If you found this video helpful, then hit the *_like_* button👍, and don't forget to *_subscribe_* ▶ to my channel as I upload a new Machine Learning Tutorial every week.
Best channel to learn Deep Learning. Nothing can beat this.Just love it. Can you please complete the series on RNN and create something on LLM as your explanation is top notch. Thanks in advance
Excellent tutorial, was able to run both tanh (86.86% accuracy) and relu activations. It is great experience to test hand made code for neural networks. Your mathematical derivations helped lot too
Good job. But dz1 may have been accidentally scaled by 1/m in the backpropagation subroutine, leading to error observed by some users. Could you please confirm that?
wasy bhai ak bat khao smjhaya boht zabardast ha .brilliant explanation,concepts and coding.agr hindi ma smjaty to cha jaty india ma shardaa didi sa b agy 👍👍🏾
Thanks for a video, now I finally start to understand how a neural networks work. As for me, ReLu works better, 78% based on Tangent, and 86% based on ReLu
@CodingLane Bro You are awesome. Became a big fan of your work. I am saying this because I can understand these concepts even without learning the basics of ML and Deep learning.
@@MachineLearningWithJay Great Videos!!!...Please give some guidance on changing the number of hidden layers and neurons in each layer easily using code
No need. This videos is just to help you understand what goes under the hood. You can also use libraries like sklearn and framework like tensorflow/pytorch, which provides built in methods and models. You can directly use them.
As you said in previous video while deriving for softmax function with multiclass. It will give same results in vector form as binary classification so can we use dw1,dw2,dw3 vector formation of binary classification for multiclass classification with more than 1 hidden layer?
bro why did you use argmax row wise(axis=0) for a_out? . what i'm thinking is a2 is (n2,m) and has each column representing one hot encoding. if thats true then axis=0 would find the max probability out of 1st row having hot encodings from different examples.
Thank you for the link sent. I have already tried the code for my dataset. I am getting the learning rate correctly, but the accuracy is not good. I am not able to identify where am going wrong. They have done for Binary, I tried for Multiclass. But not getting good accuracy.
You may want to check dz1 in backpropagation subroutine; dz1 may have been accidentally scaled by 1/m in the backpropagation subroutine, leading to error observed by some users. Could you please confirm that?
thank you for your explanations! Can I implement dogs&cats datasets from scratch or can you suggest another dataset for practice, because I want to practice these algorithms on different datasets. Again, thank you!
You can use dog&cat dataset… I have made a video on it… though in that I am not implementing from scratch, but if ypu want, you can use that dataset. Also you may find different datasets from kaggle
Thanks for the amazing videos that you upload on the coding lane channel on youtubeI had doubts regarding the elementwise multiplication at 15:00 timestamp in your Derivation of backpropagation in neural networks video. Could you please explain that point to me?
@@MachineLearningWithJay Thank u for ur immediate reply...I have got stuck up in N layer deep network....It would be of great help if you can take me out of this issue.
@@sunithap915 I won't have the time to take you out of the issue personally. But I have an article that you can refer. It has code and explaination for N layer Neural Network Here is its link: github.com/enggen/Deep-Learning-Coursera/blob/master/Neural%20Networks%20and%20Deep%20Learning/Building%20your%20Deep%20Neural%20Network%20-%20Step%20by%20Step.ipynb I hope this can help for now. Also you can search online, there may be some other articles as well on this topic.
Hello, your video is very good, thank you, I have a "small" question, suppose I have a data set with a combination of 1000 functions, let's say Gaussian functions and Lorentzian functions (which are similar functions), the data set would be something like this [[[x1,x2,x3,x4,..,x70],[y1,y2,y3,y4,..,y70],[gaussian]], [[x1',x2',x3',x4',.. ,x70'],[y1',y2',y3',y4',..,y70'],[gaussian]] , [[x1'',x2'',x3'',x4'',.. ,x70''],[y1'',y2'',y3'',y4'',..,y70''],[Lorentzian]] , ....... -1000 functions ] , the idea is that the neural network learns to classify between one function or another depending on the entry points, seeing your code more or less as it could start from your development, perhaps the biggest problem is how do I compute the input of these pairs of data from each function in training.
bro, isn't it your derivative_tanh(x) function is wrong?? I think it might be, def derivative_tanh(x): return (1 - np.power(tanh(x), 2)) please do correct me!!!!!!
If you found this video helpful, then hit the *_like_* button👍, and don't forget to *_subscribe_* ▶ to my channel as I upload a new Machine Learning Tutorial every week.
Best channel to learn Deep Learning. Nothing can beat this.Just love it. Can you please complete the series on RNN and create something on LLM as your explanation is top notch. Thanks in advance
Excellent tutorial, was able to run both tanh (86.86% accuracy) and relu activations. It is great experience to test hand made code for neural networks. Your mathematical derivations helped lot too
Great work!! And gald the video was helpful
All the videos are simple and easily understandable. Only from these videos I could understand about NN.
Glad these are helpful!
Many Thanks! These are so great. Excited to build the network now.
@@vont1360 great! All the best!
You are too good brother no one can't give this types of easy explanation
Haha… Thank you very much 😇
At 14:40 why are we multiplying with (1/m), even though it is not mentioned in the formula?
Thank you so much for this video!! I've just coded "my" first ever CNN with your help and I can't wait to learn more about it!!!
Superb videos bro. Make more videos like this I will share your videos with my friends.
Thank you so much. Your sharing will be of great help to me !
Amazing and super helpful. I am learning a lot from your videos.
Thank you so much Malik ! I am very happy that these videos helped you.
very useful tutorial
@@darshananghan1555 thank you so much!!
Thank you so much for posting these videos! I learned a lot. Thank you once again! Your videos are awesome! Please keep updating it!
Sure! Glad to help 😀
Good job. But dz1 may have been accidentally scaled by 1/m in the backpropagation subroutine, leading to error observed by some users. Could you please confirm that?
wasy bhai ak bat khao smjhaya boht zabardast ha .brilliant explanation,concepts and coding.agr hindi ma smjaty to cha jaty india ma shardaa didi sa b agy 👍👍🏾
Hi, thank you!
Thanks for a video, now I finally start to understand how a neural networks work.
As for me, ReLu works better, 78% based on Tangent, and 86% based on ReLu
Bro this is awesome, make more implementations like these👍👍🎉🎉
Thank you! Sure, I will keep uploading more implementation videos.
Very helpful, thank you so much Jai
Welcome!
@CodingLane Bro You are awesome. Became a big fan of your work. I am saying this because I can understand these concepts even without learning the basics of ML and Deep learning.
Hi, Means a lot to me! Thanks!
argmax(a2,0) --> represents column vector
The assignment link is not working. And all the text below that are crossed out in desc. box.
Hi , if possible can you do a video on the model were we can change the number of hidden layers within the script?
Okay... I will upload it soon
@@MachineLearningWithJay Great Videos!!!...Please give some guidance on changing the number of hidden layers and neurons in each layer easily using code
Isn't axis=1 row wise summation ?
Hi...can you help to create h5 dataset of images for multiclass classification
Is it necessary to implement back propagation and and sift max by our own
No need. This videos is just to help you understand what goes under the hood. You can also use libraries like sklearn and framework like tensorflow/pytorch, which provides built in methods and models. You can directly use them.
How can we directly run on images rather than csv file?
really helpful, I wish can subscribe to you 10 times
Thank you so much! Very creative comment 😊
Great explanation as always
Thank you so much 😇!
How do we create that dataset?
As you said in previous video while deriving for softmax function with multiclass. It will give same results in vector form as binary classification so can we use dw1,dw2,dw3 vector formation of binary classification for multiclass classification with more than 1 hidden layer?
bro why did you use argmax row wise(axis=0) for a_out? . what i'm thinking is a2 is (n2,m) and has each column representing one hot encoding. if thats true then axis=0 would find the max probability out of 1st row having hot encodings from different examples.
Thank you for the link sent. I have already tried the code for my dataset. I am getting the learning rate correctly, but the accuracy is not good. I am not able to identify where am going wrong. They have done for Binary, I tried for Multiclass. But not getting good accuracy.
You may want to check dz1 in backpropagation subroutine; dz1 may have been accidentally scaled by 1/m in the backpropagation subroutine, leading to error observed by some users. Could you please confirm that?
As a Muslim i give you dua for each good work ."ALLAH increases your knowledge 100 times"
Hi, Salam. And thank you so much!
thank you for your explanations!
Can I implement dogs&cats datasets from scratch or can you suggest another dataset for practice, because I want to practice these algorithms on different datasets.
Again, thank you!
You can use dog&cat dataset… I have made a video on it… though in that I am not implementing from scratch, but if ypu want, you can use that dataset.
Also you may find different datasets from kaggle
@@MachineLearningWithJay i have implemented iris dataset that I took from kaggae and it worked well.
@@abdulkadercerkezi1448 thats great!
anyone knows how to import the dataset in google colab its showing error for me
@codingLane plese reply
Thanks for the amazing videos that you upload on the coding lane channel on youtubeI had doubts regarding the elementwise multiplication at 15:00 timestamp in your Derivation of backpropagation in neural networks video. Could you please explain that point to me?
Could you please make video on multiple hidden layers and how the results may change on changing the number of hidden layers? Thank you for this video
Yes, thank you for the suggestion. This video is in my ToDo list. Will upload it when I can.
thank you very much! ❤
Kindly share the code for Multiclass classification for N layer network
Hi Sunitha... I will share a video on it soon.
@@MachineLearningWithJay Thank u for ur immediate reply...I have got stuck up in N layer deep network....It would be of great help if you can take me out of this issue.
@@sunithap915 I won't have the time to take you out of the issue personally. But I have an article that you can refer. It has code and explaination for N layer Neural Network
Here is its link: github.com/enggen/Deep-Learning-Coursera/blob/master/Neural%20Networks%20and%20Deep%20Learning/Building%20your%20Deep%20Neural%20Network%20-%20Step%20by%20Step.ipynb
I hope this can help for now. Also you can search online, there may be some other articles as well on this topic.
Hello, your video is very good, thank you, I have a "small" question, suppose I have a data set with a combination of 1000 functions, let's say Gaussian functions and Lorentzian functions (which are similar functions), the data set would be something like this [[[x1,x2,x3,x4,..,x70],[y1,y2,y3,y4,..,y70],[gaussian]], [[x1',x2',x3',x4',.. ,x70'],[y1',y2',y3',y4',..,y70'],[gaussian]] , [[x1'',x2'',x3'',x4'',.. ,x70''],[y1'',y2'',y3'',y4'',..,y70''],[Lorentzian]] , ....... -1000 functions ] , the idea is that the neural network learns to classify between one function or another depending on the entry points, seeing your code more or less as it could start from your development, perhaps the biggest problem is how do I compute the input of these pairs of data from each function in training.
Bro, you are just awesome as always. But my question is "why i need to implement these code when there was tensorflow"??
Yes, good question. You can implement these in tensorflow very easily. This is just for understanding on how things work behind tensorflow.
I will implement these in tensorflow as well and make video on it.
the partial derivation was hard for me i gotta ask some questions to my calculus teacher however things going well now let's code
Thank you 🙏
thank you a lot !!
Welcome!
helpful .
Thanks
You made ANN a piece of Cake ❤
Good bro
Thanks
bro, isn't it your derivative_tanh(x) function is wrong??
I think it might be,
def derivative_tanh(x):
return (1 - np.power(tanh(x), 2))
please do correct me!!!!!!
Hi Avinash, Thanks for bringing this to my attention. You are correct. I will make correction in assignment. Thanks again 🙂
Wlcm bro!!!!😊😊😊
Bro your videos are super helpful for me
@@Itsavinashdubey I am very glad to hear that 😇
Bro please gimme your LinkedIn profile link I wanna meet you there
@@Itsavinashdubey www.linkedin.com/in/jaimin-patel-4bb166162/
Can u please use loops and if statements next time
Okay... I will try.. Thanks for the feedback
@@MachineLearningWithJay thank you great work
@@apiwejoyi7267 Your Welcome !
Hello Jay. I wrote the code in assignment notebook. But the cost function is not changing with the iterations. Can you help with that?