@@deeplearningexplained I love how you're covering the basics with a repeatable framework. I am trying to learn the end to end workflow of designing and deploying ML projects and experiments on the job for both research and industry purposes (I work at a startup and as a deep learning research assistant with just a swe background as a new college grad) in the future, topics like 1. Data - how to analyze a dataset for deep learning tasks , particularly image datasets (given a source of data , how do I decide how much data I need based on the goal at hand (say image classification). I don’t need specifics here because that’d be too in depth but just a general process to follow (and the right questions to ask) 2. How to choose a deep learning model to fit the dataset in hand. similar to the last one. Given a dataset , and several sources of literature for a task, how do I design an approach to combine/choose the best adaptations to proceed. What are the best questions to ask that when answered would yield favorable approaches. 3. Selecting experiments - (how to design deep learning experiments on smaller chunks of data , and how to evaluate if the idea is working out before moving on to more expensive training on massive datasets) 4. Evaluation - what plots and eval metrics can I use while running smaller experiments to know what adjustments to make Im looking for something very similar to your current video format - something like a checklist of things to do while im doing these tasks that I can apply to any general use case.
very good! I'm kind of winging my way in an undergraduate level research internship in canada (I'm from brazil) and I'm sure this is gonna be of great use
@@deeplearningexplainedboth of them, my problem is that I can’t make the link between the article and the implementation, how can SSM segment an image? Usually it is convolution that we use for feature extraction! for the code also I can't understand how it works, it is difficult for me you know😢 I have another question, what is local features, and what is global features ?
Thanks for the method, I'm curious about your use of colab to read the code. Don't you find using vscode and ot functionalities of jumping through code more convenient especially the use of the debugger to understand the flow of the program? Or is the process I'm describing more detailled than the understanding you are describing or just time consuming?
Hey there, awesome question! I do use VScode or Pycharm debugger when I’m dealing with debugging a program that wrap around a model or pipeline that is running on an analysis I built. However, when I’m trying to understand how a model is running algorithmically I found that there isn’t a better way than taking a pen and paper and going through the architecture by hand. The issue is that what I usually don’t understand isn’t in the infra code, it’s in the implementation of the algorithm that was stipulated in the paper. When I clear the algorithm part, the rest is usually a breeze.
You might be my favorite creator at this point
Thanks for the kind words man, if you have any requests for a tutorial topic let me know and I'll prioritize it.
@@deeplearningexplained I love how you're covering the basics with a repeatable framework. I am trying to learn the end to end workflow of designing and deploying ML projects and experiments on the job for both research and industry purposes (I work at a startup and as a deep learning research assistant with just a swe background as a new college grad)
in the future, topics like
1. Data - how to analyze a dataset for deep learning tasks , particularly image datasets (given a source of data , how do I decide how much data I need based on the goal at hand (say image classification). I don’t need specifics here because that’d be too in depth but just a general process to follow (and the right questions to ask)
2. How to choose a deep learning model to fit the dataset in hand. similar to the last one. Given a dataset , and several sources of literature for a task, how do I design an approach to combine/choose the best adaptations to proceed. What are the best questions to ask that when answered would yield favorable approaches.
3. Selecting experiments - (how to design deep learning experiments on smaller chunks of data , and how to evaluate if the idea is working out before moving on to more expensive training on massive datasets)
4. Evaluation - what plots and eval metrics can I use while running smaller experiments to know what adjustments to make
Im looking for something very similar to your current video format - something like a checklist of things to do while im doing these tasks that I can apply to any general use case.
Absolutely awesome list of topic, will get these prioritized 🫡
very good! I'm kind of winging my way in an undergraduate level research internship in canada (I'm from brazil) and I'm sure this is gonna be of great use
Heck yeah man this is so cool, I did the same when I was at McGill. Which lab are you part of?
@@deeplearningexplained I'm working at the PRIMe (Perception, Robotics and Intelligent Machines) lab at Université de Moncton, New Brunswick!
very good 5 ⭐
🤩 thanks!
Great tips, thank you!
Glad it was helpful! 🌹
I was looking for that! Thank you
You're welcome! 🌹
@@deeplearningexplained would you like to make more vidéos like that please, because I’ m novice in this field
What do you think about Mamba and vision Mamba, would you explain how can I reproduce it please? Help me please
Hey for sure, are you trying to understand the theory or the implementation?
@@deeplearningexplainedboth of them, my problem is that I can’t make the link between the article and the implementation,
how can SSM segment an image? Usually it is convolution that we use for feature extraction!
for the code also I can't understand how it works, it is difficult for me you know😢
I have another question, what is local features, and what is global features ?
Thanks ❤
Glad it was useful!
Really helpful
Glad to hear it!
hey brother thank you so much, you have amazing content, Ukraine greets you!!
Glad you enjoy it! 🌹
Hope everything is OK on your side, stay strong!
Thanks for the method, I'm curious about your use of colab to read the code. Don't you find using vscode and ot functionalities of jumping through code more convenient especially the use of the debugger to understand the flow of the program? Or is the process I'm describing more detailled than the understanding you are describing or just time consuming?
Hey there, awesome question!
I do use VScode or Pycharm debugger when I’m dealing with debugging a program that wrap around a model or pipeline that is running on an analysis I built.
However, when I’m trying to understand how a model is running algorithmically I found that there isn’t a better way than taking a pen and paper and going through the architecture by hand.
The issue is that what I usually don’t understand isn’t in the infra code, it’s in the implementation of the algorithm that was stipulated in the paper.
When I clear the algorithm part, the rest is usually a breeze.
Are you active on X ?
Not really no, a bit more on LinkedIn.
I mainly go on X to check out the latest research!
thats the neat thing, you don't
😂
Bon travail
Merci beaucoup! 🌹