"Dynamics-based Feature Augmentation of Graph Neural Networks" derives dynamics of spatio-temproal spread and combines it with GNNs to predict when to expect the arrival of new variants based on observations around the world

"Aligning Time-series by Local Trends" develops a time-series alignment method based on local trends and uses it for generating consensus and clustering of multiple time-series.

Two papers accepted at #AAAI2025.

RT @CAIS_USC: SMH’s annual meeting brought together researchers and public health officials to reflect on the past year and explore the fut….

RT @sriniv_venkat: "Just when I thought I was out. they pull me back in" . @ajitesh47 channeling his inner Corleone to kick off SMH Meeti….

RT @EpidamikW: The epiDAMIK workshop at #KDD2024 is happening next Monday (08/26)! We have keynotes by Mauricio Santillana @MauSantillana a….

RT @EpidamikW: We're excited to announce that the Epi meets Data workshop will be held in conjunction with @kdd_news 2024 in Barcelona! We….

RT @CAIS_USC: Congrats to CAIS Associate Director @ajitesh47 for receiving the Honorable Mention award for Best Paper for the AI for Social….

Giving a talk on better survey techniques to estimate population level trends.

RT @CAIS_USC: Join @ajitesh47 for a hybrid seminar on leveraging indirect surveys to estimate hidden populations over time. Learn how the p….

RT @BDilkina: The second two talks in this session

And if we have the responses over time, we prove that we can apply smoothing but not too much!.

We show that while conducting surveys to estimate population level numbers (how many people are sick, how many have taken a vaccine, …), it is better to ask “how many do you know who are sick” rather than “are you sick”.

RT @EpidamikW: Panel of opportunities for data-driven methods on the post-emergency era of COVID-19 w/ @2plus2make5 @svscarpino @ajitesh47….

RT @EpidamikW: 📢Deadline Extension .Thank you for your epiDAMIK submissions! In response to your valuable feedback and to provide more flex….

Our paper on detecting Autism from fMRI data using Graph Neural Networks has been accepted at #ICASSP2023. Key innovations are combining voxel-level graph structural features with multiple region-level attributed graphs, and a good acronym "IMAGIN". Paper:

Initially, designed for the ECE building only, but later opened to all. Game 2 starting next week. Detailed analysis at my page: (4/4).

The move with the highest frequency of suggestion is played. If at least 3 moves have been suggested, all unique, I can pick one (I could pick the weakest move). This helps me have winning chances, otherwise, there are some strong players at USC (3/4).

I played white. Others could suggest moves for black on a sheet outside my office or suggest via email/message (2/4).

Crowdsourced #chess Game 1 of Srivastava (myself) vs @USCMingHsiehEE (+ others) ended in my victory. After losing decisive material, @USCMingHsiehEE voted to resign. Green arrows show unique suggested moves (1/1)