Global AI and Data Science

 View Only

Data Science Community News | Volume 1, Issue 2

By Christina Howell posted Mon September 09, 2019 02:24 PM

  
IBM Data Science Community Newsletter
September 2019 | Volume 1, Issue 2


Spotlight

spacer.gif
ds_newsletter-feature.jpg
Facial Recognition Tech is Growing Stronger, Thanks to Your Face
Should we start using tools that add adversarial noise to images of ourselves that we publicly post to reduce their efficacy on training or cause mis-classification, or will we see a time in which people wear clothing and accessories like glasses that distort our identity to an algorithm?

AI Skills

AI Explainability Tutorials
Recently, AI researchers from IBM open sourced AI Explainability 360, a new toolkit of state-of-the-art algorithms that support the interpretability and explainability of machine learning models. How does it match up to other recent efforts in the community on AI Explainability? Check out the AIX360 tutorials and share your thoughts. Read more

NEW IBM Professional Certificates on Coursera
IBM and Coursera have partnered to deliver Data Science and AI Professional Certificates. Is online training the way for professional development? Read more


Tools & Libraries

New SOTA Optimizer "Rectified ADAM" Shows Immediate Improvements for Model Training
Does the latest adaptive learning rate optimization method "Rectified" ADAM (RAdam) really show improvement over traditional ADAM optimization? Does RAdam solve the generalization issues that have led researchers and ML scientists to return to more traditional SGD with momentum optimization methods? This article conducts a hands on investigation. Read more


Useful Resources for Data Science
Paco Nathan asked his friends what the best resources for learning data science can be found. What do you think? Did he leave anything out? Read more


TF Explain
Despite advances in our ability to solve new problems with deep neural nets, these models are consistently challenged by questions of interpretability. Can TF explain give ML Engineers and Scientists the long sought after tools needed to fight this challenge?Read more



Solutions & Products

Diabetes Prediction with Ensemble Techniques
Why does combining multiple models (ensembeling) lead to better performance than if they are run individually? A practical example bsed on Diabetes Prediction by Neeraj Jangid. Read more


Research

RoBERTa: a Robustly Optimized BERT Pretraining Approach
What we're reading in the IBM Data Science Community:
Google's BERT broke all the NLP records late last year, but is very hard to train independently. Does RoBERTa mean that you can use this power to train for custom use cases without Google resources? Read more


Events

See all upcoming community events, here.


#In-the-know-feature
#GlobalAIandDataScience
#GlobalDataScience
#In-the-know
0 comments
43 views

Permalink