Software and tutorials | Mark Allen Thornton

Tutorials

• Introduction to programming artificial neural networks (PyTorch; from MIND 2023)
• Introduction to autoencoders for psychologists (PyTorch; from MIND 2023)
• Introduction to representational similarity analysis (Python; from S4SN 2021)
• Facial expression annotation (py-feat; undergraduate class activity)
• Sentiment analysis (NLTK; undergraduate class activity)
• Tone of voice analysis (librosa; undergraduate class activity)
• 3D body pose annotation tutorial (psypose; VIBE; from NIMH workshop)
• Cross-validation workshop (R/RStudio; from Harvard Methods Dinner series)
• Dimensionality reduction tutorial (R/RStudio; from MIND 2019)

MySocialBrain.org

A web research platform for cutting-edge investigations of social cognition. MySocialBrain.org features a wide variety of fun, interactive studies with personalized visualizations and feedback. Its account system allows participants to review past results, and facilitates research into individual differences in social abilities. Studies focus on person perception, mental state representation, and social traits. Experiments integrate modern tools into research, including online cognitive modelling, adaptive design optimization, variable length experiments, and API-mediated web services.

affectr

affectr is an R package for quantitative sentiment analysis of text. The built-in dictionaries permit accurate scoring of text along three affective dimensions: rationality vs. emotionality, social impact, and valence (positive vs. negative). These dimensions have been found to organize the neural representation of others' mental states in analyses of functional magnetic resonance imaging (fMRI) data (see here). The motivation, construction, and validation of these dictionaries is described here, with code available here. The raw weighted dictionaries are freely available in a portable (csv) format here.

Matlab TFCE

A pure MATLAB implementation of multiple testing correction for functional magnetic resonance imaging via maximal statistic permutation testing with threshold free cluster enhancement (TFCE). Works with NIfTI and not dependent on use of a particular analysis suite (SPM, FSL, etc.). More restricted functionality than FSL's randomise, but validated against it for the TFCE transformation and t-tests.

Swarm simulator

An interactive simulation of collective animal behavior. Swarmers approximate real flocking/schooling behavior by attempting to maintain fixed distances from, and match velocity to, their neighbors. Implemented with d3.js.

Network Centrality Demo

An interactive introduction to the concepts of network centrality. Build up a network from scratch by pulling and dragging from an initial node. See how centrality and other statistics change depending on which connections you add or remove. Implemented with a combination of d3.js and JSNetworkX.

Book Recommender for Project Gutenberg

A book-to-book(s) text-analysis based book recommender for the Project Gutenberg corpus. Find stylistic and content matches for a particular book. See blog post here for development and implementation details.