I usually read the experimental details and then straight up dive into the code.
Pick it apart till I’m down to bare metal. Create dummy tensors and then play around with individual modules to understand what is happening. Colabs are helpful here too, to isolate the code.
MACHINE LEARNING
-
Learning ML by Dissecting Code and Dummy Tensors
By
–
-
ResiDual Solves Gradient Vanishing and Representation Collapse
By
–
Optimizing Transformers: Microsoft & RUC’s ResiDual Solves Gradient Vanishing and Representation Collapse Issues
-
Flan-T5 Performance Analysis: Efficiency Across Language Models
By
–
Pretty cool idea! Great to see Flan-T5 (despite being the smallest model here) hold it's ground pretty well . It even outperforms other LMs like Dolly or StableLM. Also another noteworthy point is that at "compute-match", Flan-T5 3B is equivalent to the cost of a 1.5B
-
Reinforcement Learning Algorithm Enables Outdoor Robotic Navigation
By
–
Presenting an algorithm that uses #ReinforcementLearning to train a navigation policy in simulated indoor environments and then transfers that same policy to real outdoor settings in order to enable efficient long-range robotic navigation outdoors → https://t.co/Gome7XTHXb pic.twitter.com/AqMMBZZwIx
— Google AI (@GoogleAI) 3 mai 2023Presenting an algorithm that uses #ReinforcementLearning to train a navigation policy in simulated indoor environments and then transfers that same policy to real outdoor settings in order to enable efficient long-range robotic navigation outdoors → https://
goo.gle/3VvV9Ii -
Teaching Differential Privacy Mini Course in Bangalore
By
–
So excited to be teaching a mini course on differential privacy in Bangalore! It's been 11 years since I've been to India, looking forward to being back.
-
SAS Innovate 2023: Learn Analytics and AI, Register Now
By
–
At #SASInnovate 2023 (a complimentary event), you’ll have opportunities to learn, be inspired by, and guide the future of Data #Analytics and #AI. Register now and build your Agenda here: https://
sas.com/gms/redirect.j
sp?detail=PLN2755_1799220624
… by @SASsoftware ———
#DataScience #MachineLearning #ML #SASVisionary -
MLOps Environment: 30 Requirements Simplified by Abacus AI
By
–
Building #MachineLearning Systems is hard. Here are 30 requirements for an #MLOps environment. @abacusai handles all that for you. You bring the data and the use case. They deliver the #ML environment: https://
abacus.ai/mlops
———
#BigData #DataScience #AI #DataScientists #ML -
Task Emergent for Different Metrics Than Initially Measured
By
–
sure, the task would not be emergent *for that metric*, but its still emergent *for the metric we care about*
-
Discussion on Different Metrics in Previous Research Paper
By
–
Finally, some discussion around using different metrics was given in our previous paper, which could be worth taking a look at: https://
openreview.net/pdf?id=yzkSU5z
dwD
… -
Chain-of-Thought Prompting: Emergence in Large Language Models
By
–
Additional response 2: Another popular example of emergence which also underscores qualitative changes in the model is chain-of-thought prompting, for which performance is worse than answering directly for small models, but much better than answering directly for large models.