Alex 徐瀚

Senior Data Scientist at Huntington National Bank

[Alex 徐瀚 @runninghsu]

It’s Dr. Hsu, not Dr. Seuss.

[scikit-learn @scikit_learn]

🚀 Exciting News! 🚀 Thrilled to announce that @NVIDIA has joined @sklearn_inria as a corporate sponsor! 🤝💡 Their commitment to advancing open-source software in #MachineLearning is stronger than ever. Together, let's push the boundaries of innovation! https://t.co/gWBLrS1vLt

[Harry Dempsey @HJ_Dempsey]

I am grateful to have been a part of this work!
I really enjoyed using @DeepLabCut and B-SOiD @runninghsu to perform behavioural classification in ~185 hours of video.
This would not have been possible without these tools.

[Alex 徐瀚 @runninghsu]

Best cloud experience I’ve had!

[Khuyen Tran @KhuyenTran16]

If you want to remove unwanted objects or people from your pictures, use Lama Cleaner.

Lama Cleaner is a free and open-source tool powered by the SOTA AI model.

Link to Lama Cleaner: https://t.co/y9KdkboW3v

[Michael Black @Michael_J_Black]

PhD students, don't worry. Technologies, trends, and even whole fields come and go. A PhD makes you an expert in a field but, more importantly, teaches you how to become an expert. Once you know that you can learn anything, you can adapt to major disruptions in your field.

[Alex 徐瀚 @runninghsu]

Great work from the Datta Lab! Super excited to see iron sharpening iron in the behavioral analyses space!

[Alex 徐瀚 @runninghsu]

@MoeNeuro Google drive can work too.
Mounting it and you can access it in your finder.
Google drive would work better than Dropbox because of functionalities like slides, doc, and sheets.

[Alex 徐瀚 @runninghsu]

2/7! For whatever reason I live in the past.

[Alex 徐瀚 @runninghsu]

Come see my Three Minute Thesis presentation on “Building Behaviors from Brain”. Today 2/6 at 5pm, Hunt Library. @CarnegieMellon @CMULibraries @cmualumnihouse

thanks to @YttriLab and @Markolas11 for making the data possible!

[Khuyen Tran @KhuyenTran16]

Have you ever taken a while to come up with the visualizations for your data? With Lux, you can get some suggestions on what to visualize.

To use Lux, simply import it.

Link to Lux: https://t.co/1oZt76xwaw

#Python

[Alex 徐瀚 @runninghsu]

@ashleyruba Read the article, specifically talking about Australia.
57% satisfied (compared to 62% in 2019)
61% workload too high (compared to 49% in 2019)
47% affected by questionable research practice
I’m guessing 48% stress daily and 58% considering leaving academia is from your polls?

[Alex 徐瀚 @runninghsu]

@laubach_mark Welcome 🐍

[Alex 徐瀚 @runninghsu]

@S__Baksh Definitely cannot relate, and sympathize with your frustration. However, this said source your provided can only be used as Long term trend and not between degrees, unless they account for average age of degree entry…

[Alex 徐瀚 @runninghsu]

Great instructions for M2 users! Retweeting for future me 📝
The comments on “why not pay for colab pro” baffles me. Clearly they don’t work on large datasets…

[Santiago @svpino]

Apple just released their MacBook Pro laptops with their M2 chip.

Their GPU is extremely fast, yet many people have no idea how to use it to run their machine learning models.

Here is how you allow TensorFlow to use your GPU:

[⚡️GREG_CORDER⚡️ @FlyBottleEscape]

Now witness the power of this fully armed and operational battle (@DeepLabCut + B-SoiD) station

[Alex 徐瀚 @runninghsu]

@theHumanBorch Really cool but I always caution with these low d color plots because you’re plotting colors sequentially…

[Alex 徐瀚 @runninghsu]

Mostly agree if you’re in a steady env and uses cloud storage.
If you change your work station every 3-5 years, 1/12 is more likely to be unique than 1/3 or 1/5.

[Brian Heseung Kim @brhkim]

The weirdest hill I will die on is that YYYY-MM-DD is absolutely the best date format. Why?

1. Big to small units, logical!
2. Month+day are helpfully contextualized by year first
3. Filename sorting is perfectly chronological without any weird logic required (see pic)

fight me

[Khuyen Tran @KhuyenTran16]

A #Python function should only do one thing and have fewer than four arguments to make testing easier.

If a function only does one task but has more than three arguments, consider grouping the arguments using higher-level objects such as a dataclass.

[Bastijn van den Boom @BastijnvdBoom]

💥Preprint!💥 I am excited to show you our work “The therapeutic mechanism of deep-brain stimulation”! @_the_ing and I have worked on this for over 4 years. Big thanks to our co-authors @AlfredoElhz, @EnnyvBeest, @aishu1803, Peter, and Damiaan Denys (1/14) https://t.co/k6VLPngsXQ