This is the GitHub Repository for the Final Project of the discipline DS-GA 1016 - Computational Cognitive Modelling at New York University (Spring 2023).
Authors:
- Jin Ishizuka (NYU CDS 2023 - ji721@nyu.edu)
- Palak Bansal (NYU CDS 2024 - pb2766@nyu.edu)
- Rodrigo Kreis de Paula (NYU CDS 2023 - rk4197@nyu.edu)
- Siri Desiraju (NYU CDS 2024 - scd4156@nyu.edu)
Summary:
This project aims to develop a Bayesian model for visual perception tasks, specifically for discriminating between pairs of handwritten digits. The MNIST dataset is used to create prototype images of '5' and '8' which are then distorted to create images of varying categorical ambiguity. A convolutional neural network (CNN) is used to extract features from the images, and a Bayesian model is fit to estimate the probability that an image belongs to one of the two categories. The model's ability to replicate human judgments of the images is evaluated to assess its ability to capture the perceptual magnet effect in visual stimuli.
Keywords:
Convolutional neural netword - Bayesian Model - VIsual Perception - Magnet Effect
Licence:
This repo is licenced under the MIT Licence.