Jordan Meyer
@TechingForward
Machine learning engineer solving hard problems
Generating randomness is expensive. TIL cryptographers used radioactive emissions to generate random keys.
Pay closer attention to behaviors that deviate from assumed incentives from the environment
One of the best ML bloggers out there opens the hood on Diffusion models.
Diffusion models are another type of generative models, besides GAN, VAE, and flow models. The idea is quite smart and clean. It is flexible enough to model any complex distribution while remains tractable to evaluate the distribution. lilianweng.github.io/lil-log/2021/0…
Disturbed by the blatant obstacles to authorization and payment today. I'd encourage fellow engineers in healthcare to consider how their work shapes the field. Algorithms intended to predict patient conditions may be the basis for black box claim denials.
MISADVENTURES IN AMERICAN HEALTHCARE 1. Doctor orders scan that's clearly indicated for patient & which they've had many times before (allowing for apples-to-apples comparison) 2. Despite order weeks in advance, receive denial day-before from insurer without any explanation
"Fighting clutter is like fighting weed-- the writer is always slightly behind." -William Zinsser. Fresh complexities spring up always.
Advice from a former mentor: try to eliminate "I have to..." from your speech to train yourself to recognize your agency.
Loguru removed the hassle around logging in python. Multiprocessing file logging is no longer a headache. 🙏
Another good day spent progressively overloading my mental schemas
Working with the garage door open means writing audience-friemdly readmes
Love that line that writing gives poor thinking no where to hide. This all requires a good deal of courage to take inventory of the quality of your ideas. It's unpleasant to realize an idea is flimsy or rubbish in its current state.
The best way to improve your ability to think is to spend large chunks of time thinking. One way to force yourself to slow down and think is to write. Good writing requires good thinking. Writing gives poor thinking nowhere to hide. A lack of understanding becomes visible.
Even a short walk often ushers fresh perspective on some code I'm working on
In my experience it's tricky to avoid a vanilla AE from compressing all samples to a narrow latent space. Variational AEs offer a nice way to avoid this sort of collapse.
One of my favorite applications of neural networks: autoencoders. Here is how they work and how you can use them: 1 of 12
Every idea seems great when it only exists in our heads.
Writing is the best way to get distance from your mind. We're emotionally attached to the ideas in our head, but if we write them down and step away for a bit, we can think more clearly.
Our focus has tradeoffs. Reminds me of the study of the observers who entirely missed a gorilla walking through a snippet of video while they were counting some action taken in the video. What am I missing in my research because of where I place my focus?
In 2009, a Stanford business professor: - Split her class into 14 teams - Gave each team $5 and 2 hours - Told them to get as much ROI as possible - Said they’d give a presentation after finishing Here’s how it went down + the foundational lesson you can learn from it: 🧵
Attention is all you need sometimes
"An Impartial Take to the CNN vs Transformer Robustness Contest" Are vision transformers really better than CNNs? This paper strongly suggests an answer, based on a robustness throwdown between {ViT, Swin} vs {BiT, ConvNeXt}. [1/10]
![davisblalock's tweet image. "An Impartial Take to the CNN vs Transformer Robustness Contest"
Are vision transformers really better than CNNs? This paper strongly suggests an answer, based on a robustness throwdown between {ViT, Swin} vs {BiT, ConvNeXt}. [1/10]](https://pbs.twimg.com/media/FZd_6YYVQAAg6UR.jpg)
Action reveals information about our environment and ourselves.
Falling short of a goal doesn't mean failure. Often it's progress. 14 experiments, 10k people: people counted results that didn't meet targets as losses—even if they were gains. The key measure of success isn't reaching your destination. It's improving from your starting point.
Yes. Encourage your software teams to embrace a culture of writing. Clearly define your audience and compel them to act.
Jeff Bezos said: “There is no way to write a six-page narratively structured memo and not have clear thinking.” Here’s the writing framework Bezos uses (that you can too):
Rust for sure-- how large of an explosion depends on how effectively python 3.11 and future iterations improve performance.
Which programming language has the most chance of exploding in the next 5 years?
Unexpected windows updates exist to dash this feeling.
The Machine Learning for Healthcare call for papers is out! Deadline: April 14th. We will be *in person* (at @DukeU) this year. Get your best ML papers & clinical abstracts in shape & submit to MLHC, the premier venue in this exciting intersection. mlforhc.org/paper-submissi…
United States الاتجاهات
- 1. Sonic 3 8,346 posts
- 2. GOTY 19.8K posts
- 3. Expedition 33 30.6K posts
- 4. Moana 28.3K posts
- 5. #TheGameAwards 35K posts
- 6. Zelda 55.4K posts
- 7. Comey 28.6K posts
- 8. Ted Cruz 15.7K posts
- 9. FEMA 6,878 posts
- 10. Cher 20.6K posts
- 11. Clair Obscur 19.7K posts
- 12. Silksong 36.1K posts
- 13. #RestoreTheSnyderVerse 38.4K posts
- 14. Until Dawn 2,867 posts
- 15. Hades 2 14.5K posts
- 16. Geoff 5,699 posts
- 17. Rex Ryan N/A
- 18. Larosas N/A
- 19. Grok 4.1 6,981 posts
- 20. David Richardson 2,527 posts
Something went wrong.
Something went wrong.