Category Theory
Zulip Server
Archive

You're reading the public-facing archive of the Category Theory Zulip server.
To join the server you need an invite. Anybody can get an invite by contacting Matteo Capucci at name dot surname at gmail dot com.
For all things related to this archive refer to the same person.

Stream: deprecated: chemistry

Topic: random biological Petri nets

Kris Brown (Jun 06 2022 at 01:58):

To follow up on our group's project, I'm trying to think of how to operationalize our goal of "capturing the essence of biological reaction networks via constraints on some process to randomly generate Petri nets." The simplest way I can think of is that we pick some property $\phi$ (a function $\mathbf{Petri}\rightarrow\mathbb{R}$) and observe some distribution of $\phi$ over our whole dataset of biological Petri nets. We then seed our random Petri net generating process with a small number of motifs yet recover a similar overall distribution of $\phi$.

A related way to measure our success is to do data augmentation for ML. To summarize how that works:

• A simple ML model is trained to take in reaction network and make a guess at $\phi$.
• We only trained with a subset of our total data - evaluating on the hold-out data shows the black box is overfitting!
• We train a different model by augmenting our training set with random Petri nets (informed by the motifs of our training data).
• Training a ML model on this reduces overfitting, whereas training a ML model augmented with completely random networks does not do as well.

This wouldn't be the focus of the project, just a detail of our methodology for how we measure our success. This was just one idea that struck me - would love to hear other ideas for quantifying results!