By Alicia Wang, Arjun Khare, Arsh Zahed, James Jiao, Olivia Jain, Nareg Megan and Reina Wang

Introduction

The goal of this project was to be able to build a network that would be able to answer a question about a given image.

We used the MSCOCO dataset for our model. The scenes are simple, and the questions below were generated from asking many people to write a question that is required to use the picture. The images with the questions were then given to others to answer the questions. We trained our model on the image, question pairs as our input, and the answers as our output.

This task can be broken down into three parts: image embeddings, word embeddings, and recurrent attention.

Baseline Model

The first model we tried is relatively simple. We use the K-Nearest Neighbors algorithm to find the most suitable answer to a given image, question pair in the following process. Using the question embeddings we find the K-nearest questions to the test question…


Providing a Vision Of The Future Using Spatio-Temporal Machine Learning Models on Weather Images

Motivation

Current weather predictions in the industry involve measurements of data such as humidity, temperature, wind speed, etc. that are inputted into physics equations to be computed for predictions. While a reliable method, it is an extremely computational process that can take days to computer, which also makes the predictions independent of more recent weather data and suffer from a natural range of predictions.

The goal of this project is to test current ML research in video frame prediction on weather prediction. Particularly, we will be using consecutive radar images that are fed into a model to extract weather patterns…


Predicting NBA Finals and Other Match Statistics Using ML

Background

The standard set of game-level basketball statistics include but are not limited to points, rebounds, blocks, etc. Though these statistics relay valuable information about a team’s performance across a game, often more advanced statistics are computed such as a team’s offensive efficiency which is defined to be the total points scored normalized by the total number of possessions. These statistics were pulled using an NBA API that spans games from the late ’70s to the current season.

While non-linear models often tend to perform well for regression tasks, linear models offer more interpretability of features, which is a great option…

Launchpad

A technology club at UC Berkeley that fosters a community of creative and passionate engineers to tackle real world problems using machine learning.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store