1 minute read

Updated:

Throughout my academic career I've been interested in sensing, in biological inspiration, and in understanding biology in the lens of its function. In 2016, I switched paths from bioinspired nanoscale materials engineering to computational neuroscience. Learning about deep learning was instrumental in this switch.

Seeing beyond the CV

In a book of work by Studio Gang, a brilliant architectural firm led by Jeanne Gang, there is a sketch of the studio’s themes and interests over time, mapped out in pencil just like the figure above. I thought it was brilliant, both an exercise and as a communicative device.

At the same time I was also recreating my CV, as required by my department, and I was feeling disappointed by the linearity and flatness of the format. And so as an exercise in communication and introspection I decided to make my own. But not in pencil – in code.

Making colors represent topic similarity

I knew immediately I wanted color to represent the similarity between my academic interests. Without that cue, the 30+ interests of mine would fade into each other.

Instead of choosing colors by hand, I felt it would be easier to map the similarity of my interests directly into color algorithmically. This involves three steps:

1) Make a similarity matrix for each interest. 2) Use this matrix to create a 3D embedding of my interests that preserves distances as best as possible (using MDS, multidimensional scaling). 3) Interpret the 3D embedding as locations within a 3D color space in which Euclidean distance corresponds to perceptual similarity. I chose the CIELAB color space.

2D embedding of similarity matrix
You can get a sense of the process here. This is a 2D embedding of the similarity matrix – distances correspond to topic similarity. I've colored each point by its location in a 3D color space.