it’s been some time since the last volume, but even though the pause was long, DSB is back. Back again! So at first you might test your ability to randomize in Analytical. The huge topic of this volume is in Graphs and Visualizations about pictures generated by AI, it’s an amazing experience to try it by yourself. And before you go to sleep, give a shot to mCoding YouTube channel in Video & Podcast.
And as always, enjoy your reading.
https://www.expunctis.com/2019/03/07/Not-so-random.html – Old but gold. This simple game will show you that you are not able to be random (or at least your fingers aren’t).
https://netflixtechblog.medium.com/machine-learning-for-fraud-detection-in-streaming-services-b0b4ef3be3f6 – How Netflix detects fraud and abuse on its streaming service with help of ML. More technical details are in paper here.
https://blog.paperspace.com/building-a-checkers-gaming-agent-using-neural-networks-and-reinforcement-learning/ – Learn how to implement a Deep Q-Learning to create an AI agent capable of playing checkers.
Computer Science & Science
https://www.joeantonakakis.com/posts/kmp-algorithm/ – KMP (Knuth-Morris-Pratt) string search algorithm explained with code in Python.
https://spectrum.ieee.org/floating-point-numbers-posits-processor – Posit-based processors could make training of large models like GPT-3 much more effective. Posits are basically hardware-friendly numbers.
https://www.theverge.com/2022/9/28/23375691/google-search-multisearch-visual-keywords – After two decades Google is trying to come up with a new type of search being more natural and visual to compete with TikTok or Instagram.
Graphs and Visualizations
https://openai.com/blog/dall-e-now-available-without-waitlist/ – DALLE is now available to the public, so sign up and generate your own unique pictures. It’s fun, it’s interesting and you can even create something beautiful. But if you want to go more deeply into the creating process of AI content generation, then read this guide, even though it’s written mainly for Stable Diffusion, which was also publicly released a few days ago. And one more thing, intro to Stable Diffusion in KerasCV is here.
https://nightingaledvs.com/how-to-visualize-a-graph-with-a-million-nodes/ – How to deal with big network visualizations with Cosmograph.
Business and Career
https://www.finextra.com/newsarticle/41012/samsung-launches-credit-card-in-india – Samsung is entering the world of finance in India with a credit card.
https://www.coinspeaker.com/digital-euro-green-european-nations/ – European digital currency is being discussed and maybe in the near future there will be a virtual euro.
https://thepaypers.com/payments-general/amazon-partners-lendistry-for-the-amazon-community-lending-programme–1258595 – Amazon in US has loaned 35 million USD to over 800 sellers and wants to more focus on small business with help of Lendistry.
https://www.hokej.cz/hokej-v-cislech-prehled-predikci-modelu-na-startu-sezony/5067877 – Hokej.cz představil pro nadcházející sezonu extraligy predikci sezon pomocí svého Game score. Každý hráč podle svých statistik za poslední 2 roky dostane své vlastní „hráčské“ game score (počet očekávaných gólů, kvalita bránění atd.) a každý tým má předpokládanou Game score sílu složenou ze svých hráčů. Model nasimuloval 50 000 odehraných sezon: mistr bude z Pardubic, Kladno půjde zase do baráže. (rcmd by reader)
https://www.politico.com/news/2022/08/15/artificial-intelligence-health-care-00051828 – While the investments into AI in healthcare are growing, the level of usage is poor. The reasons are for example inability to explain why part of the algorithm results, government regulations, differences between health systems and a lack of quality data.
https://garymarcus.substack.com/p/the-new-sport-of-misrepresenting – Gary Marcus is defending criticism of AI and explains why it’s justified and important, and should not be misinterpreted or ridiculed.
https://towardsdatascience.com/uncertainty-in-deep-learning-brief-introduction-1f9a5de3ae04 – 6 parts series of articles about uncertainty in deep learning – from maximum likelikhood estimation to transformers.
https://pyimagesearch.com/2022/09/05/a-deep-dive-into-transformers-with-tensorflow-and-keras-part-1/ – Comprehensive multiple parts intro to transformers with TensorFlow and Keras.
Datasets & Libraries
https://www.assemblyai.com/blog/how-to-run-openais-whisper-speech-recognition-model/ – Install and run OpenAI’s Whisper which is a speech recognition model for Python.
https://github.com/modin-project/modin#readme – Modin introduces multi-threading into pandas. Now even large datasets can be handled effectively and quickly.
https://github.com/axa-group/Parsr#readme – Parsr is a minimal-footprint document (image, pdf, docx, eml) cleaning, parsing and extraction tool.
https://blog.wolt.com/engineering/2022/08/11/project-template-for-modern-python-packages/ – Description of template for modern Python package designed by Wolt, includes cookiecutter template.
https://towardsdatascience.com/machine-learning-systems-versus-machine-learning-models-3955d038ea1f – ML model is just a small part of a whole complex ML system. And you need that system if you want to deliver value to a customer.
https://medium.com/@Not4j/what-we-are-missing-in-data-ci-cd-pipelines-c3d7f02e0894 – What is missing in current CI/D tools for data pipelines?
Video & Podcast
https://www.youtube.com/c/mCodingWithJamesMurphy – Have you watched all the videos on ArjanCodes? Then definitely try this YouTube channel about Python created by James Murphy. Especially useful for advanced users. (rcmd by reader)
https://youtu.be/VMj-3S1tku0 – Andrej Karpathy left Tesla, and at least for now he is creating introductory ML tutorials. And it’s amazing! For now he has released three comprehensive videos where you can see the code and he goes through and explains every step.
https://podcasts.google.com/feed/aHR0cHM6Ly9sZXhmcmlkbWFuLmNvbS9mZWVkL3BvZGNhc3Qv/episode/aHR0cHM6Ly9sZXhmcmlkbWFuLmNvbS8_cD00MTQ5 – Lex Fridman Podcast with Michael I. Jordan. It’s an old episode from 2020, but the talk is still very inspirative. (rcmd by reader)
Papers & Books
https://arxiv.org/abs/2002.10990 – Inverse reinforcement learning approach (GIRL) to goal based wealth management problems. (rcmd by reader)
Behind the Fence
https://www.crunchboard.com/jobs/154144748-senior-data-engineer-at-rural-innovation-strategies-inc-risi – Senior Data Engineer in Rural Innovation Strategies, Inc. (RISI), Baltimore, USA.