Discover and read the best of Twitter Threads about #ML

Most recents (24)

Deep dive into "ZeRO: Memory Optimizations Toward Training Trillion Parameter Models" by Samyam Rajbhandari, Olatunji Ruwase, Yuxiong He & @jeffra45

It proposes an optimizer to build huge language pre-trained models.

Thread👇🏼 🔎…
Zero Redundancy Optimizer (ZeRO) is an optimization module that maximizes both memory and scaling efficiency.

It tries to address the limitations of data parallelism and model parallelism while achieving the merits of both…

Read 7 tweets

Quite frankly, I don't (and couldn't know). I've mucked around from simple to ensemble #ML models over the last year, only to realize that something this complex cannot be modelled or predicted.

However, you can pick up what's happening, currently, based...
... off the standard epidemiological metrics - changes in trends, TPR, testing rates, recoveries etc. And what that means in terms of the near and medium-term horizons.

The most sensible benchmark would be TPR declining or levelling off against testing rates that are rising...
... against/ahead of case growth rates. If the data is passably reliable, it's a good indicator that a peak is to be expected, followed by a decline.

This would also show in R[t] hitting 1.0 on a decline and continuing to go under. If testing is doubtful, TPR is rising ...
Read 5 tweets
@PaulBarba_ held a first @joinClubhouse room with @MuazmaZahid in the Data and AI club. The topic was "How to Curate a Good Dataset for NLP?"

There were a lot of interesting questions asked and at the end of the call lots of interesting people asked for follow up notes...
Below is a thread of the room - topic, intro, why this topic, 4 main tips. I hope this adds value to people and we can do this with more calls in the future. The call itself conveyed a lot more value but I tried to highlight the important bits!
Flagging this for the folks that followed in the call @AiTechDoc @EgboDaniel1 @BabaKirito @gboye_baba @zerotousers @LahijaniAli @talks @yalda2009 @trojkast @JMontoro3 @sroussey content to follow in the thread below
Read 17 tweets
Trump was propped up by everyone standing besides him. Birx has blood on her hands.

“Leaders”, sadly, often sadly, women, like her are exactly why I left public service. Why would I rubber stamp death?

Birx and others like her are responsible for my mother’s #COVID19 death.
Compare Birx, who is on a reputation management tour, to a true #publichealth #servantleadership role model like Dr. Hasan Gokal

Is Birx defending him, her peer?

No, she’s playing victim exactly per the playbook of American-style feminism.…
It was evangelical “science” that ruled the last administration - Birx included.

Where was Birx’s loyalty?

To vetted, verified facts and preserving human life?

I judge by actions and outcomes, not by “narrative” spun after a change of administration.…
Read 14 tweets
#Universal Sign-on #VS #Single Sign-on: Imagine logging into an app.

1) First you load the site

1.5) Load the login page or locate the sign in forum

2) Then you click single sign-on with examples: @apple @google or @microsoft @facebook @twitter

3) A pop up window loads
4) In that pop up you have to either login again but to a different service or if you’re lucky already signed in

5) Grant access to the site to one of your above accounts FYI @apple has great privacy email options now
6) Pop up closes

7) you are sent to a new page which triggers the login and now the logged in version of the site loads (Multiple page loads here)
Read 10 tweets
How to invest in disruptive innovation?

Learn from investment veteran Simon Erickson the Founder & CEO of @7innovator at #ftw2021

He is focused on identifying disruptive innovation and finding developing trends before others may even be aware of them.
Simon shares 5 principles to spot disruption:

1. Theory of resource dependency
2. Small markets don't solve growth needs
3. Markets that don't exist can't be analyzed
4. Capabilities define disabilities
5. Technology supply vs. market demand

(read on for more details)
1. Theory of resource dependency

-Customers & investors dictate how the money will be spent
-Incumbents focus on sustaining technology and find it difficult to allocate capital for disruptive ideas.

$NFLX as an example
Read 8 tweets
CLIP + StyleGAN + #mylittlepony A thread 🧵starting with @ElvisPresley

"A pony that looks like Elvis Presley"
#AI #art #NLP #ML
CLIP + StyleGAN + #mylittlepony + @Beyonce
CLIP + StyleGAN + #mylittlepony + @billieeilish
Read 9 tweets
Incoming Tweet Storm ⛈

I'm so happy to announce my #TensorFlowJS book with @OReillyMedia

The book is geared towards #JavaScript devs looking to learn #AI / #ML

It's been a complete pleasure to work with #OReillyMedia

They are professionals, but they let me keep my personality which I feel is critical to all the #CodeNewbie folk in #tech.

I've learned a lot from the #DEVCommunity.

@TensorFlow is a new way of thinking for most #JSdevs. So each chapter opens with a quote and ends with a challenge.

I found it was important to get your hands dirty as early as possible to make the content engaging and fun 🎉🎉

Read 5 tweets
Let's talk about NEURAL NETWORKS. 🧠

Most of you are probably familiar with them. 🧑‍💻

But not many know HOW they actually work and—even more importantly—WHY they work. 🔧

So, let's take a journey to understand what makes Neural Networks so effective... 📊


Let's start with a simple question...

❓ What problem are NNs trying to solve? 🤔

Generally speaking, NNs are trained on examples, to produce predictions based on some input values.

The example data (input + desired output) draws a curve that the NN is trying to fit.
In a nutshell, NNs are—like most ML tools—a fancy way to fit the curves inherently generated by the examples used to train it.

The more inputs it needs, and the more outputs it produces, the higher the dimension of the curve.

The simplest curve we can fit is ...a line! 😅
Read 21 tweets
The Evolution of Kit: Automating Marketing Using Machine Learning - @ShopifyData and @ShopifyEng

This thread will pull some interesting elements out of…
by @vincentchio

1. Intro
2. Motivation
3. #ML
4. New Businesses

1 Intro
"As a virtual assistant, Kit interacts with business owners through messages over various interfaces including Shopify Ping and SMS."

Kit serves as a nice UI to make ads and helps them
"create more effective and performant ads through marketing recommendation"
2 Motivation
Initial rule-based recommendations had the budget ranges hard coded into the application where the user can can choose from.

But these may not fit their needs and it's a difficult decision to make in order to maximize returns.
Read 10 tweets
Important to keep in mind that #AI is made by humans. Our flaws and biases get encoded and amplified. We need the right diversity of people at the table to design and implement #ArtificialIntelliegence and #ML. #humancentereddesign for all of humanity. #design
Who determines definitions, labels of #data cut offs? Too often we get all nerdy on which variables & factors in the model. But are we even using the right data? #tech #technology types might consider reading Maya Angelou & Toni Morrison and others to think outside the box
Example: many people are interested in #SDoH #socialdeterminantsofhealth & #data for it.
Who has existing large #databases?
Banks are traditionally risk averse, block/exclude those who are labeled high risk
So can their data improve health equity?…
Read 5 tweets
Andy has hours of video over many years on the internet between keynotes, interviews, and more. I found this fireside chat from 2017 insightful into some of the basics that make AWS AWS. There are foundational, they're not hidden, they just execute
"It's something we passionately believe will make a huge difference in the year, and we're investing a gigantic amount of resources"
Andy on artificial intelligence and machine learning

#AI #ML #MachineLearning #ArtificialIntelligence #AndyJassy #AWS
"If you look at the history of AWS, in every part of our business, and what's driven us, is democratizing technologies that were only available to a small number of companies"
Andy on technology democratization
#AndyJassy #AWS #JeffBezos
Read 20 tweets
(1) Why DeFi needs @ReefDeFi?

🌈Reef Finance not only makes #DeFi easy, but is also essential for the progress of the DeFi industry.

The first @Polkadot project on @binance, #Reef solves the 3 biggest problems which plague DeFi today.

Read on to know more👇
(2)1️⃣Liquidity Problem

Both CEX & DEX suffer from unique problems
- Prone to hacks
- Contain isolated silos of liquidity for different coins
- List selected coins

- Difficult to use
- Unstable liquidity
- Trades are slow
- Usually suffer from congestion on #Ethereum
(3) 2️⃣Yield Farming's Technicalities

- High technical barrier of entry
- Users have to understand projects technicalities first
- Difficult to stay well diversified & keep funds safe simultaneously
- High gas fees on #Ethereum eats away profits
Read 9 tweets
*4th PLACE* of our most impactful articles published in @RSeriesa in 2020 according to @altmetric:

Tanner et al:…

*CONTEXT* Dynamic prediction models provide predicted probabilities that can be updated over time as data become available. There is growing interest in the use of #machinelearning 🤖; yet, their use in the context of dynamic survival #prediction has been limited.
*AIM* Show how #landmarking can be combined with a #machinelearning ensemble to improve prediction performance

*METHOD* Use of discrete time #survivalanalysis techniques to enable the use of #machinelearning algorithms for binary outcomes
Read 4 tweets
Machine Learning models can be classified regarding how much human supervision they need.

This affects the algorithms used and the types of tasks that it can solve.

You can categorize Machine Learning models in 4 major categories:

1/5🧵 #ML #MondayMotivation
Supervised Learning is when you train a model from the input data and ALL their corresponding labels.

Examples of
- Tasks: classification and regression
- Algorithms: kNN, Linear and Logistic regression, SVM, Decision Tree, Neural Networks(*)

Unsupervised Learning is when you use unlabelled data to train your model.

Examples of
- Tasks: Clustering, Anomaly Detection, Visualization and Dimension reduction, Association rule
- Algorithms: K-means, PCA, DBSCAN

Read 5 tweets
Reinforcement Learning and Planning? Submissions are welcome to the workshop "Bridging the Gap Between AI Planning and Reinforcement Learning (PRL)." Deadline Feb 24. Workshop date: June 8 or 9 (TBD). #RL #AI #Planning #ML #Reasoning #icaps #prl2021 /1
In the last edition, we accepted 20+ papers, had 5 invited speakers, 4 discussions and 100+ Zoom participants. See papers, posters and talks recording at /2
Why a WS on PRL? 1) Pure RL and Planning deal with different problems, but both have been looking into each other techniques for their own challenges. /3
Read 6 tweets
Yes, the #vaccine rollout is slow, uneven, not matched to need. "Botched"? That is a bit too strong. Much blame to go around.

American #healthcare is built for silos, turf, acute care, reactive..not #populationhealth #publichealth #prevention. It shows. Why is anyone surprised?
And #MedTwitter, despite your hashtags & publications, when it comes to real life actions inside departments, hospitals, or even on Twitter, all you all do is turf, ego, cliques, us vs them, etc. I am unimpressed. Build *systems* Yes #impostersyndrome makes you crave validation.
I am a #pediatrician and if you need validation, I've got a whole box of stickers for you. I will even let you choose your own sticker. How is that? Because, yes, you are special. You matter. You are wonderful. Happy?

Then after that, could we get 💩 done?
Read 12 tweets
I was a product manager on Samsung #ecommerce and we were testing a hypothesis that users were ready to adopt purchase of products via chat i.e. conversational commerce. I got a $1M budget to validate it. Here’s how lack of model explainability dooms business decisions. Read on…
We selected the most popular chat app, FB Messenger, to test this hypothesis. To accelerate the test, we ran ads with a promotion that opened a chat window directly into our commerce chatbot.
Now about the ad target. We had a database of millions of users who had previously engaged with the Samsung brand. The target of 300k users for the ad campaign was decided by an #ML model.
Read 10 tweets
A deeply interesting tutorial by @fchollet @MelMitchell1 @ChrSzegedy at #NeurIPS2020

"#Abstraction & #Reasoning in #AI systems: Modern Perspectives "

or "What are abstraction, generalization and analogic reasoning?"

@fchollet begins by laying some foundations in notation and background.

E.g., Generalization is a spectrum (and we are looking at lower bands in #ML nowadays)

And Abstraction? It is the engine behind generalization!

And of course it comes in different flavors:

- program-centric: that is akin to high-level reasoning (inducing & merging programs)

- value-centric (interpolating examples) 👈 essentially what #DL does and excels at!

Read 15 tweets
Got the @TensorFlow Developer Certification.

Thanks to the #ML @GoogleDevExpert program for sponsoring the exam.

In this thread, I will summarize my experience.

If you use @TensorFlow in your work moderately, I think you already have the prerequisites. Definitely take the *TensorFlow in Practice* specialization by @lmoroney & @DeepLearningAI_. It will get you up to speed.

Study the contents rigorously.
Review the certificate handbook carefully. It really has all the information you need to know about the certification -….

* Install @pycharm & get sufficiently comfortable with it.
* Set up the exam environment properly.
Read 6 tweets
#DSFthegreatindoors presents

Going postal: how to craft a cutting-edge route optimisation engine in-house that suits your business needs
A talk by Fabrice Durier, Hugo Galy and Louisa Sober of @RoyalMail


We are off

Going postal: how to craft a cutting-edge route optimisation engine in-house that suits your business needs
A talk by Fabrice Durier, Hugo Galy and Louisa Sober of @RoyalMail


The diverse team
Going postal: how to craft a cutting-edge route optimisation engine in-house that suits your business needs
A talk by Fabrice Durier, Hugo Galy and Louisa Sober of @RoyalMail

Read 78 tweets
Want to become a Data Scientist? Here are some great resources that you should watch in order;Statistics & Linear Algebra Not Included 🧵
#100DaysOfCode #100DaysOfMLCode #DataScience #AI #MachineLearning #ArtificialIntelligence #ML #Coding #Coders #PyTorch #Tensorflow #Developer
Part 1 of Microsoft's Intro to Python Series:
Gets you up and running in python and introduces you to the basics of setting up your development environment.…
Continuation of the series above.…
Read 17 tweets
Language Generation with Multi-hop Reasoning on Commonsense Knowledge Graph - Despite the success of generative pre-trained language models on a series of text gen. tasks they still suffer in cases where reasoning over underlying commonsense knowledge is req during generation.
Existing approaches that integrate commonsense knowledge into generative pre-trained language models simply transfer relational knowledge by post-training on individual knowledge triples while ignoring rich connections within the knowledge graph.
Multi-Hop Reasoning Flow (GRF) that enables pre-trained models with dynamic multi-hop reasoning on multi-relational paths extracted from the external commonsense knowledge graph



#NLProc #ml #PyTorch
Read 3 tweets
#DSFthegreatindoors presents

Workshop - Text Classification by Transfer learning with Deep Transformers - Depop @depop
A talk by Oduwa Edo-Osagie @odieED, Data Scientist, Depop

ND Image
About the talk

Workshop - Text Classification by Transfer learning with Deep Transformers - Depop @depop
A talk by Oduwa Edo-Osagie @odieED, Data Scientist, Depop



ND ImageImage
Read 57 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!