Discover and read the best of Twitter Threads about #DeepLearning

Most recents (24)

Amazon Deep Learning Book! 📚📊🚀

If you are looking for a resource to learn Deep Learning, I recommend checking the Dive into Deep Learning book created by @amazon scientists - Aston Zhang, Zack C. Lipton, Mu Li, and Alex J. Smola (main authors). 🧵 👇🏼 [1/n]
#DeepLearning Image
What?
It focuses on the foundation of DL, from the basic linear Neural Network to complex modeling. It covers the math behind it while illustrating the functionality with interactive examples implemented with Python libraries such as @ApacheMXNet , @PyTorch , and @TensorFlow Image
That includes the following topics:
- Linear neural networks
- Multilayer perceptrons
- Deep learning computation
- Convolutional and recurrent neural networks
- Optimization algorithms and performance
- Computer vision and NLP models
- Generative adversarial networks Image
Read 8 tweets
1/ Our contribution to drug repositioning (DR) for metabolic disease (MD)
In HUMANS Insulin resistance (IR) in muscle & adipose shares COMMON molecular pathways & COMMON treatment responses (n= >2,000 samples). @mackinprof @eLIFE @JPTK_TechBio @chvolmar
doi.org/10.7554/eLife.…
2/ Using the common molecular features from muscle & adipose we built a 120-gene RNA IR assay & found this multi-gene assay could rank families of active drugs by potency #novel Active drugs have many protein targets #not-novel
@mackinprof @eLIFE
doi.org/10.7554/eLife.…
3/ The assay identified >200 +ve & -ve acting drugs – many proven to improve or impair insulin signalling. Not all +ve drugs will be safe to explore further but many had efficacy in recent MD models @mackinprof @eLIFE
doi.org/10.7554/eLife.…
Read 10 tweets
H2O new release! 🚀🚀🚀

This week, H2O had a major release of their ML open-source library for #R and #Python, introducing two new algorithms, improvements, and bug fixing. ❤️👇🏼 🧵

#MachineLearning #ML #DeepLearning #rstats #DataScience #DataScientists
New algorithm (1/2):
✨ Distributed Uplift Random Forest (Uplift DRF) - The Uplift DRF is a tree-based algorithm that uses a Random Forecast classifier to estimate a treatment's incremental impact. See demo on the notebook ⬇️
github.com/h2oai/h2o-3/bl…
#randomforest #ML #UpLift
New algorithm (2/2):
✨ Infogram & Admissible Machine Learning - is a new tool for machine learning interpretability. More details are available on the algorithm doc ⬇️
h2o.ai/blog/h2o-relea…
#machinelearning #ML
Read 5 tweets
Did you know that An Introduction to Statistical Learning (ISLR) book has an online course? 🎥 🌈❤️
@edXOnline is offering an online course by @Stanford University, following the book curriculum: 🧵 👇🏼

#rstats #Statistics #ML #datascience
The course instructors are two of the book authors - Prof. Trevor Hastie and Prof. @robtibshirani. While the book is based on #R some awesome people translate it to #python, #julialang, and other #Rstats flavors (see links on the comments below 👇🏼).
The course covers the following topics (aligned with the book curriculum):
Read 9 tweets
[🧠 Paper Summary 📚] An interesting paper was recently published to arxiv: "Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets" (although it originally appeared in May 2021).

The main idea is this:
1/ 🧵
If you have an overparametrized neural network (more params than the # of data points in your dataset) and you train it way past the point where it has memorized the training data (as suggested by the low training loss, and high val loss), all of a sudden the network will...

2/
learn to generalize, as suggested by a rapid decrease in the val loss (colloquially known as "grok" i.e. the NN understood/figured it out).

Practitioners usually stop training networks at the 1st sign of overfitting (as evidenced by an increasing gap between train/val loss).

3/
Read 7 tweets
[🧠 collective intelligence 🧠] I've been intrigued by the cellular automata (CA) concept for a long time, and by the potential, mutually beneficial interaction between CAs and deep learning, so I decided to dig a bit deeper. Here are some interesting resources I found:

1/🧵
distill.pub/2020/growing-c… <- one and only Distill. A nice introduction to how neural CA works and how it may be a potentially useful model of morphogenesis and regeneration processes in developmental biology.

@zzznah

2/
(think how we humans are formed from a single egg cell - when and how does this multiplication of cells stop and stays stable?)

3/
Read 8 tweets
🙌 Today was a great example of how the #rstats community can help getting learning resources! Thank you for all the amazing material about #deeplearning & #reproducibility with R 🙏 This feels like a good preamble to the remaining poll results about learning strategies 👇 Results to poll about learn...
With all these materials out there I am now wondering when will I have time to read it all, same with practicing code and the new skills I will learn after going through them!
68% of replies claimed to practice when they have time, but I think this could have been more accurate 👇
Read 7 tweets
2/n
Types of magic commands

Line magics - starts with % character. Rest of the line is its argument passed without parentheses or quotes.

Cell magics - %% - can operate on multiple lines below their call.
#DataScience #MachineLearning #100DaysOfMLCode #Python #DataScientist
Read 21 tweets
2/n

GauGAN2 combines segmentation mapping, inpainting and text-to-image generation in a single model

#Computervision #AI #ArtificialIntelligence #TensorFlow #PyTorch #DeepLearning #DataScience #MachineLearning #100DaysOfMLCode #Python #DataScientist #Statistics #Mathematics
3/n

Unlike GauGAN1 the GauGAN2 can translate natural language descriptions into landscape images. Typing a phrase like “sunset at a beach” generates the scene

#Computervision #AI #ArtificialIntelligence #TensorFlow #PyTorch #DeepLearning #DataScience #MachineLearning #Math
Read 7 tweets
Image interpolation occurs when you resize or distort your image from one pixel grid to another.

1/n

#computervision #IMAGE #DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #programming #Math #Stat #dataviz #DataAnalytics #AI #ArtificialIntelligence #data
Image interpolation works in two directions, and tries to achieve a best approximation of a pixel's intensity based on the values at surrounding pixels.

2/n

#computervision #IMAGE #DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python #programming #Math #Stat
Image resizing is necessary when you need to increase or decrease the total number of pixels, whereas remapping can occur when you are correcting for lens distortion or rotating an image.

3/n
#computervision #DataScience #MachineLearning #DeepLearning #100DaysOfMLCode #Python
Read 8 tweets
زندگی میں اک مقام درد کا ہوتا ہے
جسں میں انسان کو کِسی نہ کِسی سہارے کی ضرورت ہوتی ہے
اور محبتوں کے دعویدار اکثر ہمیں اسی مقام پہ تنہا کر جاتے ہیں
اور جب انسان گرتے پڑتے اس مقام کو پار کر جائے تب اسکی تنہائی کامل ہوجاتی ہے
اور اسے جذباتی طور پہ کسی انسانی سہارے
#DeepLearning
++
جذباتی طور پہ کسی انسانی سہارے کی ضرورت نہیں رہتی اور مقامِ درد پہ چھوڑ جانے والوں کی چاہتیں تب بہت بےمعنی سی لگنے لگتیں ہیں
اور ان چند سالوں نے زیادہ کچھ نہیں بدلا بس مجھے مقامِ درد سے نکال کر میری تنہائی کو کامل کر دیا ہے

#DeepLearning
#LateNightVibes
Read 3 tweets
1/5 Finding a do-operator in a @DeepMind article is a tectonic progress that deserves welcoming blessing. The "delusions" treated in this article are endemic of "Evidential Decision Theory" which Causality (ch 4.1.1 ucla.in/38bmhnO)
summarizes in a mnemonic limerick:
2/5
- Whatever evidence an act might provide
- On what could have caused the act,
- Should never be used to help one decide
- On whether to choose that same act.
Typical real life ramifications of these delusions are:
(1) patients should avoid going to the doctor “to reduce the
3/5
probability that one is seriously ill”
(2) workers should never hurry to work, to reduce the probability of having overslept, and more.
The deployment of the do-operator eliminates these "delusions" and has led to the "sequential backdoor criterion" of Sec. 4.4.3.
Read 5 tweets
Laura Manocci (biologiste) et ses confrères ont imaginé une façon de recenser certaines espèces de la faune sous-marine calédonienne !
1) Les animaux sont filmés en survolant le lagon grâce à un ULM.
(1/3)
2) Ensuite, le #DeepLearning est utilisé comme outil de reconnaissance de forme grâce aux CNN, un type de réseau neuronal artificiel, pour identifier et traiter les images. Les animaux sont classés par comparaison avec les vidéos en provenance d’Internet.
(2/3)
Un bon outil pour protéger la faune comme certains requins, les tortues ou encore le… dugong.
Source :
actuia.com/actualite/lint… (@ActuIAFr)
(3/3)
Read 3 tweets
Web 3.0 part 2.

You know, everyone is talking about Web 3.0 on twitter or anywhere and you probably might not even have gotten a hang of it.

I kid you not, no matter how technical the name is sounding, it's actually something you can understand.

A thread 🧵.
Before a 3.0, there was 1.0 and 2.0.

So let's do a run down from 1.0.

The first stage of the web started in mid 90's and that's what we know today as 1.0.

They were the first stage of the world wide web (www) and they were static pages.

I mean like just static. No convo
Users passively consume contents. They couldn't yunno give feedback, or create contents.

1.0 was just one big Wikipedia of hyperlinks.

What we know today as the dotcom era.

Actually, 1.0 fundamentally changed the world. It opened internet. Opened internet potentials.
Read 14 tweets
Deep learning & theoretical machine learning instruction at UTexas (Masters) is the real deal. Here, reading one of professor's recent papers on keypoint object detection - we are expected to code this on our own. B/t this, & Adaboost theory, fascinating stuff. @haltakov
And @AlejandroPiad , I was preping for exams yesterday & came across this, which says that if you don't assume that a consistent (zero-training-error) hypothesis is possible, it is NP-hard to employ Empirical Risk Minimization to estimate generalization error w/ in certain bounds
Notice the part about NP=P which you tweeted about. Fascinating stuff, looking forward to the midterm exams to then focus on object detection using heatmaps and keypoint detection..#deeplearning #machinelearning #theory #practical #SOTA
Read 3 tweets
#ArtificialIntelligence and ECG by Fabio Badilini #ESCDigitalSummit @escardio /1 A lot of research on the field in the last years ImageImageImageImage
#ArtificialIntelligence and ECG by Fabio Badilini #ESCDigitalSummit @escardio /2 the different ways of #CNN #DeepLearning and the value on Cardiology ImageImageImageImage
#ArtificialIntelligence and ECG by Fabio Badilini #ESCDigitalSummit @escardio /3 The importance of type of input signals, data and clinical output ImageImageImageImage
Read 5 tweets
Memory Efficient Coding in #PyTorch
20 tricks to optimize your PyTorch code

Let's break some of the bad habits while writing PyTorch code 👇

A thread 🧵
1. PyTorch dataloader supports asynchronous data loading in a separate worker subprocess. Set pin_memory=True to instruct the DataLoader to use pinned memory which enables faster and asynchronous memory copy from host to GPU Image
2. Disable gradient calculation for validation or inference. Gradients aren't needed for inference or validation, so perform them within torch.no_grad() context manager. Image
Read 25 tweets
A paper on "AlphaFold-multimer", a version of AlphaFold that works on protein complexes, was released by @DeepMind.

Accurately predicted structures can lead to better understand the function of such protein complexes that underpin many biological processes!

#DeepLearning 1/4
Before "AlphaFold-multimer", people discovered that AlphaFold can predict complexes if you connect them with a long linker (this tweet was cited in the above paper!) 2/4
The new model, which had various adjustments to handle the larger protein complex structures, shows improved performance over this linker approach, along with other approaches 3/4
Read 4 tweets
1/7 #SeaTurtleTalks
Automatic identification of immature green turtle behavior from accelerometer reveals key marine conservation area in the Caribbean

Accelerometer records animal movement & body posture. Signal interpretation can be difficult & limits its use on #seaturtle
2/7 We deployed multi-sensor recorders (accelerometer, gyroscope and time-depth recorder) combined with animal-borne video-recorder to validate the signal on immature green turtles in Martinique.
3/7 From labelled signals, we trained a #deeplearning algorithm, #Vnet, to automatically identify green turtle #behavior.

🚩High ability to detect behaviors with low discriminative signals such as Feeding and Scratching, and precision down to one centi-second. #rblt
Read 7 tweets
Surprised #deeplearning methods were not massively used to detect Covid? Some tried, but most failed: e.g. they rely on patients' age instead of 'truly' analyzing the medical scans. This is because networks learn correlation and not causation. Our preprint Fishr tackles this. 1/4
When dealing with multiple datasets, we intuited that the learning should unfold consistently across those: i.e., the gradient distributions should be independent of the dataset. After a few approximations, we simply bring closer the dataset-level gradient variances! 2/4
Our approach is closely related to the Fisher Information and the Hessian. This explains the name of our work - Fishr - and why this aligns the loss landscapes per dataset. This works beyond expectations, outperforming all previous methods on the public DomainBed benchmark 3/4
Read 4 tweets

Related hashtags

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!