Lab 1: Hello Julia!

Lab Teaching Objective:

Train you with the necessary tools and skills you need to complete your modeling project

Student Objectives:

• Identify an engineering or scientific problem in Artificial Intelligence, Psychology, or Cognitive Neuroscience.
• Identify & assess the quality of the data required to empirically investigate the problem.
• Develop a 1 page project proposal abstract.
• Use the machine learning skills acquired through the lab and theory to investigate the problem.
  • Recode the data using unsupervised learning (if necessary).
  • Implement a simple logistic regression model, shallow neural network, and a deep neural network model.
  • Evaluate the performance of your model (Confusion Matrix).
  • Do 2-fold cross-validation.
  • Compare the results of competing models.
• Present the results of your study as a 15-minute talk.
• Write and submit a ~10 page project report in APA Journal Article format.

SO 1: Identifying the problem

• You fill out a project interest survey
  • Artificial Intelligence Project Examples:
    • Email Spam Filter, Speech Recognition System, Twitter Hate Speech Detection
  • Psychology Project Examples:
    • Human Emotion Classification using Eye-Feature, Psychopathy Diagnosis
  • Cognitive Neuroscience Project Examples:
    • Alzheimer’s detection using fMRI, Analysis and Modeling of Neural Ensemble Rehearsal During Sleep, Tumor detection using MRI, Autism diagnosis using fMRI
• You form groups by yourself or we help you find project groups

SO 2: Identify & assess the quality of the data

Some places to look for Data

• Kaggle: For all sorts of datasets.
• UCI Machine Learning Repository: For all sorts of datasets.
• UCLA Library: For Psychological datasets.
• CMU Datashop: For Educational & Psychological datasets.
• OpenNeuro: For Neuroscience datasets.
• CRCNS: For Neuroscience datasets.
• RDatasets: A collection of common datasets used in all Statistical/Machine Learning textbooks.
• Google Dataset Search

SO 3: Project Proposal Abstract

SO 4: Machine Learning Skills/Tools

 

Common ML Languages

Why Julia ?

• Open Source (Free to Use, No license required to do commercial project in Julia)
• Less verbose (compared to R & Python)
• Best-in-class interoperability (with R, Python & C)
• Really fast when implementing computationally complex algorithms
• Good software engineering principles instantiated in the language design
• Used widely in High Performance Scientific Computing

Shortcomings

• Relatively young community –> fewer packages compared to R or Python (RCall.jl & PythonCall.jl to the rescue)
• Package installation can take some time (continually improving with each update)

Topics to cover

• Installing Julia
• Change language type in a new file
• Know your IDE
• Julia REPL Modes
• Installing Packages
• Creating a Julia Script
• Project Environments