Hikari Sorensen
Hikari Sorensen
Summer School Info
Summer School Info
Mentor: Giulio
Track: Science
Track: Science
Email: laurensorensen@college.harvard.edu
Phone: 647-244-5186
Phone: 647-244-5186
Bio
Bio
Hikari Sorensen has just finished her freshman year at Harvard College, where she studies mathematics and computer science. She really likes thinking about structure underlying real phenomena, and especially loves abstracting things, considering things in their abstract form, which things include sentences in bios in which she talks about stuff she likes. She likes meta-stuff, and also meta-meta-stuff, and also meta-meta-meta-stuff...
When her head's not way up in the clouds (which, frankly, is not very often), she likes to do machine learning, or sing songs, or catch Pokemon, or (with nonzero probability) do all three at the same time. She also likes writing and talking philosophy, she supposes, although with regard to the latter, beware if you want to engage her, because she's been known to get pretty heated in debate about these kinds of things. She's most recently been enjoying how traditionally first-person expressions in writing sound when done in third-person.
When her head's not way up in the clouds (which, frankly, is not very often), she likes to do machine learning, or sing songs, or catch Pokemon, or (with nonzero probability) do all three at the same time. She also likes writing and talking philosophy, she supposes, although with regard to the latter, beware if you want to engage her, because she's been known to get pretty heated in debate about these kinds of things. She's most recently been enjoying how traditionally first-person expressions in writing sound when done in third-person.
Topic Exploration Homework
Topic Exploration Homework
Summer School Project
Summer School Project
Visualization of Hidden Layer Transformations in Neural Networks
Visualization of Hidden Layer Transformations in Neural Networks
Goal: A dynamic visualization of features and weights learned in a neural network over the course of training.
Neural networks are commonly used in machine learning applications, but their internal workings — how exactly they learn the models they do — are still somewhat mysterious. Indeed, neural networks are largely considered a “black box” for predictive modeling. This project seeks to visualize, in a humanly interpretable way, the transformations to data a neural network makes in its training process.
The following information is immediately accessible, and will be interesting to track over iterations:
- Gradients
- Weights
- Convolved output images
What is not immediately available is information about exactly which features are learned by the network over many layers, and, in particular, if there is any meaningful way to interpret these features. Understanding the interactions between learned parameters and features is the ultimate goal of this project.
The following information is immediately accessible, and will be interesting to track over iterations:
- Gradients
- Weights
- Convolved output images
What is not immediately available is information about exactly which features are learned by the network over many layers, and, in particular, if there is any meaningful way to interpret these features. Understanding the interactions between learned parameters and features is the ultimate goal of this project.
Other Things Created at the Summer School
Other Things Created at the Summer School
Topic Explorations
Topic Explorations
Community Posts
Community Posts
Data Repository Entries
Data Repository Entries
Functions Repository Entries
Functions Repository Entries
Neural Net Repository Entries
Neural Net Repository Entries
Challenges
Challenges
Demonstrations
Demonstrations
NKS Book Notebooks
NKS Book Notebooks
Github Material
Github Material
Wolfram Community
Wolfram Community