Recurrent Neural Networks (RNN) – Deep Learning w/ Python, TensorFlow & Keras p.7




[ad_1]

In this part we’re going to be covering recurrent neural networks. The idea of a recurrent neural network is that sequences and order matters. For many operations, this definitely does.

Text tutorials and sample code: https://pythonprogramming.net/recurrent-neural-network-deep-learning-python-tensorflow-keras/

Discord: https://discord.gg/sentdex
Support the content: https://pythonprogramming.net/support-donate/
Twitter: https://twitter.com/sentdex
Facebook: https://www.facebook.com/pythonprogramming.net/
Twitch: https://www.twitch.tv/sentdex
G+: https://plus.google.com/+sentdex

Source


[ad_2]

Comment List

  • sentdex
    November 17, 2020

    I want to classify anomaly detection using RNN keras.tf but I have a problem where the accuracy value increases but the val_accuracy value does not change and just remains constant at 50%. this is my complete code available on google colab https://colab.research.google.com/drive/1saoNuCxj08JCxZ_7taIjhp8sJEIV-T5U?usp=sharing

  • sentdex
    November 17, 2020

    Can we have the codes?

  • sentdex
    November 17, 2020

    Either way

  • sentdex
    November 17, 2020

    I really don't get it, when I run the code each Epoch is 1875 samples, when you run it it is 60K..

  • sentdex
    November 17, 2020

    Hey man, i like your video. but actually i love your text editor!! what's it?

  • sentdex
    November 17, 2020

    @sentdex Question. Your code works beautifully. However, if I try to create the model by passing Sequential a list of the layers (i.e. `[LSTM, Dropout, LSTM, …]` instead of using `model.add`, I get an input shape error. Do you know why? (for reference, I have this on GitHub under Issue #42986)

  • sentdex
    November 17, 2020

    LOVE your work sentdex!! Long time viewer. Just gotta say though…what's up with that coffee mug?! I busted up when I saw that! No pause, no mention at all. Just hold on guys, I'm gonna take a sip here…hahaha! I want that mug. You are epic!

  • sentdex
    November 17, 2020

    How you set environment for deep learning in sublime text editor?

  • sentdex
    November 17, 2020

    what is mnist??somebody pls answer

  • sentdex
    November 17, 2020

    Can u help me the differences between conventional LSTM and Random Connectivity LSTM(RCLSTM) when it comes to code?

  • sentdex
    November 17, 2020

    How to combine CNN and RNN for Detection problem?? anybody help..

  • sentdex
    November 17, 2020

    You claim that RNNs are for time series but choose a normal supervised learning problem (MNIST dataset). What is then the difference between LSTM and other deep NN models such as CNNs? I watched this video to clear up this confusion and I am now even more confused. Aren't RNNs supposed to be for time series? I mean real-time series. I fail to see how 28×28 dimensional input is a time series. It's spatial data, not temporal.

  • sentdex
    November 17, 2020

    Thank you for making this video basic.
    I am very new to tensorflow and keras in general, just learnt it last 2 weeks.
    Thank you.

  • sentdex
    November 17, 2020

    Why do we use x_train.shape[1:] for the input statement. What does [1:] mean when it comes to shape?

  • sentdex
    November 17, 2020

    I have a question. If the input is 28 by 28 and I am guessing that it will be flattened to (784,1) but the cells are 128, how does that work? are we putting x1, x2, x3,…x128 at the same time?

  • sentdex
    November 17, 2020

    You sound a bit like Edward Snowden. Very good explanation (and drawings)!

  • sentdex
    November 17, 2020

    Great work, thanks; could you post an example of RNN for NLP

  • sentdex
    November 17, 2020

    Most people probably know this but when he normalizes the data and grabs "255.0" out of the sky it is because the mnist data-set is giving a 28×28 array of number digits in gray-scale by assigning a pixel shade of 0-255 ; 0 being black space and 255 being white; if you print(x_train[1]) you can tell it is a '0' and prove that by printing y_train[1]; dividing all pixels by 255 scales all image data between 0 and 1

  • sentdex
    November 17, 2020

    Hi, would you please guide me?
    what should I do when I get these errors:
    2020-05-29 14:11:57.862761: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcublas.so.10.0

    2020-05-29 14:11:58.365592: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudnn.so.7

    2020-05-29 14:11:59.372621: E tensorflow/stream_executor/cuda/cuda_dnn.cc:329] Could not create cudnn handle: CUDNN_STATUS_INTERNAL_ERROR

    2020-05-29 14:11:59.372700: F tensorflow/core/kernels/cudnn_rnn_ops.cc:1624] Check failed: stream->parent()->GetRnnAlgorithms(&algorithms)

    Have I done anything wrong with regarg to my code, or is this issue due to the installation of Tensorflow or etc.?
    Moreover, I had ran my model with LSTM for about 1000 times till today, but when I used the CuDNNLSTM it didn't work and gave these errors.

  • sentdex
    November 17, 2020

    Thanks, it was great video

  • sentdex
    November 17, 2020

    I'm a little confused… When you run yours it has like x/60000 on the left when you run it but I only have 1875 anyone knows why? I did it exactly the same as him. Also with the newer version, you don't have to do the custom layer like he did for the GPU

  • sentdex
    November 17, 2020

    I am trying this
    But I am getting this type of warning and I am unable to us CuDNNLSTM

    WARNING:tensorflow:Layer lstm will not use cuDNN kernel since it doesn't meet the cuDNN kernel criteria. It will use generic GPU kernel as fallback when running on GPU

    Please Help!!

  • sentdex
    November 17, 2020

    Are there any updates now? For doing all the same stuff in better way? As you had made this video long back !

  • sentdex
    November 17, 2020

    Can I use this for Image Classification?

  • sentdex
    November 17, 2020

    Hello sentdex,

    I'm working on my Bachelor final project and would like to implement RNN to train some sequential data. I've seen your videos on the topic but I still don't know how to organize the data. Could you help me with that?

    Great videos by the way, keep on going like this!

  • sentdex
    November 17, 2020

    You can simply say use COLAB but well done everything is clear to me

  • sentdex
    November 17, 2020

    I have a variable length input, which is a signal. How to input variable input shape without using any padding all the time?

  • sentdex
    November 17, 2020

    Note: Since Tensorflow 2.0 it will automatically take the CuDNN version if you specify no activation function

  • sentdex
    November 17, 2020

    In TensorFlow 2.0, the built-in LSTM and GRU layers have been updated to leverage CuDNN kernels by default when a GPU is available. But with conditions – must use default activation function 'tanh'.

  • sentdex
    November 17, 2020

    I have a small doubt can we use Recurrent neural networks on tabular data such as Defect predication classification task

  • sentdex
    November 17, 2020

    So the piece of code that classifies this as a RNN is 'return_sequences = True', am I right?

  • sentdex
    November 17, 2020

    THANK YOU SO MUCHHH FOR THIS VIDEO!

  • sentdex
    November 17, 2020

    I want to more about tensorflow.callbacks

  • sentdex
    November 17, 2020

    Hello thans for helping is to understand RNN
    please tell wich version of Tensorflow and keras
    because i tried your code and i have some problem
    like
    TYpeError add() got unexpected argument 'activation'

  • sentdex
    November 17, 2020

    This is really helpful. I was looking for a simple intro to RNN and LSTM, but couldn't find anywhere in tensorflow 2.1. But this one is simple, up-to-date version. Many thanks.

  • sentdex
    November 17, 2020

    You look and sound like Edward Snowden.

  • sentdex
    November 17, 2020

    Sorry, I don't understand what's the purpose of the RNN here. You're just feeding images of numbers into the RNN and then training on the labels. There isn't any order or anything. You can do the same with NN or CNN. It would have been cool if the network could fill in (predict) blank spots in the image. Like, if the image is damaged and some pixels are missing, you could use some model to fill in the gaps. I thought RNNs could do that.

  • sentdex
    November 17, 2020

    Hey when you realized it was learning slow you “normalized” the data by dividing both x_train and x_test by 255.0. Why was that number chosen?

  • sentdex
    November 17, 2020

    Which version of TF is this?

  • sentdex
    November 17, 2020

    It would be great if you tell beforehand what we are going to do with the MNIST and maybe draw a rough diagram of the structure of RNN with MNIST as input.

  • sentdex
    November 17, 2020

    If CUDNNLSTM uses tanh which goes from -1 to 1, shouldn't you renormalize the input by dividing by 127 and then subtracting 1 ?

  • sentdex
    November 17, 2020

    One question I have is when using decay, in theory it should stop at a minimum. However in practice, I find after it reaches the min it bounces out wildly. Is this just due to pythons rounding errors at e-17?

  • sentdex
    November 17, 2020

    my notebook frozze :c

  • sentdex
    November 17, 2020

    how many cups do you have

  • sentdex
    November 17, 2020

    He looks like Edward snowden

  • sentdex
    November 17, 2020

    Thank you for the tutorial.
    I have a question about the input dimensional. My input data are videos, is it true the sequences indicated the totally frames of the video and the elements indicated feature number of each frame?

  • sentdex
    November 17, 2020

    why don't you try vscode

  • sentdex
    November 17, 2020

    Why return sequences set to True?

  • sentdex
    November 17, 2020

    where to get that cool coffee cup.

  • sentdex
    November 17, 2020

    How does one learn to make these codes without seeing tutorials or videos like SENtdex?

Write a comment