PhD, CS @VT_CS, AWS AI -- Ex-Yahoo. Likes machine & human learning. -- Opinions : Strictly Personal Retweets != Agreement/Endorsement
Seattle, WA
Joined September 2014
- Tweets84
- Following349
- Followers123
- Likes960
While there are many resources/blogs floating around, not all are as comprehensive as this one on LoRA, discusses details one actually needs to know in practice. Worth a read. thinkingmachines.ai/blog/lor…
42
Illegal Timber Trade is a multibillion $ transnational crime. This paper describes a pilot project , w/ @simeone_john , @profnaren & collaborators from @WWF at #VT, gives a concise peak into the challenges & how ML based methods can help address it. journals.plos.org/plosone/ar…
Debanjan Datta retweeted
At the #KDD2022 conference this week, @SanghaniCtrVT director Naren Ramakrishnan (@profnaren) reunited with former Ph.D. students @VT_CS now on their own career paths in academia or the tech industry: Feng Chen, @bijayaAdh, Fang Jin, @devDdata.
Debanjan Datta retweeted
Faculty and Ph.D. students @VT_CS @SanghaniCtrVT were among the organizers and presenters at #KDD2022 this week in Washington, DC. The conference was also a good opportunity to meet up with alums and colleagues from other universities.
I'm excited to share my paper 'Framing Algorithmic Recourse for Anomaly Detection' with @profnaren @niustrategy will be presented at ACM SigKDD 2022, Research Track. If you're at #KDD2022 check out our talk & poster on 8/16!
[arxiv.org/abs/2206.14384] @kdd_news
Debanjan Datta retweeted
Framing Algorithmic Recourse for Anomaly Detection
deepai.org/publication/frami…
by @devDdata, @niustrategy, and @profnaren
#AnomalyDetection #MachineLearning
Thanks @nikhilm_1 . It's been great to have you as a colleague, and friend. Do keep in touch!
Thank you @profnaren for all your guidance, motivation and help! It was an eventful journey in graduate school that definitely enabled me to learn & explore, as well as meet & work with so many great people!
1
Debanjan Datta retweeted
Anomaly detection methods compute an anomaly score A(x), and in research, we measure their effectiveness using AUC for the binary decision "anomaly" vs. "not anomaly". But in applications, we need to choose a threshold tau. How can we set tau without having labeled anomalies? 1/
Debanjan Datta retweeted
ODD Workshop at #KDD2021 starts from 8am Pacific time. There are 6 exciting keynote talks and a panel discussion on fairness in outlier detection.
oddworkshop.github.io/
Debanjan Datta retweeted
With the Fall 2021 semester fast approaching, students @SanghaniCtrVT are finishing up their summer internships. Ph.D. student Debanjan Datta (@devDdata) has been working on time series characterization and classification @awscloud, Seattle, WA. @profnaren @VT_CS
Debanjan Datta retweeted
Useful thread w/ refs to recent happenings. cc @devDdata , @ShengzheXu.
