RNN Sentiment Analysis | RNN Code Example in Keras | CampusX

[@soumyajitpal8273]

The following correction is needed - model.add(Embedding(input_dim=18, output_dim=2, input_length=5)) input_dim = Size of vocabulary (maximum integer index + 1); in this case 18 (17+1) output_dim = Desired embedded vector length; in this case 2 input_length = Length of input sequences after padding; in this case 5

[@pranavreddy9218]

looks like sir didn't run the code, just copied from some where and teaching, anyway he is gathering nice info..

[@k-AE-MDWazidAnsari]

I thought bacause of your menrotship program you are stopping deep learning playlist. ;-;. Please don't ever stop making these dl videos. I have much more to learn and you are my only hope.. ;-;.. please. dont stop...

[@te_b4_73_sushant_yelurkar4]

Same bro Mea to dar hi gya tha

[@SleepeJobs]

Thank you sir for such quality content for free. Keep them coming at you comfortable pace and don't exhaust yourself 🙌🙌

[@rajkachhadiya6192]

Hello Sir, I really like your work and no one provides this type of good content even in a 2lakhs rupees course also. I really appreciate your work and heartily thank you. Even if we are paying too much money for our degree then also we prefer your content for our base and then see other high levels of content. After your content its looks like a very simple thing. Thank you so much.

[@muskangupta4864]

Please keep uploading deep learning videos.. It's one of the best channnel.. Don't ever stop these videos plz plz plz🥺🥺🥺🥺🥺

[@lug__aman]

correct

[@abhishektehlan7814]

can u suggest some similar channel like campusx for deep learning

[@sayanroy281]

@@abhishektehlan7814 campusx

[@near_.]

@@abhishektehlan7814 if u get pls let me know

[@suvankarpaul7483]

6:63

[@pammasinghkainth]

Good Sir, Your lectures are really good. We appreciate the amount of effort you put teach these topics!

[@JALAJBANSAL-k5b]

Use the following to make the model: model = Sequential() model.add(Embedding(input_dim=88586, output_dim=2, input_length=50)) model.add(SimpleRNN(32, return_sequences=False)) model.add(Dense(1,activation='sigmoid')) model.summary() this gives a higher val_accuracy: 0.7865 when I ran it.

[@JourneytoWINover_PORN]

why r u taking such high input_dim ? ig taking such high inout_dim means having more data and dl is data hungry that is why you got high val_accuracy and wht abt train_accuracy?

[@aanchalsharma2567]

input dimension(88586) shows here number of sentences which are made using different unique words???

[@pranavreddy9218]

@@aanchalsharma2567 i tried to get number of unique words in training set is 24902 before padding and 24893 after padding. where did we reach 88586, but working fine with this number

[@thatsfantastic313]

you are no doubt the best teacher

[@sarojdahal4406]

sir after completing each section plz solve one project with best model ,it will helps a lot ......lots of love from nepal...

[@hassanjaved906]

Haye brother what is my first word when I see this "f**k what are you doing" damn that is the best explaination I have seen in DL you are awesome, Thank God I just found you in a right time, sorry for sharing my thoughts after while taking this lecture, I have no words to explain, you are champ for me.

[@akashprabhakar6353]

Thanks a lot. Awesome video. Embedding seems powerful. Drastic improvement in model.

[@alroygama6166]

God bless u sir with everything u need in life

[@ZEENATFATIMA-jk1zl]

at time 33.32 it you give input size also model.add(Embedding(10000, 2,50)) then it raises error that it doesnot recognise identifier 50

[@JourneytoWINover_PORN]

r u single? ready to mingle?

[@mushtaqahmed-my9vy]

you really great sir

[@trishanudas438]

The best on Yt

[@guyswapnil]

Sir Please Upload Videos on DL we are highly dependent on you.

[@Eswar.]

Should I also need to learn deep learning I have completed his 100days of ml and done some projects

[@guyswapnil]

Go ahead with DL

[@JourneytoWINover_PORN]

@@Eswar. wht r u doing now? ig even whole ml is very much so no need for dl actually

[@narendraparmar1631]

Nice explanation of RNN, Thanks

[@mdabdullahalhasib9081]

Thank you sir.. Awesome lecture..

[@piyushpathak7311]

Sir please upload videos in Deep learning playlist daily pls

[@lug__aman]

sir video daily upload kiya kro please

[@Eswar.]

Should I also need to learn deep learning I have completed his 100days of ml and done some projects

[@mushtaqahmed-my9vy]

love u sir

[@sahuchiragshyamlal3684]

word embedding can work in less dimension bcz dense represet and semantic meaning capture. pretrain word2vec

[@ParthivShah]

Thank You Sir.

[@Ishant875]

I have a doubt, please reply if anyone knows , if RNN can take input of different sizes then why we are using padding to make each input of equal size?

[@VickyPatel1293]

because of use of batch-wise training>>Padding is needed if we want to train them batch-wise. A batch is a rectangular tensor of fixed size. This does not work with sentences of varying lengths, which is why you can either pad or truncate the sentences to a uniform length. for better understanding check article on hugging face or stackoverflow.

[@JaishreeramCoder]

@@VickyPatel1293 Can I say that as in RNN, the weight dimension depends on word vector dimensions and not on each sentence size? So padding will not increase no. the parameters? which was the one of reasons why ANN wouldn't work well for texts

[@VickyPatel1293]

@@JaishreeramCoder Yes, in RNNs, the weight dimension depends on the word vector dimensions, not on each sentence size. Padding doesn't increase the parameters because the weights are shared across the sequence length, which helps address the issue of variable-length inputs in text data. This is one reason why traditional artificial neural networks struggled with text processing tasks. Or you can take one example like you are training robot with some set of sentences. So if we use extra words(padding) to make equal length of sentences to train the robot then it will not increase their brain size as we are not using dedicated weights to each words while training

[@rajamurishiva2084]

please continue the series

[@hiddenworld1445]

Plz keep uploading

[@Akhilesh-bq8kl]

i am using this archhitecture model=Sequential() # units are nodes are added model.add(Embedding(input_dim=1000,output_dim=2,input_length=50)) # embedding converts vectors with semantic meaning model.add(SimpleRNN(units=32,return_sequences=False)) model.add(Dense(1,activation='sigmoid')) model.summary() model.compile(optimizer='adam',loss='binary_crossentropy',metrics=["acc"]) r=model.fit(x_train,y_train,epochs=5,validation_data=(x_test,y_test)) but getting this eror indices[0,47] = 38286 is not in [0, 10001) [[{{node sequential_7/embedding_6/embedding_lookup}}]] [Op:__inference_train_function_12881] why??

[@JourneytoWINover_PORN]

wht r u doing now?

[@Naman_Bansal102]

Wow Good Video Sir

[@rafibasha4145]

Nitish bro ,next video pls ..i felt you finishied CNN in hurry .pls cover each topic in depth like how you do and upload wekkly 2 videos atleasr

[@campusx-official]

Will upload the next video on Thu. Also, let me know what topics did I miss in CNN. Or any other feedback.

[@rafibasha4145]

@@campusx-official ,thanks for the reponse bro ,different cnn archutectures not covered like resnet inception etc ,tensorboard not covered,learning rate scheduler....i am adding NLP also here dependency parsing coreference r3solution topic modelling

[@campusx-official]

@@rafibasha4145 ok, will come back to these topics once I complete rnn

[@rafibasha4145]

@@campusx-official ,thanks for your brillinat teaching bro ..i would love to learn MLOPS,Time series ,GAN ,advance NLP from you ..Your channel will be number 1 in India with millions of subscribers soon

[@ZEENATFATIMA-jk1zl]

@@campusx-official at time 33.32 it you give input size also model.add(Embedding(10000, 2,50)) then it raises error that it doesnot recognise identifier 50

[@sandipansarkar9211]

finished watching and coding

[@skplk9221]

Sir RNN mai hmm jo pad_sequences ki help se textual data ko number mai convert krte hai. Too kya RNN mai hmm pad_sequence ki jagah CountVectorizer use kr skte hai text ko number main convert krne ke liye

[@miscellaneoushub1956]

Sir amazing video, RNN to kamal hi h. sir lstm and gru kab aaraha hai...

[@satyamgupta4808]

vry very vry nice channel

[@sejalninni3076]

Could you please make a video on LSTM. its hard to understand from others and theory in google resources.

[@near_.]

If u get please let me know. Iam in learning phase. 😊🙏

[@machineinsurgence87]

Did you get it yet?

[@madhuril1397]

Did u get any good video on LSTM?

[@JourneytoWINover_PORN]

wht abt now?would like to share?

[@kkkk150984]

But can you tell how can same RNN model can retrained with more data?

[@Shuraim843]

sir update deep learning playlist

[@Eswar.]

Should I also need to learn deep learning I have completed his 100days of ml and done some projects

[@mdriad4521]

Of course

[@Eswar.]

@@mdriad4521 but i don't have enough time brother I love computer vision (DL) but i need a job now learning deep learning may kill my time

[@nikhiliyer8436]

@@Eswar. then dont do it ,why you are asking when you know you dont have time

[@rahuljha3686]

Done✔️

[@aiforeveryone]

Cool

[@aninditasaha0307]

I have a question...if in this line "X_train = pad_sequences(X_train,padding='post',maxlen=50)" I am not using maxlen=50 and taking all words, what should i write in this line model.add(SimpleRNN(32,input_shape=(50,1),return_sequences=False)) got an error...thats why asking

[@brahmadevrai4369]

Bro ur if u use entire data then change ur input shape

[@priyadarshichatterjee7933]

keep your input shape as the dimension of X_train or X_test which ever is higher. In this dataset keep it at (2492,1)...it will work fine

[@ekramshethil7932]

when i using this code my pc : model.add(Embedding(10000,2,50)) my pc error give me : ValueError: Could not interpret initializer identifier: 50 How I solve this please tell me

[@thequietkid5212]

change your runtime from cpu to gpu

[@thequietkid5212]

remove '50' he didnt run that cell it was only for explaining

[@mritunjay3723]

From where you have copied this ... 😅😅😅

[@asrahussain8642]

unsubscribed btw

[@himeshsarkar3259]

kyu bhai?

[@AshishSharma-tf3fy]

@@himeshsarkar3259 bro read his name , he is muslim . since sir has used modi ji ki jai his community got offended

[@AshishSharma-tf3fy]

good for u bro , deep learning won't help u . go fix tyre's

[@himeshsarkar3259]

@@AshishSharma-tf3fy oh. Just saw modi speech in KZbin saying mutton machli logic and connotating with muslims and now your notification. Irony!

[@AshishSharma-tf3fy]

@@himeshsarkar3259 algorithm on his best ! hahah

[@MB-yw8ky]

hi pls help me out,,, model.predict(sequences) is generating an error InvalidArgumentError ----> 1 pred = model.predict(sequences) InvalidArgumentError: Graph execution error: indices[9,1] = 17 is not in [0, 17) [[{{node sequential/embedding/embedding_lookup}}]] [Op:__inference_predict_function_128]

[@yashsoni8043]

is your problem solved cause i am facing the same problem tell me how to fix this

[@soumyajitpal8273]

model.add(Embedding(input_dim=18, output_dim=2, input_length=5)) input_dim = Size of vocabulary (maximum integer index + 1); in this case 18 (17+1) output_dim = Desired embedded vector length; in this case 2 input_length = Length of input sequences after padding; in this case 5

[@aayovichaariye]

i am not able to take only 10000 sentences during embedding Node: 'sequential_14/embedding_12/embedding_lookup' indices[23,8] = 38275 is not in [0, 10000) [[{{node sequential_14/embedding_12/embedding_lookup}}]] [Op:__inference_train_function_103656] getting this error

[@rounakshow2150]

##Define a vocabsize. vocab_size = 10000 (X_train,y_train),(X_test,y_test) = imdb.load_data(num_words =vocab_size) X_train = pad_sequences(X_train,padding='post',maxlen=100) X_test = pad_sequences(X_test,padding='post',maxlen=100) model = Sequential() model.add(Embedding(input_dim = vocab_size, output_dim= 2,input_length = X_train.shape[1])) model.add(SimpleRNN(32,return_sequences=False)) model.add(Dense(1, activation='sigmoid'))

[@sumanthnimmagadda4472]

@@rounakshow2150 thanks alot