Omid Sani

[Omid Sani @omidsani]

💻We have added the IPSID code to the PSID Python/Matlab repos:
https://t.co/RhYLFr5veb

For python 🐍, you can run
pip install PSID --upgrade
to install the latest version that includes IPSID.

📒We have also prepared a tutorial notebook for IPSID here:
https://t.co/8gomwBL4hW

[whenlotto @whenlotto]

#Powerball
⏲️ Next draw: Dec. 24, 2025, 7:59 p.m. PT (11 hours from now)
💰 Jackpot: $1.7 billion
💵 Cash value: $781.3 million
🎫 Ticket: $2
💵 Expected prize cash value
Before tax: $2.82 ⇨ $2.14 (ROI: 7%)
✅ Could be worth it

Source: https://t.co/2URdXJI2g3

[Omid Sani @omidsani]

📢 We have a new notebook that demoes DPAD (@NatureNeuro 2024) in public neural-behavioral data from start to finish, answers FAQs, & gives best practice tips! Check it out!

🔗 Link is in the repo: https://t.co/FqkKzyqsYs

🙌 Thanks @Vahidi_Parsa & @MaryamShanechi!

Paper 🧵⬇️

[Maryam Shanechi @MaryamShanechi]

Excited that our paper made the cover of @NatureNeuro!

We developed DPAD, an AI framework for dynamical modeling of neural-behavioral data and dissociating behaviorally relevant neural dynamics.

📜https://t.co/Uz71hQX2ZL
💻https://t.co/ghiUGzxs0e
👏@omidsani

[whenlotto @whenlotto]

#Powerball
⏲️ Next draw: Sept. 6, 2025, 7:59 p.m. PT (10 hours from now)
💰 Jackpot: $1.8 billion
💵 Cash value: $826.4 million
🎫 Ticket: $2
💵 Expected prize cash value
Before tax: $2.98 ⇨ $2.26 (ROI: 13%)
✅ Could be worth it

Source: https://t.co/2URdXJHuqv

[Omid Sani @omidsani]

You can subscribe to my thing to get notified every time it happens:
https://t.co/LQ4qEnTj95

And follow its twitter account:
@whenlotto

[Omid Sani @omidsani]

🚨 Oh this rare occurrence is happening again!

The expect value of the Powerball prize, per ticket, after tax, is more than the $2 ticket price.

Even very rational people who are curious about lotteries but can't justify the game being rigged against the players can also play.

[whenlotto @whenlotto]

#Powerball
⏲️ Next draw: Sept. 6, 2025, 7:59 p.m. PT (2 days, 17 hours from now)
💰 Jackpot: $1.7 billion
💵 Cash value: $770.3 million
🎫 Ticket: $2
💵 Expected prize cash value
Before tax: $2.78 ⇨ $2.11 (ROI: 6%)
✅ Could be worth it

Source: https://t.co/2URdXJHuqv

[Omid Sani @omidsani]

This has happened ONLY 5 times in the past 12 years by the way! 🤯

[Paul Graham @paulg]

Are they really this clueless about how rocket development works, or do they hate SpaceX because of Elon? As so often happens, it's hard to distinguish incompetence from politics.

[Omid Sani @omidsani]

56% net margin is crazy! And we're still far from feeding the market's appetite for server side compute.

VCs are subsidizing the investment now:
VC ➡️ Startup ➡️ Nvidia

In hopes that they will get to:
Customer ➡️ Startup ➡️ Nvidia

And a key winner in both cases is Nvidia!

[Earnings Anatomy @EarningsAnatomy]

NVIDIA $NVDA Q2 FY2026 results are in! 🚀
✨ Record Revenue: $46.7B (+56% YoY)
💸 Staggering Net Income: $26.4B (+59% YoY)

The AI engine roars, with Data Center revenue at $41.3B! 🧠 Reinvesting big in R&D ($4.3B, +39%). But watch out for export control risks 🌍.

https://t.co/P9YOxdwnnB

[Earnings Anatomy @EarningsAnatomy]

$MSFT and $GOOG have had similar net incomes for the past decade, but somehow $MSFT is worth $4T but $GOOG is worth only $2.35T. Why? 🤯

[Omid Sani @omidsani]

Updated ScholarEnhancer to Manifest 3, so it works in Chrome again!

https://t.co/tYzvyWafo0
https://t.co/IpfObKSYRF

[Omid Sani @omidsani]

Have you ever had to endlessly scroll & click to load all the papers on a Google Scholar profile? I once made an open source browser extension that fixes this and can also
- export a bibtex of all papers
- export a list of all co-authors
You may like it!🙂
https://t.co/sOq47koOeI

[Omid Sani @omidsani]

Our latest paper uses a creative dual RNN architecture to dissociate intrinsic vs input dynamics: a predictor RNN extracts latent states while a generator RNN explains how intrinsic dynamics evolve from those.
@Vahidi_Parsa is presenting at #ICLR2025 👏
Check 🧵 for details 👇

[Maryam Shanechi @MaryamShanechi]

BRAID:

✅ Disentangles intrinsic behaviorally relevant neural dynamics from input, neural-specific & behavior-specific dynamics
✅ Captures nonlinearity

It is a multi-stage RNN: each stage consists of a predictor network & a generative network to learn the intrinsic dynamics.

[Maryam Shanechi @MaryamShanechi]

Excited for BRAID to expand our neural-behavioral models: PSID & DPAD (Nat Neuro 2021 & 2024), IPSID (PNAS 2024)!

See @Vahidi_Parsa at #ICLR2025

📍Poster Session 5, Hall 3+Hall 2B #57 | Sat Apr 26 | 10AM - 12:30PM @iclr_conf

📜https://t.co/dnOMZzVdZf
💻https://t.co/qG0C4MgAUR

[Maryam Shanechi @MaryamShanechi]

🎉 New in #ICLR2025, we present BRAID for input-driven nonlinear dynamical modeling of neural-behavioral data.

BRAID disentangles the intrinsic dynamics shared between modalities from input dynamics and modality-specific dynamics.

👏@Vahidi_Parsa @omidsani
🧵, Paper & Code ⬇️

[Maarten Ottenhoff @MC_Ottenhoff]

Check out our new preprint where we decode goal-directed hand movements from brain-wide intracranial recordings. Find the full preprint here https://t.co/Qxf1zfaX6e. I did move to the where the sky is blue, so please find a brief summary over there!

[whenlotto @whenlotto]

#MegaMillions
⏲️ Next draw: Dec. 27, 2024, 8 p.m. PT (2 days, 10 hours from now)
💰 Jackpot: $1.15 billion
💵 Cash value: $516.1 million
🎫 Ticket: $2
💵 Expected prize cash value
Before tax: $1.82 ⇨ $1.38 (ROI: -31%)
❌ Still NOT worth it
Source: https://t.co/2URdXJI2g3

[Omid Sani @omidsani]

A lottery has a Billion dollar jackpot again, but surprise... It is still not worth it! 😀

P.S. My account is not hacked. I made @whenlotto; That's why I some times tweet about it. 😅

[whenlotto @whenlotto]

#MegaMillions
⏲️ Next draw: Dec. 24, 2024, 8 p.m. PT (1 day, 10 hours from now)
💰 Jackpot: $970.0 million
💵 Cash value: $439.9 million
🎫 Ticket: $2
💵 Expected prize cash value
Before tax: $1.57 ⇨ $1.19 (ROI: -41%)
❌ Still NOT worth it
Source: https://t.co/2URdXJI2g3

[Maryam Shanechi @MaryamShanechi]

Can we efficiently learn models of collective dynamics in multimodal discrete-continuous time-series, e.g. spike-LFP?

We develop a new multimodal subspace identification method to do so & enable causal multimodal decoding!

👏@ParimaAhmadi @omidsani
J Neural Eng paper, code,🧵⬇️

[Omid Sani @omidsani]

Congrats @loganesian! 🎉

Check out our new multimodal preferential learning method, developed & presented by @loganesian @NeurIPSConf!

Like PSID, this method is analytical, so its learning is not iterative or gradient-based.

👇 Code and #NeurIPS2024 poster info in the 🧵

[Maryam Shanechi @MaryamShanechi]

New in @NeurIPSConf, we develop PGLDM, an analytical subspace identification method to jointly model two generalized-linear time-series & dissociate their shared vs private dynamics.

#NeurIPS2024 Poster Sess 5 East #3808, Fri Dec 13 11am
👏@loganesian @omidsani
Paper, Code &🧵⬇️

[Maryam Shanechi @MaryamShanechi]

Excited that our paper made the cover of @NatureNeuro!

We developed DPAD, an AI framework for dynamical modeling of neural-behavioral data and dissociating behaviorally relevant neural dynamics.

📜https://t.co/Uz71hQX2ZL
💻https://t.co/ghiUGzxs0e
👏@omidsani

[Omid Sani @omidsani]

Google's NotebookLM (https://t.co/7GQon8NAeC) makes surprisingly convincing podcasts! This one was made by just pasting the link to https://t.co/MubFB2AQ5H into NotebookLM! 🤯

[whenlotto @whenlotto]

Anyone can make a podcast these days!! 🤯
We turned our blog post into a podcast just using Google's NotebookLM (https://t.co/CtMsjgLjSQ) and https://t.co/e06Uf2jckJ! Check it out!
https://t.co/8CxGQxf8Lo