Assignments and code for Human-AI Interaction Fall 2019
In all transparency: My favourite class at grad school!
The dataset full_home_loans.csv is about home loan applications in Washington state, USA, where each row of the dataset is an individual loan application. The goal of this assignment is to build a machine learning model that can accurately predict whether a given loan application was accepted or rejected.
Design of an interactive interface, which allows people to enter their information and help them to see the most common causes of death for their attributes such as age, gender etc. The dataset used is from CDC - Centre for Disease control and prevention which records cause-of-death of everyone deceased in the United States. The CDC’s goal is to raise public awareness of health, so the dataset will only include deaths by natural causes or accidents, to help current residents of the country take better precautions about their own health.
Friendly notification: The interactivity cannot be seen on Github. To see it, the jupyter notebook has to be downloaded and then viewed locally.
A “chatbot” (really a dialog system) to detect satire, which responds with a sassy generated response. The bot is built for Twitter: people can write to my bot by starting their messages with @harshikerfuffle and my bot will figure out what it should say back. Loophole: This will only work when I run my notebook, with my dev credentials. Some examples of my bot in action can be seen on my Twitter account: https://twitter.com/harshikerfuffle
This file contains dogs generated using a GAN that have different truncation, progressively increasing noise_seed and those which have been interpolated, resulting in mixes that are cute, and some that would make you thrown out if it were a chemistry lab.The most visually appealing notebook in this folder!