Practical Consequences of Underspecification in ML

There was a recent paper talking about Underspecification in ML. Aside from making only weak evidence for this phenomenon, the paper doesn’t actually go into much detail about how to fix it. With that in mind, I figured a short post was in order to clarify what underspecification in ML is, what it looks like, and what you can do about it.

References

• Anonymous. (2020). LambdaNetworks: Modeling long-range Interactions without attention. ICLR 2021 Conference Blind Submission.

Cited as:

@article{mcateer2020underspecml,
    title = "Practical Consequences of Underspecification in ML",
    author = "McAteer, Matthew",
    journal = "matthewmcateer.me",
    year = "2020",
    url = "https://matthewmcateer.me/blog/ml-underspecification/"
}

If you notice mistakes and errors in this post, don’t hesitate to contact me at [contact at matthewmcateer dot me] and I will be very happy to correct them right away! Alternatily, you can follow me on Twitter and reach out to me there.

See you in the next post 😄