Masthash

#DeepLearning

Hourly fractals
20 minutes ago

Surprise! Newton! #DeepLearning #VisualArt

Hourly fractals
5 hours ago

Hope you have an amazing Sunday! #DeepLearning #VisualArt

Rene Schulte
5 hours ago

I always like to ride a good ol' steam train, even better in mountain regions. It also reminds me that these majestic steam locomotives, are a symbol of the Industrial Revolution that reshaped our world. ๐Ÿš‚

What is the driving force of today's industrial revolution? ๐Ÿค”

AI? ๐Ÿง 

#IndustrialRevolution #AIRevolution #Innovation #Progress #AI #deeplearning #llm #SteamLocomotive #steam #train #FutureOfWork

Hourly fractals
7 hours ago

Hope you're having a great Sunday! #DeepLearning #Math

Hourly fractals
11 hours ago

Have an awesome Sunday! #DeepLearning #VisualArt

PAUL COGAN
22 hours ago

Smart writing on #AI and overhyped expectations of creativity from a generic amalgamator. Measuring its value has me thinking of the clarity and murkiness of academic assessment. https://apperceptive.substack.com/p/th
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience
#tech #humanintelligence #gai

PAUL COGAN
23 hours ago

Austin-based Diligent Robotics, which makes โ€œsocially intelligentโ€ hospital service robots, raised $25M and plans to have 100+ robots in 22 US hospitals by 2024
https://techcrunch.com/2023/09/21/diligent-raises-25-million-to-triple-its-nursing-robots-reach/
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience
#tech #humanintelligence #gai

PAUL COGAN
23 hours ago

CLSA: China has 130+ LLMs, or 40% of the global total and behind the US' 50% share; experts say most have yet to find viable business models and are too similar
https://www.reuters.com/technology/chinas-ai-war-hundred-models-heads-shakeout-2023-09-21/
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience
#tech #humanintelligence #gai
#china

PAUL COGAN
1 day ago

Sequoia published a big update to their generative AI thesis shared one year ago. They now say that technological advancements like Moore's Law, the internet, and cloud computing have driven generative AI to constitute a significant innovation milestone.
https://www.sequoiacap.com/article/generative-ai-act-two/?utm_source=aittention.beehiiv.com&utm_medium=newsletter&utm_campaign=private-alternative-to-chatgpt-chinese-ai-influencers-and-the-reversal-curse-today-s-top-ai-news-9-22-23
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience
#tech #humanintelligence #gai

PAUL COGAN
1 day ago

Daniel Huynh introduced BlindChat, a privacy-centric, open-source alternative to ChatGPT. Derived from Hugging Face Chat-UI, BlindChat operates on the client side using transformers.js for local inference and stores chats in the browser cache, ensuring user privacy.
https://huggingface.co/blog/dhuynh95/introducing-blindchat?utm_source=aittention.beehiiv.com&utm_medium=newsletter&utm_campaign=private-alternative-to-chatgpt-chinese-ai-influencers-and-the-reversal-curse-today-s-top-ai-news-9-22-23
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience
#tech #humanintelligence #gai

PAUL COGAN
1 day ago

This article from the Mozilla Foundation surfaces into the human decisions that shape generative AI. It highlights the ethical and regulatory implications of these decisions, such as data sourcing, model objectives, and the treatment of data workers.
https://thoughtshrapnel.com/2023/09/23/if-llms-are-puppets-whos-pulling-the-strings/
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience
#tech #humanintelligence #gai

Hourly fractals
1 day ago

Newton vibes to brighten your Saturday. #AI #DeepLearning

thetestspecimen
1 day ago

Julia is the new upstart in the data science world, but can it keep up with the tried and tested combination of Python + NumPy + Numba?

https://towardsdatascience.com/is-julia-faster-than-python-and-numba-897c53694621

#DataScience #deeplearning #MachineLearning #Python #NumPy #Numba #JuliaLang

Hourly fractals
1 day ago

Surprise! Newton! #Art #DeepLearning

Hourly fractals
1 day ago

Smile on this lovely day! #AI #DeepLearning

Hourly fractals
2 days ago

Good Saturday! This Mandelbrot Set is for you! #DigitalArt #DeepLearning

Hourly fractals
2 days ago

Hope your Friday is awesome! #Science #DeepLearning

Hourly fractals
2 days ago

Keep shining on Friday! #Art #DeepLearning

Hourly fractals
2 days ago

Surprise! Mandelbrot Set! #DeepLearning #Nature

Hourly fractals
2 days ago

Cheers to a fantastic Friday! #DeepLearning #Fractals

Hourly fractals
2 days ago

Wishing you a joyful day! #MachineLearning #DeepLearning

Hourly fractals
2 days ago

Guess what? It's Mandelbrot Set! #DigitalArt #DeepLearning

Big things come in small packages! ๐Ÿ‘ฅ๐Ÿ†

Check out this inspiring list of small teams that have achieved great things.

๐ŸŒŸ๐ŸŽ‰ From startups to non-profits, these teams prove that with the right mindset, strategy, and collaboration, anything is possible.

Check it out here: https://stevepulec.com/posts/small/
My newsletter subscribers learned about this 6 months ago!
https://dramsch.net/newsletter

โ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโœโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆ
๐Ÿ‘‹ Moin, I'm Jesper!

I share awesome machine learning finds like this every day to help you build better real-world ML!

๐—™๐—ผ๐—น๐—น๐—ผ๐˜„ ๐—บ๐—ฒ to stay in the loop!

#Data #Python #LateToTheParty #DataScience #Kaggle #DeepLearning #Career

PhantaNews โœ…
2 days ago

GRRM und John Grisham gegen #OpenAI

Ich bin SEHR gespannt, wie das ausgeht, denn am Ende mรผssten die Autoren Open AI und Microsoft vermutlich nachweisen, dass sie tatsรคchlich Material von den Autoren verwendet haben. Und in Europa ist Datamining grundsรคtzlich erlaubt. Da eBooks am Ende auch nur html-basierte Daten sind ... (allerdings meist nicht frei verfรผgbar)

#ChatGPT #KI #AI #DeepLearning

https://www.huffpost.com/entry/john-grisham-george-rr-martin-openai-copyright-infringement_n_650b3f3ee4b097b86b3bb4da

Hourly fractals
2 days ago

Have an extraordinary Friday! #DeepLearning #Nature

Hourly fractals
2 days ago

Wishing you a fantastic day! #DeepLearning #AI

Hourly fractals
3 days ago

Today is awesome! #Art #DeepLearning

Hourly fractals
3 days ago

Enjoy your Thursday! #DeepLearning #DigitalArt

Hourly fractals
3 days ago

Let this Newton inspire your Thursday. #DeepLearning #VisualArt

Hourly fractals
3 days ago

Smile on this lovely day! #DeepLearning #Programming

C & C++ Weekly
4 days ago

Today's reading:
Normalization Techniques in Training DNNs: Methodology, Analysis and Application.
Lei Huang, Jie Qin, Yi Zhou, Fan Zhu, Li Liu, and Ling Shao.

A review and commentary of normalization methods in DNNs. I skimmed this one. Good for definitions of all of the normalization terms and especially Figure 1.

arXiv: https://doi.org/10.48550/arXiv.2009.12836
Publisher: http://doi.org/10.1109/TPAMI.2023.3250241
pdf: https://arxiv.org/pdf/2009.12836.pdf

bibtex etc. https://amytabb.com/today-i-read/2023/09/20/tir/
#AmyPostsPapers #deepLearning #normalization

The different normalization operations -- centering, scaling, decorrelating, standardizing, and whitening.
PAUL COGAN
4 days ago

Google updates Bard to use data from Gmail, Docs, and Drive, not just the web, to help users find and summarize emails, highlight points in a document, and more
https://www.theverge.com/2023/9/19/23878999/google-bard-ai-chatbot-gmail-docs-drive-extensions
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience
#tech #humanintelligence

Interested in the global AI talent landscape? ๐ŸŒŽ๐Ÿ‘ฅ

Check out MacroPolo's Global AI Talent Tracker!

This interactive tool lets you explore trends in AI talent across countries and cities, and compare the relative strengths of different regions in the field.

Check it out here: https://macropolo.org/digital-projects/the-global-ai-talent-tracker/
My newsletter subscribers learned about this 12 months ago!
https://dramsch.net/newsletter

โ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโœโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆ
๐Ÿ‘‹ Moin, I'm Jesper!

I share awesome machine learning finds like this every day to help you build better real-world ML!

๐—™๐—ผ๐—น๐—น๐—ผ๐˜„ ๐—บ๐—ฒ to stay in the loop!

#Programming #DeepLearning #DataScience #Coding #Visualization #Python #LateToTheParty

Alan Kotok
5 days ago

Researchers from a university biochemistry lab and company developing synthetic enzymes designed techniques using engineered bacteria to detect buried land mines at longer distances.

https://sciencebusiness.technewslit.com/?p=45222

#News #Press #Science #Business #LandMines #Biotechnology #Biochemistry #Biosensor #Enzymes #Proteins #SyntheticBiology #ArtificialIntelligence #DeepLearning #Algorithms #Israel #EColi

Looking to make your machine learning research reproducible? ๐Ÿ“š๐Ÿ’ป

Check out ml.recipes!

This collection of "easy wins" and easy-to-use recipes focuses on basics that work and gets you 90% of the way to top-tier reproducibility.

Check it out here: https://ml.recipes

โ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโœโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆ
๐Ÿ‘‹ Moin, I'm Jesper!

I share awesome machine learning finds like this every day to help you build better real-world ML!

๐—™๐—ผ๐—น๐—น๐—ผ๐˜„ ๐—บ๐—ฒ to stay in the loop!

#Jupyter #MachineLearning #Coding #DeepLearning #Career #Kaggle #Communication #Software #Ml4science

PAUL COGAN
6 days ago

What happens, when different groups of people talk to #AI about science? Kaiping Chen and her team investigated how #GPT fared in conversations on controversial science topics as well as other social issues. In our interview, she shares their discoveries ๐Ÿ‘‰https://www.wissenschaftskommunikation.de/
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience
#tech #humanintelligence

PAUL COGAN
6 days ago

AI fake news detector helps detect fake news through binary classification methods. It helps build experiences by controlling the flow of disinformation. It's built on top of various powerful machine learning libraries. The tool works by training a neural network to spot fake articles based on their text content.
https://kandi.openweaver.com/collections/artificial-intelligence/build-ai-fake-news-detector
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience

Hourly fractals
6 days ago

Cheers to a fantastic Monday! #DeepLearning #AI

#AI #DeepLearning isn't as fascinating as many people believe. Image recognition AIs, for instance, mostly rely on #ImageNet which has 14 million hand classified images. This mean AI's can at best perceive reality as good as humans can. Of course, they are faster and more accurate. However, AIs can never see through reality beyond human visual perception and classification, at least not Deep Learning #ais

https://paperswithcode.com/dataset/imagenet

PAUL COGAN
1 week ago

Here come the โ€œmetahumansโ€: Virtual avatars have real jobs in Indonesia. Metahumans are occupying trusted roles like news anchors and government communicators.
https://restofworld.org/2023/metahumans-go-mainstream-indonesia/
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience

PAUL COGAN
1 week ago

The workers at the frontlines of the AI revolution. The global labor force of outsourced and contract workers are early adopters of generative AI โ€” and the most at risk.
https://restofworld.org/2023/ai-revolution-outsourced-workers/
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience

PAUL COGAN
1 week ago

Forget subtitles: YouTube now dubs videos with AI-generated voices. YouTubeโ€™s automatic dubbing system is competing with human firms to help creators scale the language barrier.
https://restofworld.org/2023/youtube-ai-dubbing-automated-translation/
#AI #AGI #artificialintelligence #intelligence #ai #generative #machineLearning
#AI #AInews #AIupdate #GenerativeAI #dotnet #LLM #deeplearning #DataScience

There is a fundamental difference between traditional #DeepLearning and #LargeLanguageModels, because in traditional deep learning you train the model multiple epochs until convergence, and in #LLMs it's just one epoch, typically in fine-tuning as well.

In multi-epoch training, the training task becomes of memorization. The training objective improves and loss decreases as the training example is memorized better. Generalization happens, but isn't the quality which is actually measured.

In a single epoch training, every training example is shown to the network only once. Still the training loss decreases every step, for new examples the model has never seen before! That's a fundamentally different task, a different objective, and the loss gradients are different as they cannot utilize any knowledge of this example having been seen before.

I don't think many people appreciate this fundamental difference. Even though the loss function is the same, the architecture is the same, the training regime is similar except this one small difference, it is a fundamentally different task.

In a single epoch learning the network is directly trained to generalize, not memorize. Any memorization that happens is just a side effect, like in a multi-epoch training any generalization is a side-effect of memorization.

There are indications that training LLMs multiple epochs or showing the same document to them more than once in training causes kind of catastrophic forgetting for any generalization they have learned.

I think this could be formulated in formal mathematical terms as well, but it's a bit tricky and probably requires new formalism.

This isn't new, by the way, just badly understood. We have trained neural models on synthetic, simulated or generated data where no training example repeats for decades.

"Overfitting? Pfft. I just don't give it any data twice."

In deep reinforcement learning especially overfitting has never been an issue because there as well (with some exceptions in replay trajectories and such) you don't tend to reuse training examples but simply generate new trajectories every round.

But weirdly enough I don't think anyone has formulated this mathematically. Not many people even speak of it.

It doesn't seem to be a coincidence that the first echoes of true generalist AI come from large reinforcement learning agents and LLMs which have been trained in this manner.

Alex Zap
1 week ago

#Python #AI #deeplearning #DataScience #wildfire #classification #riskmanagement #environment
2023 is one of the worst years for forest fires. We have implemented the fast TensorFlow CNN algorithm in Python tested on public-domain fire images.

http://newdigitals.org/2023/09/16/image-based-fast-forest-fire-detection-with-tensorflow/

Learn about Stable Diffusion ๐Ÿงฎ๐Ÿ“ˆ

with Jay Alammar's visual explanation

Illustrative animations to clarify the method's intuition and demonstrate its efficacy in machine learning.

Check it out here: https://jalammar.github.io/illustrated-stable-diffusion/
My newsletter subscribers learned about this 11 months ago!
https://dramsch.net/newsletter

โ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโœโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆ
๐Ÿ‘‹ Moin, I'm Jesper!

I share awesome machine learning finds like this every day to help you build better real-world ML!

๐—™๐—ผ๐—น๐—น๐—ผ๐˜„ ๐—บ๐—ฒ to stay in the loop!

#DeepLearning #Coding #LateToTheParty #MachineLearning #ArtificialIntelligence #Ml #Tech

Wout Bittremieux
1 week ago

I enjoyed presenting our work on #DeepLearning to uncover the #immunopeptidome at the #BMSS2023 conference.

For more information:
- Slides: https://doi.org/10.5281/zenodo.8346713
- timsTOF fragment ion intensity prediction: https://doi.org/10.1101/2023.07.17.549401
- Casanovo: https://doi.org/10.1101/2023.01.03.522621

Why investigate the immunopeptidome?
Fine-tuning the Prosit HCD model
Casanovo is a de novo sequencing transformer
Casanovo outperforms other methods

Bugs in #DeepLearning systems are extremely common even in reference implementations of common models.

This is mainly because in normal software engineering a bug causes a failure of some kind, a compilation error or incorrect behavior which can be caught by unit tests.

In deep learning the issue typically just causes slightly degraded performance. Degraded compared to what? Well, compared to a correct implementation which typically isn't available.

The traditional software engineering quality assurance methods and processes are very unsuitable for catching these sorts of bugs. We need to compensate that with an increased weight on review, checking the math, and checking numerical stability issues by staring at the code instead of trusting the problems will be caught by any other means.

Review and re-check is the only way to mitigate the prevalent bugs in deep learning systems. More eyeballs is more. Never trust, always check.

What happens to โ€œyourโ€ Wikipedia entry after death? ๐Ÿ’€๐Ÿ“š๐ŸŒŽ

Explore this fascinating visualization bei The Pudding!

Check it out here: https://pudding.cool/2018/08/wiki-death/
My newsletter subscribers learned about this 6 months ago!
https://dramsch.net/newsletter

โ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโœโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆ
๐Ÿ‘‹ Moin, I'm Jesper!

I share awesome machine learning finds like this every day to help you build better real-world ML!

๐—™๐—ผ๐—น๐—น๐—ผ๐˜„ ๐—บ๐—ฒ to stay in the loop!

#DeepLearning #ArtificialIntelligence #Visualization #Programming #Technology #LateToTheParty #Software

#DeepLearning seems to somehow sit between high-dimensional signals and low-dimensional signals.

High-dimensional signals are handled by e.g. convolutional filters, or dense layers, either exploiting domain symmetries or lottery ticket hypothesis where the well-performing randomly initialized weights become utilized and the rest scrapped.

Attentional layers utilize low-dimensional keys and values, typically ~128 dimensions where distance measures are still meaningful, projected into this low dimensionality from high-dimensional embeddings, and after attention, concatenated across heads and again projected and processed in high-dimensional space.

Ultimately segmentation and classification tasks again project the outputs into low dimensionality, either per pixel or e.g. per classification or regression output.

I think much of the power of deep learning comes from this interplay of high- and low-dimensional representations intertwined with manifold learning.

blake shaw
2 weeks ago

Put your most realistic #LLM prompt in a safe place, and come back in 10 years. This is how real its gonna look.

You're not experiencing a revolution on par with the printing press. You're experiencing the sublime effect of being dominated by the latest iteration of the Spectacle, just as the stupefying effects of its former appearance begins to wane.

#ai #deeplearning

The title screen of Microsoft Flight Simulator 5.0 from 1993, with what were the cutting edge 3D graphics at the time, which now seem as retro as 3D gets
A view from the cockpit landing in NYC in Flight Simulator 5.0
Alan Kotok
2 weeks ago

Results of a clinical trial show one radiologist using software with artificial intelligence can detect more breast cancer cases in mammograms than current reviews by two radiologists.

https://sciencebusiness.technewslit.com/?p=45194

#News #Press #Science #Business #Cancer #BreastCancer #Mammogram #ClinicalTrial #ArtificialIntelligence #Algorithms #DeepLearning #Radiologist #Sweden #SouthKorea #Diagnostics #Detection

The struggle of training deep learning models and being deeply impatient

I love a good deep learning model. I do.

However, I am no stranger to the challenges of training neural networks.

There's one problem that constantly nagged at me: The excruciatingly slow training times.

It was like watching a snail in a world of cheetahs, and I knew I needed a solution.

WHY WE TRAIN MODELS IN THE REAL WORLD
The situation got even worse when I started working on bigger projects that involved satellite data, radar data from trains, or what I do today, improving weather forecasts using machine learning.

Read moreโ€ฆ [https://dramsch.net/articles/the-struggle-of-training-deep-learning-models-and-being-deeply-impatient/] (2 min remaining to read)

...

#DeepLearning #MachineLearning #Training

Find more here: https://dramsch.net/articles/the-struggle-of-training-deep-learning-models-and-being-deeply-impatient/

Sharon Machlis
2 weeks ago

Deep Learning with torch in R: 2-hour workshop by Daniel Falbel, maintainer of the {torch} #rstats ๐Ÿ“ฆ and a @Posit software engineer. 28 Sept at 18:00 CEST/noon EDT/9 am PDT
This is a Workshop for Ukraine: Cost is 20 euro / $20 donation.
https://sites.google.com/view/dariia-mykhailyshyna/main/r-workshops-for-ukraine#h.bjmk3ceostwa

#DeepLearning #torch @rstats

Dan Stowell
2 weeks ago

Announcing fully-funded PhD positions on our new "Bioacoustic AI" project: bioacousticai.eu Apply now for a #PhD studying animal sounds (#bioacoustics), #deeplearning, #acoustic signals, and #ecology! 2 open to apply now, 8 more coming soon. (Pls share)

Logo. We study birds, bats, hyenas and more!

Making weather forecasting machine learning models operational!

As a team at ECMWF we have open-sourced "ai-models" and plugins for all the major open-source data-driven NWP models:

๐ŸŒ FourCastNet v2 with spherical harmonics by NVIDIA
๐Ÿค– PanguWeather 3D transformer by Huawei
๐ŸŒ GraphCast multi-mesh graph neural network by Google DeepMind

View them on the ECMWF website with the charts you know.

Or even run them yourself!
๐ŸŒ pip install ai-models-fourcastnetv2
๐Ÿค– pip install ai-models-panguweather
๐ŸŒ pip install ai-models-graphcast

These are all open-source plugins that make it easy to load data from MARS if you have access, CDS, or your own grib files.

Super proud of our work so far and that we can run these alongside our physical model now as a service to the weather community. ๐ŸŒฆ

Also, can we talk about running, ONNX, Pytorch, and Jax for this? Now just waiting for a Tensorflow model to fill my Pokedex. ๐Ÿ‘€

#MachineLearning #WeatherForecasting #DeepLearning #MLOps #Tech

(Pst, we're hiring btw! ๐Ÿ”ฅ)

๐Ÿ“ฃ New article on my website live!

The struggle of training deep learning models and being deeply impatient

Check it out here:
https://dramsch.net/articles/the-struggle-of-training-deep-learning-models-and-being-deeply-impatient/

#DeepLearning #MachineLearning #Training

Harald Sack
3 weeks ago

Of course I know RDF2vec ;-)
However, Heiko Paulheim's RDF2vec website has grown nicely and has become a rather valuable resource for this #knowledgegraph embedding method, including implementations, models and services, variations and extensions, as well as more than 150 references in scientific papers!
RDF2vec website: http://www.rdf2vec.org/
original paper: https://madoc.bib.uni-mannheim.de/41307/1/Ristoski_RDF2Vec.pdf
Petar Ristoski's PhD thesis (with RDF2vec): https://ub-madoc.bib.uni-mannheim.de/43730/1/Ristoski_PhDThesis_final.pdf
#kge #deeplearning #embeddings

Screenshot of the RDF2vec website. 
About RDF2vec:
RDF2vec is a tool for creating vector representations of RDF graphs. In essence, RDF2vec creates a numeric vector for each node in an RDF graph.

RDF2vec was developed by Petar Ristoski as a key contribution of his PhD thesis Exploiting Semantic Web Knowledge Graphs in Data Mining [Ristoski, 2019], which he defended in January 2018 at the Data and Web Science Group at the University of Mannheim, supervised by Heiko Paulheim. In 2019, he was awarded the SWSA Distinguished Dissertation Award for this outstanding contribution to the field.

When I interviewed at Amazon they asked me to explain Gradient Descent.

Iโ€™m happy where I am now, but letโ€™s talk about the algorithm and how machine learning actually influenced computational physics,

โ€ฆ and give you some fun facts for your next interview!

Gradient descent is the optimisation algorithm at the core of neural networks and SVMs.

The big idea is that we can calculate how wrong the prediction of our neural net is, using our loss.

Then we can calculate the gradient and propagate it through the network using the chain rule. After all, deep learning is the most lucrative application of the chain rule in the world.

Now, the problem is that weโ€™d have to calculate the loss on all of our data to get the best gradient. Something we figured out in physics a long time ago.

#machinelearning #python #deeplearning #career

Get ahead in deep learning with The Matrix Calculus You Need For Deep Learning ๐Ÿง ๐Ÿ’ป๐Ÿ”ข

by Terence Parr and Jeremy Howard

This comprehensive online book covers all the essential matrix calculus concepts you need to know to understand and implement deep learning algorithms, including derivatives, gradients, chain rule, backpropagation, and more. Whether you're a beginner or an experienced practitioner, this resource is an invaluable reference for anyone who wants to master the mathematical foundations of deep learning.

Check it out here:https://explained.ai/matrix-calculus/
My newsletter subscribers learned about this 13 months ago!
https://dramsch.net/newsletter

โ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโœโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆ
๐Ÿ‘‹ Moin, I'm Jesper!

I share awesome machine learning finds like this every day to help you build better real-world ML!

๐—™๐—ผ๐—น๐—น๐—ผ๐˜„ ๐—บ๐—ฒ to stay in the loop!
#LateToTheParty #DeepLearning #Python #Technology #MachineLearning #Software #Tech

Metin Seven ๐ŸŽจ
4 weeks ago

Anyone tried https://character.ai yet? It's an AI chatbot platform created by former Google engineers.

#AI #ML #ArtificialIntelligence #MachineLearning #DeepLearning #chat #ChatGPT #ChatBot

Ready to learn from failure? ๐Ÿค–๐Ÿ‘จโ€๐Ÿซ

Join me at PyData as I share insights on failed machine learning models from real-world examples in science and industry.

Let's avoid common pitfalls and maybe even achieve success!

Check it out here:https://www.youtube.com/watch?v=sfzWdCJmwiI

โ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโœโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆโ”ˆ
๐Ÿ‘‹ Moin, I'm Jesper!

I share awesome machine learning finds like this every day to help you build better real-world ML!

๐—™๐—ผ๐—น๐—น๐—ผ๐˜„ ๐—บ๐—ฒ to stay in the loop!
#DeepLearning #MachineLearning #Data #Ml #ArtificialIntelligence #Pydata #Programming #Python

Joxean Koret (@matalaz)
4 weeks ago

One question for ML people: Machine learning was "taught" to me, back in the day as something that should only be used when 1) There is no exact #algorithm or tool to code what we want, and, 2) We don't need exact answers (ie, 100% accuracy).

Does this still hold true? My understanding is that it hasn't ever changed, or did it?

#MachineLearning
#ML
#DeepLearning
#LinearRegression

Kathy Reid
4 weeks ago

It's a while since I've done an #Introductions #ConnectionList #TwitterMigration post, where I use my large reach to more richly connect the #Fediverse :fediverse: by finding interesting accounts you may wish to follow.

@tero ๐Ÿ‡ช๐Ÿ‡ธ posts about #technology, #DeepLearning, #AI and #Linux ๐Ÿ‘‹

@crossref is the official account of the #CrossRef not for profit, that works with #research #MetaData

@ANUResearch ๐Ÿ‡ฆ๐Ÿ‡บ seems to be official and active, although I don't know the person behind it. Posts about #research from #ANU

@jessperriam is a #sociologist and assistant #professor #researcher at #UniversityOfCopenhagen #Copenhagen ๐Ÿ‡ฉ๐Ÿ‡ฐ who is into #STS and #DigitalMethods

@djnavarro does a lot of work in #DataScience and #GenerativeAI #GenerativeArt ๐ŸŽจ

@notsolonecoder is Katie ๐Ÿ‡ฆ๐Ÿ‡บ and she's new to #Mastodon - please make her feel welcome! She is into #DevOps and #SRE and is also an #educator ๐Ÿ

@dalboz is into #technology, #scifi, #switch, #xbox and #gaming, and is new here too. Please make him welcome! ๐Ÿ‡ฆ๐Ÿ‡บ

That's all for today, please do share great people to follow, so we can curate our wonderful communities here โค๏ธ

Complexity and extensive customization for your neural network are NOT always indicators of a good system.

While customization is important, excessive complexity can lead to confusion, reduced reproducibility, and difficulty in maintaining and scaling your projects.

In the context of managing configuration files for neural networks, it's tempting to create highly detailed and intricate configuration files that account for every possible parameter and scenario.

Howeverโ€ฆ

This approach can backfire, especially when collaborating with others or when revisiting the project at a later date. An overly complex configuration file can become hard to decipher, leading to errors and inefficiencies in the workflow.

#machinelearning #python #deeplearning #tech

Matt Hodges
1 month ago

This is a joyful read. "The original network trained for 3 days on a SUN-4/260 workstation. I ran my implementation on my MacBook Air (M1) CPU, which crunched through it in about 90 seconds (~3000X naive speedup)"

https://karpathy.github.io/2022/03/14/lecun1989/

#AI #DeepLearning #MachineLearning #ML #NeuralNetworks

Hereโ€™s a tool you should add to your AI toolbelt:

Auto-regressive models!

Theyโ€™re models that step through data and use context of the past to predict the future.

Here are three models you should have a look at!

๐•Ž๐•’๐•ง๐•–โ„•๐•–๐•ฅ ๐•—๐• ๐•ฃ ๐”ธ๐•ฆ๐••๐•š๐•  ๐ŸŽผ๐ŸŽถ๐ŸŽต

This model broke Twitter back in the day when it generated speech like no model before. Audio is hard at 16,000 samples per second (or more), so they used these interesting dilated convolutions that iterated over its own prediction to make the next sound as consistent as possible!

๐‘ท๐’Š๐’™๐’†๐’๐‘ช๐‘ต๐‘ต ๐’‡๐’๐’“ ๐‘ฐ๐’Ž๐’‚๐’ˆ๐’†๐’” ๐Ÿ–ผ

This model takes the auto-regressive idea to image generation. We know each pixel in an image has a relation to each other. So like an old-time electron-scanner TV, we are painting the image pixel by pixel, conditioned on the pixel before! Not as used today, but certainly an interesting lecture on our modelling history.

#machinelearning #datascience #python #deeplearning #career #tech

๐Ÿ’ป benchy: 1โญ

Improve the performance of your deep learning data loaders with benchy โฑ๏ธ๐Ÿ“Š

The prototype benchmarking dataloader by Romero Josh.

๐Ÿš€๐Ÿ” Use this tool to measure the speed of your data loading pipeline and identify potential bottlenecks, so you can optimize your models for maximum efficiency and speed.

Check it out here:https://github.com/romerojosh/benchy
My newsletter subscribers learned about this 5 months ago!
https://dramsch.net/newsletter

#Technology #Tech #MachineLearning #Python #DeepLearning #Career #LateToTheParty

gyptazy
1 month ago

This is real but not - and this could be your face!

Well, let's have a look at #AI and training models.

Public social media profiles offer a big possibility in scraping data. Thatโ€™s nothing new so far but letโ€™s have a look what everyone can do with these kind of data nowadays. In the past (and still ongoing) it was more regarding data theft in a kind of text like names, addresses, birthday, etc. Today, this also applies easily to photos. Guess, where you can find tons of public available pictures of people and especially of the same person - social media! What could possibly go wrong when posting hundreds or thousands of pictures of you, your wife or even your kids? Just by scraping the public available pictures from a social media platform you are able to train and create models for #AI (of anyone).

What might happen is shown right now. This random girl on social media platform posted around 1k pictures which were used for training a model. Afterwards I generated some pictures of her in France and sent her those. She replied that this is awesome but she has never been in France before and how this could be possible. We had some talks and afterwards she granted me to show this generated picture as an example. She also set her profile to private, now.

While this is just an example, this could be generated in every constellation in a kind of location, clothing (or even not!), other people, fore- & backgrounds.

Please, be aware of the data you post in public and keep your personal data safe!

#ai #ml #dl #ArtificialInteligence #MachineLearning #DeepLearning #stablediffusion #diffusionbee #model #generating #security #safedata #socialmedia #bigdata #data #DataScience

Learn the secrets of Data Science and Machine Learning with these must-read books! ๐Ÿ“š๐Ÿค–๐Ÿ’ก

From crafting great resume points to reviewing 70 years of machine learning in geoscience, there's something for everyone.

Don't miss out on the ML Recipes, Stable Diffusion Lookbook, and more!

Check it out here:https://dramsch.net/books

#MachineLearning #Career #DeepLearning #Data #Geoscience #ArtificialIntelligence #Technology #DataScience

Training neural networks can be a challenging task.

We have to learn about all these machine learning models, principles, evaluation, and adjust our Python coding style to how data scientists expect you to import numpy as np and everything else.

But with the right techniques and tools, you can improve the performance and efficiency of your models.

In this blog post, weโ€™ll touch on five intermediate tips that can help you enhance your neural network training using PyTorch and, frankly, make it easier. These tips cover various aspects, from data preprocessing to model architecture and optimization. Let's dive in!

DATA AUGMENTATION TECHNIQUES:
Data augmentation is a powerful technique that can significantly improve the performance of your neural networks.

When we apply random transformations to the training data, you can increase its diversity and reduce overfitting without even collecting more data!

...

#DeepLearning #MachineLearning #Pytorch

Find more here: https://dramsch.net/articles/5-intermediate-tipps-to-train-better-pytorch-models/

Qiusheng Wu
1 month ago

Segment-geospatial v0.9.1 is out. It now supports segmenting remote sensing imagery with the High-Quality Segment Anything Model (HQ-SAM)

Video: https://youtu.be/n-FZzKirE9I
Notebook: https://samgeo.gishub.org/examples/input_prompts_hq
GitHub: https://github.com/opengeos/segment-geospatial

#segmentanything #geospatial #deeplearning

Take your AI models to the next level with Stable Diffusion ๐Ÿš€๐Ÿค–

from Stability.AI

Access this powerful toolkit for improving the stability and robustness of your machine learning models, and ensure consistent and reliable performance on a range of tasks and environments, including real-world scenarios.

Check it out here:https://stability.ai/blog/stable-diffusion-public-release
My newsletter subscribers learned about this 11 months ago!
https://dramsch.net/newsletter

#Python #Data #Career #Ai #DeepLearning #LateToTheParty #MachineLearning

TT
1 month ago

@aral

As sb who holds an LL.M. (Master of Laws) in:

โถ #IntellectualProperty & Information #Technology Law

โท Substantive #EuropeanUnion Law (+ #EU #Competition)

&

โธ International Commercial Arbitration #Law (Alternative Dispute Resolution)

๐Ÿ“Œ I treat #LLM & all #AI hype โ€” #MachineLearning #LargeLanguageModels #DeepLearning #GenerativeAI #ArtificalIntelligence #LLMs โ€” the same way I DO #Cyber (#InfoSec #CyberSecurity, etc.) hype๐Ÿ”ดhttps://rb.gy/g5zcy #Tech #Risk #News #Politics #Securityโ–ผ

Hourly fractals
2 months ago

Hello Wednesday! This Mandelbrot Set is yours! #Math #DeepLearning

Alan Kotok
2 months ago

A system using deep learning algorithms to analyze ultrasound video images is shown to detect a difficult-to diagnose type of heart failure with clips from one view of the heart.

https://sciencebusiness.technewslit.com/?p=45055

#News #Press #Science #Business #HeartDisease #HeartFailure #Diagnostics #Ultrasound #Echocardiogram #ArtificialIntelligence #DeepLearning #Algorithms #University #UK #Entrepreneurs

Neuromatch
2 months ago

It's the first day of school ๐ŸŽ’ for #NMA2023! After yesterday's โœจopening ceremonies โœจ, today more than 2000 students have started their exciting journey. If you are not inspired enough, see the video to glimpse what awaits you!

Credits to Alireza Tanha for the amazing video!

#NeuromatchAcademy #Neuromatch #neuroscience #ComputationalNeuroscience #DeepLearning @neuromatch@a.gup.pe #neuromatchstodon

An introductory video to Neuromatch Academy's 2023 courses in the style of a movie trailer featuring various participants, organizers, and some course content
Stuart Gray
3 months ago

I've had a long-time interest in AI (I prefer the terms Machine Intelligence or Machine Learning) and Procedural Generation (ProcGen). Mostly the practical applications and implementation of them, rather than novel research, things like Large Language Models (LLMs), Natural Language Processing (NLP), Image Generation & Diffusion, Computer Vision, Object Recogntion, and Deep Learning.

[2/6]

#AI #MachineIntelligence #MachineLearning #ProcGen #LLMs #NLP #ImageGen #ComputerVision #DeepLearning

Today I had a thought ( a rare occurrence ๐Ÿคช ),
you know how you do not own anything anymore?!
Everything is subscription based, and is gone as soon as you can no longer afford it?!

AI might do the same with skills, in the future people might no longer be able to do things without AI that seem simple to us now. So you will no longer be able to access your skills without money, very dystopian if you think about it. ๐Ÿค”

#AI #KI #ChatGPT #ML #OpenAI #DeepLearning #LLM #LLMs

Nick Byrd
3 months ago

The 2nd #AI session talk at #SPP2023 was given by Sam Whitman McGrath and Jacob Russin: "Can Deep Learning Inform Explanation in Cognitive Science?"โ€”the slides also credited Roman Feiman, Ellie Pavlick, and Randy O-Reilly.

They take a middle ground view between two existing viewsโ€”an underrated move in academia?

You can follow this kind of work via Jacob's ResearchGate page: https://www.researchgate.net/profile/Jacob-Russin/research

I could not find a followable webpage for Sam.

#AI #philMind #cogSci #DeepLearning #LLM

Sam and Jacob introduce the talk.
Sam explains the middle-ground view.
Jacob explains the implications of the "past tense" case.
A picture from the Q&A.
Alan Kotok
3 months ago

Neuro Biomarkers, Home EEG Assessed in Upcoming Trial

A #digital #biomarkers company and developer of #neuroscience analytics are collaborating on a study of #EEG measurements and #algorithms with a #wearable device at home.

https://sciencebusiness.technewslit.com/?p=44890

#News #Science #Business #ClinicalTrial #MedicalDevice #Collaboration #Sleep #ArtificialIntelligence #DeepLearning #Neurology #Brain

Brain stimulation illustration
Neuromatch
4 months ago

Check your emails! Application results are being released ๐ŸŽ‰โ€‹ With 5000+ student and 550+ TA applications, #NeuromatchAcademy and #ClimatematchAcademy 2023 broke a new record.

โš ๏ธโ€‹Remember to accept the offer by June 7th!

Many thanks to the reviewing team, who worked days and nights to provide you with an answer as soon as possible.

Congratulations to all applicants who got accepted!! Unfortunately, we cannot accommodate everyone, but this should not stop you from continuing to learn #ComputationalNeuroscience, #DeepLearning, and #ClimateScience.Thank you for the effort you put in preparing your applications and good luck to the people admitted to the courses this July! We all look forward to starting this amazing adventure together!

Stay tuned for news and updates!

#neuromatch #NMA #CMA @climatematch #neuromatchstodon

Neuromatch
4 months ago

Would you like to mentor a group of amazing students from all over the world? The students at #NeuromatchAcademy need you this summer!

Itโ€™s only a 2-3 hour commitment over July 10-28.

Please sign up here for #ComputationalNeuroscience or #DeepLearning: https://airtable.com/shr7lhMw39FnKYCyN

#neuromatch @neuromatch@a.gup.pe #neuromatchstodon

Map of the World with states coloured according to the number of applicants received. The Neuromatch Academy name and logo on the upper right. Below that in the center is the text: "MENTORS NEEDED, Apply Now" and a bullet point list: "Computational Neuroscience and Deep Learning", "Influence the careers of future researchers", "Make a WORLD of difference to the students"
Grace Lindsay
4 months ago

I need a way as a non-hardcore mathematician to keep up with deep learning theory. Anyone know of good sources of review articles or other easier-to-digest summaries of findings?

#AI #MachineLearning #deeplearning

Alan Kotok
4 months ago

Rare Disease Search Adds Large Language Model A.I.

A company offering a search engine to help diagnose rare diseases is integrating #generative models from Open AI into its artificial intelligence #algorithms.

https://sciencebusiness.technewslit.com/?p=44792

#News #Science #Business #RareDiseases #OpenAI #ArtificialIntelligence #SearchEngine #DeepLearning #ChatGPT

Artificial intelligence, left-brain/right-brain, illustration
Alan Kotok
5 months ago

Computational Drug Discovery Company Acquires A.I. Businesses

Two companies using artificial intelligence to advance #biotechnology and design new #therapies are merging with a #computational #engineering drug #discovery enterprise.

https://sciencebusiness.technewslit.com/?p=44767

#News #Business #Science #Finance #Merger #Acquisition #Canada #Chemistry #ArtificialIntelligence #Algorithms #DeepLearning

Artificial intelligence, DaVinci Genesis-parody,  illustration

๐Ÿ“ฐ #AI #Godfather "Dr. Geoffrey Hinton has been inundated with requests to talk after quitting Google to warn about risk of digital intelligence."

Hinton believes digital brains might be about to supersede biological ones.

He's a socialist who says private ownership of media and "means of computation" is not good.

How long before for-profit "Big AI" labels him a "Commie" and succeeds in marginalizing Hinton as an anti-free speech Luddite?

#AI #DeepLearning https://bityl.co/IVnr

Skullvalanche
5 months ago

Hinton Today: https://arstechnica.com/information-technology/2023/05/warning-of-ais-danger-pioneer-geoffrey-hinton-quits-google-to-speak-freely/

Hinton a month ago: https://youtu.be/qpoRO378qRY

I find the "oh, I know them, they're good people" argument about A.I. researchers to be such a tired and garbage statement. It doesn't matter how *good* the scientists who worked on the atomic bomb were, they still ushered in the age of mutually assured destruction.

I'd like to reiterate that my concern about #AI is not that it's going to become conscious, or that it's going to displace a huge portion of the workforce (it will) but before it does *any* of that it's going to be leveraged by bad actors to flood society with truthy sounding bullshit the likes of which we've never seen and it will lead to killing in the streets.

QAnon is just a canary in the coal mine.

#DeepLearning #Google #LLM #GPT #ChatGPT #Microsoft

blake shaw
5 months ago

"I don't know if any of you have really seriously dived into #DeepLearning, but it is a truly massive and unwieldy zoo of extremely abstract, seemingly tailored concepts that appears almost impossible to get to the bottom of. There is certainly no foundational, theoretical treatment of the field that would allow for it to be approached with the sort of global understanding with which we treat programming languages... so that that is what I want to expose in the book, how can build up this tower of abstractions from nothing but the simplest functions, elucidating what has been obscured in the production process."

โ€” Anurag Mendhekar on his new book with Dan Friedman, the 800 page tour de force The Little Learner

#lisp #scheme #clojure #ai #ml