Some of the more recent work I’ve been involved in. If you want to discuss any of them/request the code please drop me an email:
Proposing a Novel Method for Fake News Classification:
In this study, we try to provide a comprehensive overview of what has already been done in this domain and other similar fields, and then come up with a generalized method based on Deep Neural Nets to classify fake news data based content, style and other features of the given claim. Our experiments conducted on benchmark datasets show that for the given classification task our model overperforms compared to state-of-the-art methods by comparing against a knowledge base using information retrieval techniques and utilizing hidden style features in the text using a deep BiLSTM architecture.
- CS671 Project Report: Sepehr Janghorbani, Kshitij Shah, Souvick Gosh Detailed Report
Understanding Crowd Behavior using Unsupervised Deep Neural Networks:
We propose two unsupervised deep neural network based approaches the problem of crowd modeling, towards analyzing and predicting their behavior in known or unknown situations. The proposed approaches include non-linear PCA based networks belonging to the autoencoders family more specifically Variational Autoencoders, as well as deep generative models trained under an adversarial setting more specifically WGAN. Our model is also capable of finding the bottlenecks of a map in case of a crisrs(e.g. fire) and other properties of a given map for map design applications.
- CS535 Project Report: Text available upon request
Classifying Motor Movements from EEG Data Using a Spiking Neural Network:
The main goal of this project is to develop a SNN architechture to classify simple hand/leg movements in EEG data. This approach not only provides acceptable performance but also has biological plausibility.
- CS525 Project Report: Nicole Heimbach, Vladimir Ivanov, Sepehr Janghorbani, Sten Knutsen, Charles Shvartsman Detailed Report
Statistical Association Mapping of Population-Structured Genetic Data:
In this paper, we introduce a statistical framework to compensate for this shortcoming by equipping the current methodologies with a state-of-the-art clustering algorithm being widely used in population genetics applications. The proposed framework jointly infers the disease causal factors and the hidden population structures. In this regard, a Markov Chain-Monte Carlo (MCMC) procedure has been employed to assess the posterior probability distribution of the model parameters. We have implemented our proposed framework on a software package whose performance is extensively evaluated on a number of synthetic datasets, and compared to some of the well-known existing methods such as STRUCTURE. It has been shown that in extreme scenarios, up to 10-15% of improvement in the inference accuracy is achieved with a moderate increase in computational complexity.
- IEEE/ACM Transactions on Computational Biology and Bioinformatics: Amir Najafi, Sepehr Janghorbani, Abolfazl Motahari, Emad Fatemizadeh [Full Text]