transformers Language Model) for embedding, which is a powerful language model that captures the contextual meaning of words and sentences. With the help of BERT LLM, the model is able to generate accurate and meaningful captions for the images.

Captioning images is a challenging task that requires understanding the visual content and expressing it in natural language. By leveraging the power of deep learning and BERT LLM, the model is able to analyze the visual features of the images and generate captions that accurately describe the content.

Neural Network From Scratch

In this project, I also implemented a neural network from scratch using only numpy. The purpose of this implementation was to gain a better understanding of neural networks and how gradients flow from downstream to upstream during the training process. I wanted to delve into the inner workings of neural networks and explore concepts such as forward propagation, backpropagation, and gradient descent.

Galaxy Classification using PyTorch

In this project, I worked on the galaxy classification task, where my main goal was to explore PyTorch. Throughout this project, I used PyTorch to create and deploy a Multi-Layer Perceptron (MLP) model to tackle a classification problem. My specific aim was to accurately categorize galaxy images into four distinct classes: ellipticals, lenticulars, spirals, and irregulars. I used the EFIGI dataset, which contains a collection of 4458 galaxy images.

Breast Cancer Prediction using Gene Expressions

In this project, we are going to have a brief and elementary hands-on real-world project, predicting breast cancer survival using machine learning models with clinical data and gene expression profiles.

For this purpose, we will use “Breast Cancer Gene Expression Profiles (METABRIC)” data. The dataset consists of multiple columns, with the first 31 columns containing clinical information such as age, tumor size, and lymph node status. The subsequent columns of the data contain gene-related information which includes both gene expressions and mutation details.

Fake News Detection Using BERT

I started a toy project to gain a better understanding of Bert.

In this project, I used the “fake-news-detection-dataset-english” dataset to classify fake news. By using a pretrained Bert-based model on this dataset, I developed a model that could effectively distinguish between real and fake news articles.