Paper submission advice?

I’ve nearly finished my first ‘technical’ machine learning application paper and am finishing plans on how and where to submit it. Wondering if anyone has any advice…

( Background: I’m a “late-career” teaching prof who got “out of” physics research a while ago but recently developed a love for ML research, at a small college, where they will cover costs for presenting at one conference per year. I alternate years between the fall conference for each of two fields I straddle and this leaves no budget to travel to anything else, e.g. no ML conferences like ICML or N(eur)IPS. And my uni regards conference presentations as “less than” compared to journal publications. Even so, grants I’ve looked into either stipulate that they don’t cover travel, or if they do cover conference travel then generally their eligibility requirements…I fail to meet for reasons of race, gender, or age/“years since PhD”. )

So the plan has been to submit to one of the journals in my discipline(s), however this will (a) take a while (e.g., a year) to get reviewed & published and (b) they don’t yet have a way to anonymously submit code, datasets, or demos as part of reviews, and I doubt my collaborator wants to hand over our code with our names stripped out on some public forum. But I would like this work to be available to members of the ML community too, preferably sooner rather than later…

Thus, to counteract (a) I can post to arXIv and hope reviewers won’t see it (or care if they do) – this is pretty common & acceptable in my field – and for (b) I can make my demo page “anonymous” and then…what? Maybe wait until notice of acceptance before making the code (and dataset(s)) public? (Currently the code is in a private GitHub repo).

Any thoughts or feedback on how to conduct this, e.g. to do things differently? If this were about my original area of expertise years ago, I’d know what to do, but in this area I’m new. Thanks!


PS- Because I’m overly excited: here’s the (work-in-progress) demo page: Note it can be slow to load/respond; it’s composed of two single-worker Heroku apps. The graphs, sliders & NN inference are handled server-side via Bokeh Server (& a stripped-down, CPU-only version of PyTorch), b/c ONNX export in PyTorch omits too many special functions I needed.

1 Like

as i understand it most researchers post on arxiv to “show who came first”/notoriety while going through the much longer review process for “normal” papers. Anyway you better verify that your self.

I hope you will be on axiv, i rarely find papers elsewhere:)

1 Like