00:51 Starting AN practical implementation with data set and code download 03:43 Practical implementation and understanding of different optimizers. 08:41 Implementing Artificial Neural Network with TensorFlow GPU in Google Colab 10:51 Setting up TensorFlow and importing necessary libraries 15:12 Identifying independent and dependent features for data set division 17:16 Implementing pd.get_dummies in feature engineering 20:59 Updating and concatenating categorical features. 22:47 Implementing Train-Test Split for Data Set 26:51 Feature scaling is essential for quicker convergence in regression 28:39 Explanation of fit-transform vs. transform 32:44 Tensorflow 2.0 integrates Keras, making it a wrapper over TensorFlow for better usage 34:45 Importing activation functions and dropout layer in neural network implementation. 38:23 Introduction to activation functions and dropout layer 40:12 Understanding Dropout Layer in Neural Networks 43:49 Setting up multiple hidden layers in a neural network 45:44 Compile neural network using Adam optimizer and binary cross entropy 49:22 Implementing early stopping in ANN training for optimal epoch selection 51:08 Validation split and stopping the training process based on accuracy 54:54 Implementing early stopping in training neural networks 56:47 Implementing early stopping in model training 1:00:53 Practical implementation of early stopping and prediction using classifier 1:02:49 Understanding Neural Network Weights and Storing 1:07:19 Using dropout in neural networks and understanding black box vs white box models 1:09:24 Understanding the difference between white box and black box models 1:12:54 Monthly community sessions covering various topics. Crafted by Merlin AI.
Don't worry about the people who are attending the session... Your session is amazing and beneficial....it is helping many people....keep doing great thing
Hi krish, ur teaching is amazing. By watching ur videos only I have learning deep learning and machine learning. I have finished two course work by learning from you. Thanks a lot💗💖💓🙏🙏🙏🙏
As always you are the BEST.... We may not join live sessions.. but we make sure to watch recorded sessions.. as your trainings are always interesting to explore....So keep doing live sessions
We love your explanation and also the references you share. I have been following all your lectures back to back. Thanks for creating tutorials and sharing so much knowledge with us.
thankyou so much, it was a pretty pretty amazing lecture, watched every minute of it and finally made my first ANN. Thankyou for the efforts. It meant a really lot for us.
Respected Sir, Your hard work is really very much helpful for the people like me. Sir, I am facing lot of problem in Reinforcement learning. I am requesting you to please upload a video based on a loan approval automated system (If above than credit score/threshold, then approve the loan else reject the loan) based on past payment behavior/History (in data set) of customer using Reinforcement learning. If possible, please also include exposure as profit for bank.
Hi Krish, greetings from Chile! I have a question, the first layer (Input Layer) I noticed that you use an activation function to the inputs. Is it OK to do that? I imagine it as a transform to your inputs doing a Relu and then that info you pass to the next layer, but you loss all inputs that are less than zero since with the StandarScaler transformation you would transform certain data and obtain data with negative values. It not will better just apply Dense (with no activation) on the Input Layer?
Krish for people like me who are trying dl for first time, need more clarification in terminology used during modeling onward day 4, like early stopping, dense and all. Also in notebook early stopping is not used instead you have used 50 epochs.
Thank you sir, I have a basic question: how do you know how many hidden layers and how many nodes in each hidden layer . Why did you choose in this example 2 hidden layers and 7 and 6 nodes for each layer ? Eran
Sir my data is having very high dimension(20000-50000)(tabulardata with numerical values) and small data size, I've done augmentation and my model is having overfitting when i used ANN, how can I reduce the dimension to select significant features? Also how to avoid overfitting?
What if let's say after 50 epochs, our accuracy is not increasing. So, let's say we decided to pause the iteraions and change the parameters like learning rate and activation function to see whether it further increases the accuracy instead of early stopping. So, from the 51 epoch, we will have new activation function and learning rate. Is this possible or we need to start from beginning if we want to use new parameters.
ameError Traceback (most recent call last) in () ----> 1 model_history=Classifier.fit(X_train,y_train,validation_split=0.33,batch_size=10,epochs=1000) NameError: name 'Classifier' is not defined how to improve it
HI Sir I am getting error while doing the prediction. Please help. y_pred=Classifier.predict(X_test) Error: ValueError: Failed to find data adapter that can handle input: ,
Hi Krish, I'm facing below error while executing the command clasifier.compile(optimizer='adam',loss='binary_crossentropy',metrics='accuracy') in pycharm. Error: OperatorNotAllowedInGraphError: using a `tf.Tensor` as a Python `bool` is not allowed: AutoGraph did convert this function. I tried various method but unable to unblock, Could you please help.?
No, the concept is totally different. Encoding is used for making the values of the column into 0 and 1 , while we do StandardScaler for bringing the values between 0 and 1. So no need to change categorical variable with StandardScaler
1:03:18 sir score is 85 % but in confusion matrices we see that for false or 0 there is 215 , 190 that's around 55 % accuracy that is not good how to resolve this
HEY KRISH! I WAS FORTUNATE ENOUGH TO LEARN FROM YOUR COMMUNITY SESSIONS INSPITE OF MISSING THE LIVE. BUT NOW WHEN I AM TRING TO ACCESS THE DATASET, THE WEBSITE IS SHOWING AN ERROR. COULD YOU PLEASE HELP ME BY SORTING IT OUT? THANK YOU
its a trial and error process , if model is under fitting add more neurons or layers, if overfitting reduce the number, the specific number of neurons itself if arbitrary
00:51 Starting AN practical implementation with data set and code download
03:43 Practical implementation and understanding of different optimizers.
08:41 Implementing Artificial Neural Network with TensorFlow GPU in Google Colab
10:51 Setting up TensorFlow and importing necessary libraries
15:12 Identifying independent and dependent features for data set division
17:16 Implementing pd.get_dummies in feature engineering
20:59 Updating and concatenating categorical features.
22:47 Implementing Train-Test Split for Data Set
26:51 Feature scaling is essential for quicker convergence in regression
28:39 Explanation of fit-transform vs. transform
32:44 Tensorflow 2.0 integrates Keras, making it a wrapper over TensorFlow for better usage
34:45 Importing activation functions and dropout layer in neural network implementation.
38:23 Introduction to activation functions and dropout layer
40:12 Understanding Dropout Layer in Neural Networks
43:49 Setting up multiple hidden layers in a neural network
45:44 Compile neural network using Adam optimizer and binary cross entropy
49:22 Implementing early stopping in ANN training for optimal epoch selection
51:08 Validation split and stopping the training process based on accuracy
54:54 Implementing early stopping in training neural networks
56:47 Implementing early stopping in model training
1:00:53 Practical implementation of early stopping and prediction using classifier
1:02:49 Understanding Neural Network Weights and Storing
1:07:19 Using dropout in neural networks and understanding black box vs white box models
1:09:24 Understanding the difference between white box and black box models
1:12:54 Monthly community sessions covering various topics.
Crafted by Merlin AI.
Don't worry about the people who are attending the session... Your session is amazing and beneficial....it is helping many people....keep doing great thing
Hi Krish just now i joined Full stack Data Science course. I like your teaching that's why i joined.
Thank you
Hat's up to you....krish while others are taking a huge amount of money ...u r giving free community session.
Hi krish, ur teaching is amazing. By watching ur videos only I have learning deep learning and machine learning. I have finished two course work by learning from you. Thanks a lot💗💖💓🙏🙏🙏🙏
Sessions helped me to learn things that i couldn't complete in 6 months. Thanks a lot krish
🔥 Fantastic Explanation! Krish, your explanation and implementation on ANN concepts is commendable. Keep up the great work!
I m going through all of ur videos, seriously krish it's quite amazing
loved the explanation, u simplified it and made it so easy to understand. u r just awesome.
As always you are the BEST.... We may not join live sessions.. but we make sure to watch recorded sessions.. as your trainings are always interesting to explore....So keep doing live sessions
We love your explanation and also the references you share. I have been following all your lectures back to back. Thanks for creating tutorials and sharing so much knowledge with us.
thank you Krish for this amaizing course it really help me a lot
much love
keep up the good deed
Hello Krish, can you please start NLP live ASAP please? This all would be in a proper sink and would be able to connect concepts.
dont worry he will start the nlp live session after finishing deep learning.He has promised👍
your sessions are amazing. Thanks for coming up with these lectures
thankyou so much, it was a pretty pretty amazing lecture, watched every minute of it and finally made my first ANN.
Thankyou for the efforts. It meant a really lot for us.
Amazing always love your video and you are the source of my knowledge in DS, also in the pin section there is no dataset, so kindly look into it.
Respected Sir, Your hard work is really very much helpful for the people like me. Sir, I am facing lot of problem in Reinforcement learning. I am requesting you to please upload a video based on a loan approval automated system (If above than credit score/threshold, then approve the loan else reject the loan) based on past payment behavior/History (in data set) of customer using Reinforcement learning. If possible, please also include exposure as profit for bank.
Hello Krish Thanks for all the videos. I have a request to make series of video on computer vision with math and code. Thanks
Had crystal clear understanding krish thankyou
finished coding .Feeling confident
hello sir, What happen to Day5? looking forward towards another amazing session .. Thank you so much for amazing initiative that you have taken.
Maaaaaan, you are awesome!
Amazing session thank you sir. Your sessions are very interesting to learn
Hi Krish, greetings from Chile! I have a question, the first layer (Input Layer) I noticed that you use an activation function to the inputs. Is it OK to do that? I imagine it as a transform to your inputs doing a Relu and then that info you pass to the next layer, but you loss all inputs that are less than zero since with the StandarScaler transformation you would transform certain data and obtain data with negative values.
It not will better just apply Dense (with no activation) on the Input Layer?
Krish for people like me who are trying dl for first time, need more clarification in terminology used during modeling onward day 4, like early stopping, dense and all.
Also in notebook early stopping is not used instead you have used 50 epochs.
thank you sir krish for everythink.
Thank You Sir !
great session
while concatenating it raises issue as 'geography is not defined but I copypasted exactly like sir !!!
Very interesting lecture.
Can I get all the
Videos of the lecture?
Thank you for the session sir ❤
Amizing thank u karish bhai😍😊
Excellent
Thank you sir,
I have a basic question:
how do you know how many hidden layers and how many nodes in each hidden layer . Why did you choose in this example 2 hidden layers and 7 and 6 nodes for each layer ?
Eran
Hi Krish, can you please create similar community session video on feature engineering + feature selection by preventing data leakage.
Big thankew Sir
I can’t join the live session but I am able watch those videos later early morning
really helpfull❤❤❤❤❤
Thankyou sir❤️🔥
Sir waiting for Day 5 and NLP session.
Hi, Sir. Can we use Label encoder here for the categorical features or map as per the unique features available in the column???
Thank you sir....
nice explaination
Sir my data is having very high dimension(20000-50000)(tabulardata with numerical values) and small data size, I've done augmentation and my model is having overfitting when i used ANN, how can I reduce the dimension to select significant features? Also how to avoid overfitting?
Sir please make a course on Algorithmic Trading
Hi sir ...from where I can get ....dataset. ...also study material....not showing in given link ...disc
how can we apply Deep Learning for small size, high dimensional data? Also how to combine different small size dataset with variations in dimensions?
What if let's say after 50 epochs, our accuracy is not increasing. So, let's say we decided to pause the iteraions and change the parameters like learning rate and activation function to see whether it further increases the accuracy instead of early stopping. So, from the 51 epoch, we will have new activation function and learning rate. Is this possible or we need to start from beginning if we want to use new parameters.
Thankyou Krish ..for the great sessions and your efforts.. learnt a lot of concepts easily .. when are we going to have the final day CNN session?
On Monday
finished watching
i love u sir
@32:00 tensor flow starts
Where is the data? There is no comment pinned.
drive.google.com/file/d/1ZRXIpzLhbtBccjauzROH5AuY3aZl7faW/view
drive.google.com/file/d/1ZRXIpzLhbtBccjauzROH5AuY3aZl7faW/view
❤
ameError Traceback (most recent call last)
in ()
----> 1 model_history=Classifier.fit(X_train,y_train,validation_split=0.33,batch_size=10,epochs=1000)
NameError: name 'Classifier' is not defined
how to improve it
Nice explanation, From where I can get the code notebook
It will get uploaded in the dashboard
@@krishnaik06 is Day 5 postponed
Dataset is in the live chat pinned comment!!
Please, how can I get the dataset used? The link pasted below is not found again
No soft max in the previous class, Can u plz teach us
Do we have Day-5 lect today?
Hey brother I work in Punjab national bank and want to be a data scientist what should I do , how should I start?
Sir i just want to know why is there an activation function for the inputs
hello Sir, I am getting this error when running the epoch command, please help - checked stack over flow but was not able to get help
HI Sir I am getting error while doing the prediction. Please help.
y_pred=Classifier.predict(X_test)
Error: ValueError: Failed to find data adapter that can handle input: ,
Hi sir, where is link for data to download for practical
Hi Krish, I'm facing below error while executing the command clasifier.compile(optimizer='adam',loss='binary_crossentropy',metrics='accuracy') in pycharm.
Error: OperatorNotAllowedInGraphError: using a `tf.Tensor` as a Python `bool` is not allowed: AutoGraph did convert this function.
I tried various method but unable to unblock, Could you please help.?
have you tried metrics='[accuracy']
Sir day 5 pls upload
Hello, May I know when will be the Day-5 community session?
not able to access the contents ,please help !!!
Should we do standardscalar for categorical features which are encoded... Like one hot, or label encoded...?
No. their values are already between 0-1. so, no need to apply standardscaler on them.
No, the concept is totally different. Encoding is used for making the values of the column into 0 and 1 , while we do StandardScaler for bringing the values between 0 and 1. So no need to change categorical variable with StandardScaler
Sir i want to apply the hybrid encryption in block chain how this is deploy
Where is the pinned comment
Kha h dataset
1:03:18 sir score is 85 % but in confusion matrices we see that for false or 0 there is 215 , 190 that's around 55 % accuracy that is not good how to resolve this
where is the dataset?
Installing tensorflow gpu is giving error _:- error: metadata-generation-failed
dont use -gpu
not required
just run !pip install tensorflow
Sir where is Churn dataset csv and this code.
HEY KRISH! I WAS FORTUNATE ENOUGH TO LEARN FROM YOUR COMMUNITY SESSIONS INSPITE OF MISSING THE LIVE. BUT NOW WHEN I AM TRING TO ACCESS THE DATASET, THE WEBSITE IS SHOWING AN ERROR. COULD YOU PLEASE HELP ME BY SORTING IT OUT?
THANK YOU
Updated the link
Sir, How to know how many hidden layers and neurons we need to add?
its a trial and error process , if model is under fitting add more neurons or layers, if overfitting reduce the number, the specific number of neurons itself if arbitrary
where we can find the link of above code
where is data set. I could n't find. Plz share
its in live chat
drive.google.com/file/d/1ZRXIpzLhbtBccjauzROH5AuY3aZl7faW/view?pli=1
sir Please share the data set link
Where is the pinned comment for the dataset? Can anyone tell please...???
drive.google.com/file/d/1ZRXIpzLhbtBccjauzROH5AuY3aZl7faW/view
No live session today?
I am not able to find the pinned comment.(Is it removed now?).
drive.google.com/file/d/1ZRXIpzLhbtBccjauzROH5AuY3aZl7faW/view
@@MuruganVeera1980 thanks
Anyone know good Channel for Django ?
Sir Day 5 session??
It will be on Monday
@@krishnaik06 thank you.
where can i find the dataset???
drive.google.com/file/d/1ZRXIpzLhbtBccjauzROH5AuY3aZl7faW/view
Where is data set.?
drive.google.com/file/d/1ZRXIpzLhbtBccjauzROH5AuY3aZl7faW/view
😇😇😇😇😇😇😇😇
Sir Deep learning live session5?
It will be on Saturday or Sunday
Thank you sir
@@krishnaik06 We need a session on CNN. Please add the 5th day
hello i am getting error while importing relu
i am getting below error
cannot import name 'ReLu' from 'tensorflow.keras.layers
❤🧡💛💚💙