Applied Scientist at Amazon I write about trust, fraud and risk: https://nitter.gq/t.co/pmmEtFB87b
Seattle, WA
Joined January 2012
- Tweets270
- Following308
- Followers627
- Likes1.3K
Prince Grover retweeted
Some final context: fighting spam is incredibly *dynamic*. The adversaries, their goals, and tactics evolve constantly – often in response to our work! You can’t build a set of rules to detect spam today, and hope they will still work tomorrow. They will not.
Prince Grover retweeted
Shocker of a day — admitted 5 patients to liver ICU with alcohol induced liver failure. Disturbingly, ALL had binged on alcohol in the previous 4 weeks.
Sadly, only two of them can be saved by a transplant.
Weekend bingeing may be worse than regular drinking.
Serious risks are: sudden liver failure, seizures, partner harm, criminal behavior, homicide, suicide.
If you suddenly see live stream of Elon Musk giving bitcoin price prediction on your YouTube recommendations, you might be seeing a new scam happening these days. The fraudsters are hijacking YouTube accounts of people with large following.
youtube.com/watch?v=EFsRw_I3…
"Fullz", a slang commonly used by fraud criminals. It means a full set of information on a prospective fraud victim, including name, billing address, cc num, ssn, birth date. Fullz is bought and sold online, typically dark web. Impact of #DataBreach.
Starting a monthly newsletter on "AI for Trust, Safety, and Fraud." Stay informed with the latest trends in this field.
Newsletter: lnkd.in/eMdc-p5i
Signup form to receive the newsletter: eepurl.com/ioJT0E
Prince Grover retweeted
We’re excited to announce that Kaggle will be sponsoring $500k in awards for outstanding papers and reviewers in the Machine Learning Reproducibility Challenge 2022. kaggle.com/reproducibility-c…
Is someone else too interested in knowing common topics in papers/posters accepted in #NeurIPS2022 ? Pre-trained, fine-tuning, language models, self supervised, gradient descent seem to be common topics in Poster Session 3 after removing obvious stop words.
Prince Grover retweeted
❤️ Feeling blessed to have had an amazing day volunteering as a group leader for education outreach at #NeurIPS2022.
250 high school students from #NewOrleans, all asking great questions about AI ethics, copyright, and applications.
Thank you for organizing, Matt and Jess! 🙏✨ – at New Orleans, LA
I am in New Orleans this week for #NeurIPS2022. DM me if you want to chat about fraud detection using machine learning or to explore food options nearby.
The following is a useful resource to migrate following in Mastadon app: fedifinder.glitch.me/
We have created a Fraud Dataset Benchmark which is a compilation of publicly available datasets catered to fraud detection, comprising a variety of fraud related tasks. The Python based library provides API for easy data loading with standardized splits.
arxiv.org/abs/2208.14417
Prince Grover retweeted
A key challenge for ML in fraud detection is developing a strategy for handling PII data. The standard tricks don't work, and the Kaggle-like challenges ignore this critical question for privacy reasons. Part 2 of my blog dives deep into this topic bit.ly/3wL8AYM.
Back to basics! Decision tree classifier and roc curve from scratch. Could be useful to someone new to ML.
kaggle.com/code/grroverpr/de…
And a vanilla random forest made up with vanilla decision tree, together with visualization of splits and comparison with sklearn's implementation:
kaggle.com/code/grroverpr/ra…
Viz idea is stolen from here: structuringtheunstructured.b….
Prince Grover retweeted
ML and NLP Research Highlights of 2021
These are the research areas and papers I found most exciting & inspiring in 2021.
ruder.io/ml-highlights-2021/