{ Nadja Rhodes }

My work year in review

Thu, 30 Jan 2020 00:00:00 +0000

I celebrated my 1-year work anniversary at ASAPP this past week. So now that I possess one year of professional experience as a machine learning engineer (aka ML engineer or MLE) – what has this past year been like for me?

Advice on OpenAI Scholars

Sun, 07 Oct 2018 00:00:00 +0000

I am taking some blogging advice that I’ve read elsewhere: if multiple people ask you similar questions offline, answer those questions in a blog post!

deephypebot: an overview

Fri, 31 Aug 2018 00:00:00 +0000

Motivation

Notes on topic modeling

Fri, 24 Aug 2018 00:00:00 +0000

I’ve talked a bit about topic modeling on this blog before, pre-Scholars program. I revisit the subject now as a potential automatic reward function for the LC-GAN I am using in my final project. My hypothesis is that topic modeling can distill the sentences in my music commentary data set down into distinct sentence types; this information can then be used to teach the LC-GAN what types of sentences to encourage the downstream language model to generate.

Notes on genre and inspiration

Tue, 14 Aug 2018 00:00:00 +0000

My final project is underway, and it feels like I get to return to my software engineering roots with these powerful, new deep learning techniques in my tool belt. This is a welcome opportunity to flex what I’ve learned this summer.¹

Perhaps this is the true meaning of “Flexing To The Study” :slightly_smiling_face: ↩

OpenAI Scholar: Final Project

Fri, 03 Aug 2018 00:00:00 +0000

This post is a replica of my OpenAI Scholar final project proposal, also available here.

Training an LC-GAN

Sat, 28 Jul 2018 00:00:00 +0000

I spent the last week before my 4-week final project understanding, implementing, and training a special kind of GAN. GAN is short for generative adversarial network, a neural network that “simultaneously trains two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G” (from the original GAN paper).

Interpreting Latent Space and Bias

Sat, 21 Jul 2018 00:00:00 +0000

For week 7, and my second week on model interpretability (see first week post), I focused in on one particularly cool VAE-based visualization example from Ha & Schmidhuber’s World Models work. I also did some broader thinking around selection bias in my song review training data.

Interpreting with Attention and More

Fri, 13 Jul 2018 00:00:00 +0000

This week, I picked up from last week’s study of attention and applied it towards model interpretability, or the ability for humans to understand a model. I also explored some other “attribution-based” methods of interpretability that credit different parts of the model input with the prediction.

Not Enough Attention

Fri, 06 Jul 2018 00:00:00 +0000

Attention week: what an ironic week to get distracted.