Thanks for the videos. They are very concise 🤩. Regarding the time series decomposition, pacf and acf, what should be done differently if the signals have high acf, seasonality and trend? Do you add additional features to capture this information?
@delali1900 for removing Trends you can do detrending using Detrending: Differencing, Log Transformation, Polynomial Detrending: Fit and remove polynomial trends if the trend is nonlinear. For Deseasonalizing, You can do Seasonal Decomposition and Seasonal Differencing: Subtract the value from the same season of the previous cycle (e.g., for monthly data, subtract the value from the same month last year). For Modelling choose either, LSTM or Temporal Fusion Transformers (TFT)
Dear Maryam, Thanks for sharing your knowledge, But there are some points which could help the videos better: - I have watched the andrej karapthy's videos, and one thing that really helped me understand the foundations and got me to stuck to his videos is that he tries to explain the backbone of the algorithms that he is trying to teach us al by examples and even he tries to write the code for them. - You really are making the point by your videos, But I think since you are trying to make short videos the consistency of the video decreases, For example the length of the Andrej's videos has been more than 1 hour but since I was really enjoying his lecture about disecting the building blocks of the neurons I was willing to watch it till the end. - Right now the videos about the generalized ideas in data science are really booming on youtube, But since You have all that experience and knowledge I think you can make a difference by really getting into the models and teach the audience about the backbone and details that are happening behind all those functions (This is the type of content and I think is missing write now and You can use it as your own advantage) - At the end I wish you a very good luck for your channel ;)
@@maryammiradi Well, Since you are mostly experienced in NLP I really would like to know more in depth about NLP techniques, how they are working underneath and also a deep dive in hugging face models and how should we use the different models ( IK you covered it before but I think there is so much more room for details andknow hows)
Very glad to hear! Would you care to explain what you mean by intermediate advanced level? Which type of things you would like to know about catboost and xlstm? Let me know.
@@maryammiradi here are few example video topics: 1. Catboost theory, and classification example with real world dataset. 2. Xlstm theory and example using it for time series analysis or NLP task. 3. Newer Vision transformer models such as LeViT, FasterViT, DaViT etc there are no videos at all on these on CZcams. 4. Another computer vision video i would love to see is "image quality assessment" for documents. If image is blank, dirty, blurred, poorly scanned etc, how to assess it's quality. 5. Crash course on the most promising boosting algorithms such as adaboost, randboost, xgboost etc. I like a bottom up approach with theory first and then code for such topics because the training code is really small.
First of all thank you for the great video ,21:00 can we use Autoencoder for feature extraction and fed UNet or any other DL model in Computer vision segmentation task ?
@RAHULR-fd4qm Yes, absolutely, After training Autoencoder, use the encoder part of the autoencoder to transform the input data into the latent space. These latent vectors serve as the feature representations of your input data.
🙋🏻♀️Get Access to my 20+ Years Experience in AI: ⚡️Free guide: www.maryammiradi.com/free-guide
⚡️AI Training: www.maryammiradi.com/training
Love this video, help me alot in implementing project
@@JustDONEItRN Very glad to hear!!
Your videos are Excellent Im eagerly waiting for the next one ! Thank you 🥰
Thank you so much and very glad to hear!!
Excellent series . excited for this week project thanks a lot Mam 😊
Most welcome 😊
Very informative! Thank you.
I am glad it was informative 👋
Thanks for the videos. They are very concise 🤩. Regarding the time series decomposition, pacf and acf, what should be done differently if the signals have high acf, seasonality and trend? Do you add additional features to capture this information?
@delali1900 for removing Trends you can do detrending using Detrending:
Differencing, Log Transformation,
Polynomial Detrending: Fit and remove polynomial trends if the trend is nonlinear. For Deseasonalizing,
You can do Seasonal Decomposition and
Seasonal Differencing: Subtract the value from the same season of the previous cycle (e.g., for monthly data, subtract the value from the same month last year). For Modelling choose either, LSTM or Temporal Fusion Transformers (TFT)
That is very useful!! Thanks!!
My pleasure and Glad it was useful 😄
Ma'am love from Nepal 🇳🇵
Keep continuing mam
Dear Maryam, Thanks for sharing your knowledge, But there are some points which could help the videos better:
- I have watched the andrej karapthy's videos, and one thing that really helped me understand the foundations and got me to stuck to his videos is that he tries to explain the backbone of the algorithms that he is trying to teach us al by examples and even he tries to write the code for them.
- You really are making the point by your videos, But I think since you are trying to make short videos the consistency of the video decreases, For example the length of the Andrej's videos has been more than 1 hour but since I was really enjoying his lecture about disecting the building blocks of the neurons I was willing to watch it till the end.
- Right now the videos about the generalized ideas in data science are really booming on youtube, But since You have all that experience and knowledge I think you can make a difference by really getting into the models and teach the audience about the backbone and details that are happening behind all those functions (This is the type of content and I think is missing write now and You can use it as your own advantage)
- At the end I wish you a very good luck for your channel ;)
Thank you for the suggestion. Which methodsdo you lije to know more about?
@@maryammiradi Well, Since you are mostly experienced in NLP I really would like to know more in depth about NLP techniques, how they are working underneath and also a deep dive in hugging face models and how should we use the different models ( IK you covered it before but I think there is so much more room for details andknow hows)
I am experienced in all different part of AI including NLP, I will do some deep dives. Thanks.
Love your content and style. Could you do some intermediate - advanced level guides covering models like catboost, xlstm etc
Very glad to hear! Would you care to explain what you mean by intermediate advanced level? Which type of things you would like to know about catboost and xlstm? Let me know.
@@maryammiradi here are few example video topics:
1. Catboost theory, and classification example with real world dataset.
2. Xlstm theory and example using it for time series analysis or NLP task.
3. Newer Vision transformer models such as LeViT, FasterViT, DaViT etc there are no videos at all on these on CZcams.
4. Another computer vision video i would love to see is "image quality assessment" for documents. If image is blank, dirty, blurred, poorly scanned etc, how to assess it's quality.
5. Crash course on the most promising boosting algorithms such as adaboost, randboost, xgboost etc. I like a bottom up approach with theory first and then code for such topics because the training code is really small.
@@ajkdrag sure Thanks alot for all the suggestions 👋 Will put them on my list but I always need to check if there is a demand broadly.
Waiting for part 2 with hyperpara tuning!!!
take away: keep it simple and do hyperparameter tuning
First of all thank you for the great video ,21:00 can we use Autoencoder for feature extraction and fed UNet or any other DL model in Computer vision segmentation task ?
@RAHULR-fd4qm Yes, absolutely, After training Autoencoder, use the encoder part of the autoencoder to transform the input data into the latent space. These latent vectors serve as the feature representations of your input data.
Data Science Guide
Of course! Please go to my website with this link: www.maryammiradi.com/free-guide
Could you get me the data-set please?
It is in the description. Please check it out there.
@@maryammiradi Found it now. Thanks a lot. Best regards, João.
@@joaobezerra2494 Glad to hear! Let me know if you have any other questions.