Austin Benson

Quant finance. Former computer science professor.

[Kiran Tomlinson @kiran_tomlinson]

If you’re at NetSci and want to learn about how we can use social networks to improve choice prediction, come check out my poster! (Research with @austinbenson)

[SIAM @TheSIAMNews]

Many models involve numerous parameters that make direct interpretation challenging. At #SIAMMDS22, @dgleich of @Purdue explores the possibilities of using ideas based on #topological analysis methods to understand and evaluate these complex prediction functions.

[Chris Riederererer @yeahtwoERs]

thank you @austinbenson , i was going to declare it a huge win if this niche post got more than 0 likes

[Giovanni Petri @lordgrilo]

Latest work out here: https://t.co/Kyiq4KKQ7j led by @arnaudon, with @RobertPeach15 and @ExpertPol. We take a careful look at Kuramoto models on simplicial complexes including frustration, harmonic order parameters and orientations. A thread 1/N

[Johan Ugander @jugander]

New work, out today in @PNASNews, joint with @jonassjuul! Please see his excellent summary 🧵!

[Jonas L. Juul @jonassjuul]

Does false news spread differently than true news online? For example, does false news spread faster or deeper into the Twitter network? How about videos vs. petitions? New paper out in @PNASnews with @jugander https://t.co/9Xl9ERjazY 1/14

[Kiran Tomlinson @kiran_tomlinson]

📝We are now accepting submissions for the NeurIPS '21 Workshop on Human and Machine Decisions! Submit new and developing research in ML x decision-making; both empirical and theoretical contributions welcome. A partial list of topics and submission link: https://t.co/JTNYrCMjgf

[Quanta Magazine @QuantaMagazine]

The traditional networks of graph theory aren’t enough to fully map today’s vast pools of data. Researchers are pushing to provide new kinds of network models that can accommodate the embarrassment of riches. @StephenOrnes reports: https://t.co/Nx5DQISbsu

[Kiran Tomlinson @kiran_tomlinson]

If you're attending KDD and want to learn about some recent discrete choice research, I'm giving a couple of talks tomorrow at 12:30pm CDT during this paper session: https://t.co/YeOn5I84Jz

[David Gleich @dgleich]

Need data (networks, hypergraphs, tensors, ML, AI)? Here’s a timestamped and linked digest of Anthony Fauci’s emails https://t.co/4Ude2KaDv4 from @JasonLeopold FOIA request. With @austinbenson and @n_veldt Data and Code https://t.co/0NBMmmOwzD

[Kiran Tomlinson @kiran_tomlinson]

Big news! 🎉The Workshop on Human and Machine Decisions will be at NeurIPS '21! 🎉We have some fantastic speakers/panelists, incl. @duncanjwatts, @2plus2make5, @jugander, @maithra_raghu, @hima_lakkaraju, @CFCamerer, ... (see site) *Call for papers* 🧵1/3 https://t.co/fvMgsXyYu8

[Austin Benson @austinbenson]

@bayrameda_ee @emaros96 Ah, I see. In the ZGL algorithm, the iterative solution is independent of initialization. In the Zhou et al. LP, the initialization matters. (In both cases, though, the indicator vector has a different bias, which is why ZGL discuss things like class mass normalization)

[Austin Benson @austinbenson]

@bayrameda_ee @emaros96 I agree that indicator vectors for categorical labels is different. But there are similar issues with implicit data assumptions. Sec 4 of ZGL paper uses class mass normalization (https://t.co/omN0q7L6MT) and Sec 3.3 of https://t.co/mpVOg2esVN has a nice example w/ label imbalance

[Austin Benson @austinbenson]

@bayrameda_ee @emaros96 If the target mean is not zero, then initializing with zero on unlabeled nodes and using algorithms like average-of-neighbors-at-unlabeled-nodes algorithm is not appropriate. For example, if mean is super large, then initializing at zero will bias towards too small predictions.

[Austin Benson @austinbenson]

@emaros96 Yeah!

[Blair D. Sullivan @BlairDSullivan]

So who's going to be at #SIAMACDA21 next week? Hoping to see some familiar faces!
@austinbenson @RichVuduc @dgleich @eapurvine @Prof_DavidBader @aydoz @DocSparse @rpeng233 ?
PS. Check your email for instructions on participating in the Introduction Blitz on Monday!

[Austin Benson @austinbenson]

@algo_class Localized / seeded clustering in graphs (and related flow-based improvement algorithms). Good survey: https://t.co/VCUaeJZVgd

[Austin Benson @austinbenson]

Looking forward to this tomorrow! First time talking about this paper with @n_veldt and Jon Kleinberg:
https://t.co/GeLbxyhuWR

[Leonie Neuhäuser @[email protected] @leoneuhaeuser]

Keynotes by @ingo_S and @austinbenson, a panel discussion with @mstrohm, @KristinaLerman and @fede7j, a software demo session by @BenFMaier and invited talks by @_giuliacencetti, @iacopoiacopini, @antoneri, @GuiFdeArruda, @HZinnbrooks, @PhilChodrow and @lee_de_ville

[DynaMo: Dynamics and motifs @networks_dynamo]

On June 28, Austin Benson (Cornell University) will kick off the Dynamo satellite with a talk entitled “Triadic data analysis in temporal and higher-order networks”. We are looking forward to hearing @austinbenson’s take on temporal networks, higher-order networks and motifs!