Neural Network in Python using Keras|How to implement a basic neural network in python using keras

[@jaypatel731]

I think accuracy is good considering very less numbers of examples are available in Training sets. In practical scenario, KNN, XGboost or Decision tree could be more useful.

[@UnfoldDataScience]

Good Observation Jay.

[@seekingthat330]

Thanks Aman Excellent tutorial! May I know why the neurons for the internal layer was chosen as 12 when the inputs were 8 ? input_dim and neurons for hidden layer were chosen as 8. Does it help to have the neurons higher than the inputs ? Is it recommended to have neurons higher than input_dim by a certain percentage ? Thanks

[@franciscoac4717]

The author of the video does not explain this part well. I think that this line of code (which the author comments as "First Hidden Layer") is actually two layers: an input one with 8 neurons and a hidden one with 12 neurons. In any case, it remains to be seen why he used 12 and not 5, 14, 17 or some other amount. Did you research anything about it?

[@ashwinikumar6461]

Quite awesome and Wonderful tutorial.. god bless.. Can you pls tell me why you are using ' _ , ' ( underscore and comma ) in the Train Accuracy alone and not in test accuracy...

[@ramar5558]

Aman in the comments everybody is asking about why you have taken 12 neuron on what basis any rule ????

[@rashidaka4314]

Thank you so much. It was a good class. 👌😊

[@UnfoldDataScience]

You're welcome 😊

[@murapakaganesh1108]

Hi aman i followed you recently , your videos also very good compare than where I have been learning data science ..in my opinion it is healthcare business case problem so we need to have a high accuracy here how we will do hyper parameter tuning please

[@UnfoldDataScience]

All the best Ganesh.

[@veerababun9312]

Hi Aman, got the 80% accuracy by using k-fold cross validation # Constants num_folds = 10 num_instances = len(x) seed = 42 from sklearn.model_selection import KFold cv = KFold(n_splits=10, random_state=42, shuffle=False) cvscores = [] for train_index, test_index in cv.split(x): X_train, X_test, y_train, y_test = x[train_index], x[test_index], y[train_index], y[test_index] # Fit the model model.fit(X_train, y_train, nb_epoch=150, batch_size=10, verbose=2) # evaluate the model scores = model.evaluate(X_test, y_test, verbose=0) print("%s: %.2f%%" % (model.metrics_names[1], scores[1]*100)) cvscores.append(scores[1] * 100) print("Keras: %.2f%% (+/- %.2f%%)" % (np.mean(cvscores), np.std(cvscores))) accuracy: 68.83% accuracy: 83.12% accuracy: 75.32% accuracy: 76.62% accuracy: 79.22% accuracy: 79.22% accuracy: 85.71% accuracy: 84.42% accuracy: 81.58% accuracy: 89.47% Keras: 80.35% (+/- 5.59%)

[@UnfoldDataScience]

ok. Goo to see your commitment.

[@sandipansarkar9211]

great explanation

[@UnfoldDataScience]

Thank you.

[@AbhiAnuP-fp5vp]

Hi, can a similar video on Caffe2 be shared? It would be useful

[@umeshtiwari800]

Good

[@surafelm.w4058]

Hi, how to normalize gridded datasets (X1 and X2) for neural network in python and split as training and testing set?

[@harshyadav1428]

sir i want to make a graphical neural network model on node classification without using ML libraries from scratch by using loops please make video on it

[@milanpatel6240]

Thanks Aman. Very good video. Can you explain me why in the 2nd layer you choose 8 neurons(unit size)? Is it because we have 8 inputs?

[@UnfoldDataScience]

Thanks Milan. Yes.

[@jawadali4392]

In my opinion this model can attain more accuracy. this model is neither good nor bad. Sir kindly tell us how to insert a pictures tensors into this model and classify it. love from pakistan.

[@UnfoldDataScience]

pictures tensors - do you want to classify pictures?

[@jawadali4392]

Yes sir

[@anirbansarkar6306]

Thanks Aman for this great video. I have a doubt although from my side.Why 'relu' as an activation funtion avoids the exploding/vanishing gradient problem? Sorry if I missed your explanation somewhere else (in this case, please refer me that video).

[@UnfoldDataScience]

Relu property is like that, please watch my playlist on neural network u will find a video explaining it.

[@ashwinikumar6461]

Further in continuation to my previous comment, Iam getting A error in the test accuracy code , where it is showing " model is not defined' and hence , would you mind in sharing the exact code link which is used in he video

[@ShaidaMuhammad]

Sir I request you to make these videos in PyTorch, because KZbin has a lot of Keras videos. PyTorch are very little.

[@UnfoldDataScience]

Thank you for your suggestion. Noted.

[@ioogg2dulce]

Hi, very well explained! I have a question, can we train a neural network to classify a dataset which doesn't have targets?

[@UnfoldDataScience]

Training can not happen without target. Predictions can happen without target.

[@ioogg2dulce]

Unfold Data Science I watched your tutorial about hotel reviews classifier and I succeeded to predict on new data! Great job! Thank you very much!

[@ioogg2dulce]

very well explained!!! keep on with the good work!

[@sarwatfawwad]

I am facing problem in uploading keras model. I have also use tensflow to remove error but it is not working

[@AlgoTribes]

I think without doing any sort of tuning to the model and not doing any sort of cleaning of the data before feeding to the model which could be done..i believe the model is doing good with regards to the accuracy metric..😀😀..F1-score probably will give the more clear picture..do correct me if am wrong in assuming this..👍👍

[@UnfoldDataScience]

Yes. up to the point

[@sivaram.k5767]

How to perform abolone age prediction in neural network model sir

[@AmithG-cw1rw]

sir where is the actual model like equation and the plot of ANN how to do it

[@Data_In_real_world]

hello sir its a big kind request please help me .. i need to find weights of all independent VARIABLES of my dataset. how can get the weights which have resulted in my output.

[@nazerinnazeer5941]

Hi sir, great class i have an doubt why u choose 12 neurons in the first layer. Is it any reason?

[@UnfoldDataScience]

Same as number of input features.

[@naveedtaimoor4007]

@@UnfoldDataScience but input features are just 8 as you discussed in video

[@dileepkonkena2319]

we can't check accuracy in classification based on accuracy, we need to consider precision,recall,f1 also

[@aomo5293]

Hello, Thank you for this tutorial, what kind of function that we should use in case we have multiclass ? Not binary ? Thank you

[@engr.simeonstevensonturay61]

In that case, you can use categorical cross entropy

[@PILAKAREDDY]

Hi Aman ,y_pred=model.predict_classes(x) i have error as no model has been created

[@UnfoldDataScience]

check model creation step again

[@srirangasuprabhathkoduru1021]

Can we manually specify the weight initialisation method?

[@UnfoldDataScience]

There are techniques we can use

[@muhammedthayyib9202]

If use scalar, It would be a good model.

[@maryjasmine6767]

This model is like classification.. How to do this for a regression model..?? Is it possible for using neural network for regression model problems.??

[@UnfoldDataScience]

Hi Mary, yes it is possible, we need to have a "linear" or "relu" activation function in output layer.

[@maryjasmine6767]

@@UnfoldDataScienceThank u so much for your response. if possible Could you please share some sample code..?

[@sachinrlu1987]

How 12 output feed from layer 1 to layer 2 with 8 neuron, what will be the processes.

[@gnayaniharshi]

Sir when we compare in terms epochs this is the best model

[@UnfoldDataScience]

Hi Nayani, did not get your question.

[@Sandykrang]

Hi, productive explanation, how can we add Local response normalization in CNN in keras? Can you help with this?

[@UnfoldDataScience]

Hi Sandhya, I am not very sure at this moment, let me note this topic and do some research. Thanks for asking.

[@Sandykrang]

@@UnfoldDataScience ohk , can you please make a video on video data forgery detecting? Can you please help me with the code of how to iterate code for a video to subtract the individual frame from the median frame and repeat the process for the complete video and so on for complete data

[@muhammedt1841]

I think if you scaled the data accuracy will increase.

[@UnfoldDataScience]

Normally yes.

[@pritishbanerjee5517]

Hi Aman, thanks a lot for the tutorial. I have checked the drive shared by you in the description but unable to find this code in the google drive shared by you. Can you please help and upload the code ?

[@UnfoldDataScience]

Hi Pritish, I will check, it was created long back. Am afraid I dont have back up :(

[@ThirdEye097]

Module " tensorflow.python.framework.ops" has no attribute '_tensorlike' What I do ?

[@ThirdEye097]

How to resolve this

[@sabafarheen4918]

Hello , I have a question as we have 8 input features so we kept 8 units in second layer .. but why 12 in first layer ? Can we put any number ?

[@ranjanibhat4417]

Basically it's a thumb rule.... number of first layer is equal to 1.5 times the input layer

[@mranaljadhav8259]

I just used standard scaler to scale those features in same scale and I got accuracy of 84 ℅ .

[@UnfoldDataScience]

Awesome 😊

[@mranaljadhav8259]

@@UnfoldDataScience thanks alot for your nice explanation 😊

[@sagartiwari279]

i have to work on rumour detection using neural networks .... i want to know how can i work with strings instead of integer values... please help me clear my doubt

[@UnfoldDataScience]

Hi Sagar, you should encode strings in numbers. Methods can be onehotendocing etc

[@mabubakerashraf1812]

Can you give the Tutorial of Extreme learning machine code in python

[@UnfoldDataScience]

Sure Mabubaker.

[@64sush]

My quick question is that what is the difference between sequential model and RNN?

[@UnfoldDataScience]

Sequential models - A group of models RNN - A member of the above group

[@bhumikshah2597]

I am getting error while doing kerasmodel.predict_classes(X_test) saying Sequential' object has no attribute 'predict_classes

[@UnfoldDataScience]

Your trained model is having issues

[@preranatiwary7690]

Thank you. I ll get code from repo.

[@UnfoldDataScience]

Sure you can get.

[@akashmanojchoudhary3290]

When I'm trying to fit the model, its shows 62/62, whereas it should show 614/614 as there 614 records in training set, Pls help.

[@UnfoldDataScience]

Check your epoch number.

[@sandipansarkar9211]

75 % accuracy not too good in terms of business accuracy .I believe

[@UnfoldDataScience]

Little higher would be better.

[@libinkoshy632]

Model is good in terms of not overfitting However in terms of accuracy, model isn't doing a great job. And also since it is a medical dataset, Recall will be the right method to guage the model. Am I closer Aman?

[@UnfoldDataScience]

Hi Libin, you r thinking in right direction, feel free to take data and code from my repo and mail me ipynb of an improved model :)

[@libinkoshy632]

@@UnfoldDataScience Thank you Aman :)

[@ajaybhatt6820]

sir make vedio on pytorch,tenserflow quickly

[@UnfoldDataScience]

Sure Ajay. Wil cover the suggested topic. :)

[@ajaybhatt6820]

@@UnfoldDataScience thanku sir 🙏🙏

[@shreyjain6447]

Can a Neural Network work with negative inputs?

[@UnfoldDataScience]

yes it can

[@eswarasatyamsetty5367]

where model is created i can't find it is giving error

[@UnfoldDataScience]

Please post error message.

[@sachinrlu1987]

and why it is decided 12 neuron in first layer why not 16 or 48

[@akashrawat4325]

Hello Sir How can I get this code

[@UnfoldDataScience]

Hi Akash, link to access code is shared in the description of every video. Please check again.

[@akashmanojchoudhary3290]

Drive mein ye wala code nahi hai aman. can you share it?

[@UnfoldDataScience]

Hi Akash, It might be in my old laptop, I will hv to search, Kindly mail me just to remind. Also follow up If I do not Respond.

[@arsiblack2404]

please. make backpropagation

[@wliiliammitiku3996]

model.fit(x_train,y_train,epochs=150,batch_size=10) Etrror here???

[@UnfoldDataScience]

Are you getting any error?

[@wliiliammitiku3996]

@@UnfoldDataScience yes it gets me an error..... C:\Users\user\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\keras\engine\input_spec.py:212 assert_input_compatibility raise ValueError( ValueError: Input 0 of layer sequential_7 is incompatible with the layer: expected axis -1 of input shape to have value 12 but received input with shape [None, 8]

[@nabeel9867]

ValueError: Failed to convert a NumPy array to a Tensor (Unsupported object type int). Getting this error there why??

[@paragpujari3961]

can you please provide the dataset

[@UnfoldDataScience]

drive.google.com/drive/folders/1XdPbyAc9iWml0fPPNX91Yq3BRwkZAG2M

[@nikhielsingh748]

no as the dataset is not balanced , hence accuracy cannot be a reliable metric

[@md.estiakahmed5599]

_, means?

[@UnfoldDataScience]

Which line/part of the video?

[@md.estiakahmed5599]

@@UnfoldDataScience _, accuracy = kerasmodel.evaluate (X_train,y_train) My qs this line what _, meaning?

[@mranaljadhav8259]

@@md.estiakahmed5599 @Md.Estiak Ahmed you can take it (_) has your variable name, instead of using any name he had used underscore.. If you print that (_) underscore you will get loss value of your model.. That evaluate function returns 2 value, 1st is your model loss and 2nd is your model accuracy

[@paragpujari3961]

Error getting while computing accuracy_scoreValueError: Classification metrics can't handle a mix of binary and continuous targets

[@UnfoldDataScience]

datatype issue

[@arpangupta5991]

i want to ipython code of this video

[@UnfoldDataScience]

drive.google.com/drive/folders/1XdPbyAc9iWml0fPPNX91Yq3BRwkZAG2M

[@longtruong9935]

@@UnfoldDataScience what is exactly folder for the code?

[@celina7538]

Based on the accuracy i dont think its a good model..

[@UnfoldDataScience]

Yes there are scopes for improvement.

[@mainuddinali9561]

no