Federated learning, wherein training data never leaves the user's device (only gradients or model parameters), is an effective way to protect the privacy of individual training points.
MACHINE LEARNING
-
Sparse Deep Networks Generalize Better Than Dense Networks
By
–
Read one of the first theoretical analysis covering optimization, generalization and approximation in deep networks: https://
bit.ly/3VKflW2 The paper proves that sparse deep networks such as CNNs can generalize significantly better than dense networks. -
Responsible AI Systems Must Prioritize Values Over Efficiency
By
–
Building responsible #AI systems starts with recognizing that #technology solutions implicitly prioritize efficiency https://
sloanreview.mit.edu/article/when-a
lgorithms-rule-values-can-wither/
… @mitsmr #MachineLearning #100DaysofCode #IoT #serverless #Robots #DataScience #DeepLearning #Python #womenwhocode #BigData #Analytics -
Google Unveils Learning Interpretability Tool and Salience Evaluation Protocol
By
–
Today at 12:30pm the Google booth will host a demo about Google's open source Learning Interpretability Tool (LIT; https://
goo.gle/3kTIiLT) along with a newly introduced evaluation protocol (
https://
goo.gle/3gZSOWl) that illustrates the difference between various salience methods -
Neural Networks Learn Better With Rest Periods
By
–
Artificial #NeuralNetworks Learn Better When They Spend Time Not Learning at All http://
neurosciencenews.com/ai-ann-rest-le… @NeuroscienceNew #AI #MachineLearning #DataScience #Python #BigData #Analytics #100DaysofCode #IoT #serverless #womenwhocode #Robots #DigitalTransformation #DeepLearning -
AI Achieves Stunning Advance in Computer Code Writing
By
–
#AI learns to write computer code in ‘stunning’ advance
#RuleoftheRobots https://
science.org/content/articl
e/ai-learns-write-computer-code-stunning-advance#.Y5VC5wLum1g.twitter
… -
Evolution as Pre-training, Experience as Fine-tuning
By
–
Evolution is pre-training. Experience is fine-tuning.
-
ResNets in the 90s: Neural Nets with Direct Linear Connections
By
–
We had resnets in the 90s, in the form of neural nets with direct linear connections between input and output.
-
Chain-of-thought Jar Jar Binks posts linked by Twitter algorithm
By
–
I've never seen Twitter's timeline algorithm more excited than when, immediately after I posted about chain-of-thought prompting GPT‑3 in the style of Jar Jar Binks, it showed me @paulg
's tweet about the chain of thought that prompted Jar Jar Binks.