Have watched almost every video on your channel. Love your content. It's one of the best courses ever. Could you please make courses on Modern Natural Language Processing and Computer Vision using PyTorch and Tensorflow showing how to implement the architectures of SOTA models?
Excellent. Thank u :) to create y_leftright, it is better to use numpy built-in functions and avoid looping in arrays for performance reasons in case dataset are big. Example : y_leftright = (y_train > 5) * 1
There is no one who comes close to your explaining and originality. I am blown away by sheer clearance and hardwork being put into this. What do you actually do man?
Amazing tutorial! Thank you very much for developing and providing such a high-quality and so inspiring tutorial series! I have just one question that I could not understand very well. How does the backpropagation work in this MNIST example where we have two different losses? How can the shared weights be updated in this case? Also, would be possible to have one output for classification and another one for regression with shared weights in this multi-branch architecture?
At 07:49, where does the new_model come from? we had created it when we created a sequential model but you seem to be using it in functional api as well. Or do you mean to say that we can create a functional model using Sequential as a starting point. The new_model variable creates a confusion for me.
Thank you so much for all of these. Your series on coding algorithms only using numpy and python really helped me understand what the algorithms are doing. I really really appreciate you. I've been learning data science/machine learning from online materials over the past couple of months, but I was wondering how you learned it? Did you teach yourself or do something along the lines of going to university for computer science? If you did teach yourself or found other materials outside of your university coursework extremely helpful, is there anywhere you can't point me to so that way I can build the strongest foundation possible to work upwards? I know numpy, pandas, and matplotlib pretty well, and some sklearn and tensorflow, but I think I need to somehow (and I know this would be difficult to teach) learn more of the logic followed when people like yourself are coding things like these
Hi, thanks so much! I totally understand that the path to learn ML can be a bit difficult and overwhelming with all those topics. I got a strong theoretic foundation in university, and picked up the Python skills in my first job. Other then that I have some useful resources here: www.python-engineer.com/posts/ml-study-guide/. I myself learned pandas, numpy, matplotlib etc.with YT videos, and I also did Andrew NG's ML course, fast.ai course, Standford lecture on DL, and I have the book Hands-on Machine Learning with Scikit-Learn...
Hello, Thank you so much for this video. It is really useful and simple to learn. I want to ask one question for the example of Functional API: For example Can I use functional API for predict the house price and city name? I think this is example of one input and two output. Am I right?
hello author, i want to train a model to predict heatmap (mean square error loss) and binary segmentation (binary cross entropy loss). i tried to train model using multi branch (2 branch duplicates for 2 output). but the the final output will favour for only one type of output. For example when i train using model.fit with equal loss weights, the output is good for heatmap, but binary mask output is wrong and gives pixels 1 for the regions similar to heatmaps. And when i train using gradtape loop, the output is good for segmentation mask, but heatmaps are wrong and looks like masks. how can i solve this, please give me your suggestion. thank you
This is complety irrelevant to the topic of this video, but: model.summary() prints the summary of the model and returns None. So print(model.summary()) prints a None after the model summary.
A real waste if time, just to see a single use case. I thought all the different use cases of the functional api has drawn on the IDE will be discussed but no, and I came juste for that 😢
Have watched almost every video on your channel. Love your content. It's one of the best courses ever. Could you please make courses on Modern Natural Language Processing and Computer Vision using PyTorch and Tensorflow showing how to implement the architectures of SOTA models?
Great video. Very clear and easy to follow.
thanks!
Your videos are amazing they help me alot
حلابة
Thanks :)
Excellent. Thank u :)
to create y_leftright, it is better to use numpy built-in functions and avoid looping in arrays for performance reasons in case dataset are big. Example :
y_leftright = (y_train > 5) * 1
You’re absolutely right, thanks for the hint!
There is no one who comes close to your explaining and originality. I am blown away by sheer clearance and hardwork being put into this. What do you actually do man?
Thanks so much! Really glad to hear that
Amazing explanation
Amazing tutorial! Thank you very much for developing and providing such a high-quality and so inspiring tutorial series! I have just one question that I could not understand very well. How does the backpropagation work in this MNIST example where we have two different losses? How can the shared weights be updated in this case? Also, would be possible to have one output for classification and another one for regression with shared weights in this multi-branch architecture?
At 07:49, where does the new_model come from? we had created it when we created a sequential model but you seem to be using it in functional api as well. Or do you mean to say that we can create a functional model using Sequential as a starting point. The new_model variable creates a confusion for me.
Superb!
Thanks a lot!
can we write this whole code in a class that is inherited from tf.keras.model, like we do in pytorch? If we can, please make a video on it. Please
Thank you so much for all of these. Your series on coding algorithms only using numpy and python really helped me understand what the algorithms are doing. I really really appreciate you.
I've been learning data science/machine learning from online materials over the past couple of months, but I was wondering how you learned it? Did you teach yourself or do something along the lines of going to university for computer science? If you did teach yourself or found other materials outside of your university coursework extremely helpful, is there anywhere you can't point me to so that way I can build the strongest foundation possible to work upwards? I know numpy, pandas, and matplotlib pretty well, and some sklearn and tensorflow, but I think I need to somehow (and I know this would be difficult to teach) learn more of the logic followed when people like yourself are coding things like these
Hi, thanks so much! I totally understand that the path to learn ML can be a bit difficult and overwhelming with all those topics. I got a strong theoretic foundation in university, and picked up the Python skills in my first job. Other then that I have some useful resources here: www.python-engineer.com/posts/ml-study-guide/. I myself learned pandas, numpy, matplotlib etc.with YT videos, and I also did Andrew NG's ML course, fast.ai course, Standford lecture on DL, and I have the book Hands-on Machine Learning with Scikit-Learn...
@@patloeber Thank you for all of this!
Great!!
Hello, Thank you so much for this video. It is really useful and simple to learn. I want to ask one question for the example of Functional API: For example Can I use functional API for predict the house price and city name?
I think this is example of one input and two output. Am I right?
why is that dense1 layer just hanging ....where it connected to the output layer? plz help me
hello author,
i want to train a model to predict heatmap (mean square error loss) and binary segmentation (binary cross entropy loss).
i tried to train model using multi branch (2 branch duplicates for 2 output). but the the final output will favour for only one type of output.
For example when i train using model.fit with equal loss weights, the output is good for heatmap, but binary mask output is wrong and gives pixels 1 for the regions similar to heatmaps.
And when i train using gradtape loop, the output is good for segmentation mask, but heatmaps are wrong and looks like masks.
how can i solve this, please give me your suggestion.
thank you
Can you add auto subtitle for this?
This is complety irrelevant to the topic of this video, but:
model.summary() prints the summary of the model and returns None.
So print(model.summary()) prints a None after the model summary.
Are you german? Great videos btw
Yes I am. It's hard to hide my accent ;)
subtitles please
A real waste if time, just to see a single use case. I thought all the different use cases of the functional api has drawn on the IDE will be discussed but no, and I came juste for that 😢
kauon ke beech me koyal ko chup rahna padta hai
thank you