Files
2025-06-05 22:07:12 +02:00

39 lines
1.8 KiB
Markdown

-
- [[Lavori Ristrutturazione]]
- [[PAPERS/Software-ecosystem-book-chapter]]
- type:: [[meeting]]
external-links::
tags::
people:: #people/claudio #people/juri #people/riccardo #people/PhuongNguyen
presenter:: #people/claudio
date:: [[15-09-2022]]
duration-min:: 72
- ![image.png](../assets/image_1663255002804_0.png)
- ![image.png](../assets/image_1663255015769_0.png)
-
- ![image.png](../assets/image_1663255130726_0.png)
- Claudio has some ideas to employ Pytorch in our context to develop recsys
- ![image.png](../assets/image_1663255235458_0.png)
- ![image.png](../assets/image_1663255291339_0.png)
- It is necessary to have in the Claudio's thesis a section about AutoML
- ![image.png](../assets/image_1663255368192_0.png)
- ![image.png](../assets/image_1663255419524_0.png)
- Way to compose existing models
- ![image.png](../assets/image_1663255453686_0.png)
- Look at what Azure ML is doing concerning recsys for supporting ML pipelines [[Ideas]]
collapsed:: true
- There is also a respository we could you ([OpenML](https://www.openml.org/)OpenML I think)
- ![image.png](../assets/image_1663255676941_0.png)
- ![image.png](../assets/image_1663255743343_0.png)
- Dataset of queries and they classified the intent of the user ie navigational, transactional, and informational
- Use this maybe for user-feedback [[Ideas]]
- ![image.png](../assets/image_1663255899417_0.png)
- ![image.png](../assets/image_1663255935871_0.png)
- [[Ideas]]
- ![image.png](../assets/image_1663256047826_0.png)
- [[Ideas]]
- #Pytorch is provided by Facebook
- It seems to be faster and also because it is becoming open source
- #Tensorflow is from Google
- Use Transformer (BERT) to extend NEMO
- Or use NLP techniques to even recommend values for recommended operations/actions