Questions tagged [tensorflow]

A Python library for deep learning developed by Google. Use this tag for any on-topic question that (a) involves tensorflow either as a critical part of the question or expected answer, & (b) is not just about how to use tensorflow.

Filter by
Sorted by
Tagged with
0
votes
0answers
9 views

Suitability of ML for coordinate/PCA translation

Forgive the newbie question, but hopefully someone can set me on the right path. I'm trying to asses the suitability of Machine Learning techniques to allow for the translation of PCA data generated ...
0
votes
0answers
8 views

Test scores are way lower than cross-validation scores

I split my Dataset with 80% of the data for training and 20% for the test in the context of a binary classification task with a very unbalanced dataset. On the training set I do a 3 folds ...
0
votes
0answers
10 views

Arbitrary threshold for sigmoid activation function for CNN binary classification?

I am classifying sentiment of reviews - 0 or 1 - using gensim Doc2Vec and CNN in ...
0
votes
0answers
9 views

Regression problem with a recursive equation, prediction of a parameter

Hello I would like to predict the parameter "alpha" from the recursive equation: y(j,k+1) = y(j,k) * alpha(j,0)+ x(j,k). Where j is the loop parameter ...
0
votes
0answers
12 views

How to use multiple inputs into a single output properly

I am building a Reddit post to upvotes predictor based on title, desc and post age and I’m wondering how to design a model with multiple inputs properly, this is a schema I have, what do you guys ...
0
votes
1answer
20 views

How do you combine an image and meta data as input in machine learning?

Well my question is really about text, but image is easier to understand. Let’s say I have some images and some meta data about the images and I want to train a model with the metadata and the image ...
1
vote
0answers
11 views

Accuracy in DNN

How does number of batch size and steps per epoch affects the accuracy of the model? Edit: Training and validation accuracy both. I trained a model using CNN where the accuracy is changing relatively ...
0
votes
1answer
24 views

What is the difference when importing tf.keras between Tensorflow 1.14 and Tensorflow 2.0?

I have trained a Deep Learning model with tf.keras. In particular, based on an open-source, I have customized my training data and tried to solve a Semantic ...
0
votes
1answer
24 views

Why is my Tensorflow training and validation accuracy and loss exactly the same and unchanging? [duplicate]

I am a beginner to CNN and using tensorflow in general. I have been referring to this image classification guide to train and classify my own dataset. I have 84310 images in 42 classes for the train ...
0
votes
0answers
15 views

Is there a way to find the most adapted NN?

I'm trying to build a NN that uses one or two time series to predict the value of another one, using history. For example, in the next graph : Blue is the input Orange is the predicted output Green ...
0
votes
0answers
14 views

Problems understanding linear regression model tuning in tf.keras [migrated]

I am working on the Linear Regression with Synthetic Data Colab exercise, which explores linear regression with a toy dataset. There is a linear regression model built and trained and one can play ...
1
vote
2answers
31 views

How many total numbers of filters are there in Conv2D?

I have a question regarding filters used in Conv2D. How many total numbers of filters are there? Max number of filters I used is 64. If it is possible to use as many numbers(like 128, 256, 512, 1024) ...
-1
votes
0answers
26 views

How does dilation affect the total ops in dilated convolution layers?

After studying dilated-convolution from Google's WaveNet work using diagram at page-2 (also used in their DeepVoice model), it seems to me that dilated-convolutions on a series of input samples like ...
-1
votes
0answers
7 views

Why does Nvidia consider waveglow model faster than Google's wavenet for text to speech?

I've spent a lot of time trying to understand the Google's WaveNet work (also used in their DeepVoice model), but am confused when comparing it to Nvidia's WaveGlow model. I'm referring to this ...
-1
votes
0answers
12 views

How to get number of operations or Flops in a model?

I have been studying the Nvidia WaveGlow model that's used by Nvidia in text-to-speech synthesis, but couldn't find any estimate of # of Flops in the inference for this model. One can easily print out ...
0
votes
1answer
34 views

Training a machine learning model to recognize sentiment in text

I am just getting started with TensorFlow and machine learning. I have watched the following demo video showing how TensorFlow can be used to train a model to recognize sentiment in the text. In the ...
2
votes
1answer
50 views

Weird behaviour in toy RNN (Keras, LSTM)

I'm trying to learn more about RNNs and I'm tackling a toy problem. I'm generating some data that has a pattern, two 1s followed by three 0s which keeps repeating infinitely without any noise. So my ...
0
votes
0answers
21 views

Why is my validation loss always larger than my epoch loss [duplicate]

I am training a CNN and it seems no that matter what I do my validation loss is always much greater than my training loss. To my understanding, this should mean my model is always overfitting my data ...
1
vote
1answer
34 views

How do you handle missing data in neural nets for LSTM or CONV1D?

I have a device that outputs signals from 26 channels. I have 4 sessions of data that gave out signals from 26 channels and on the 6th session, I lost 5 channels. I have trained my model with 1-5 ...
2
votes
0answers
14 views

Resize images before training object detection

I am training an object detector. I didn't resize my image before labeling because the of assumption that the model does this automatically to fit its input shape. ...
0
votes
1answer
34 views

Why does one need Google's WaveNet model to generate audio if it already takes audio as input? [closed]

I've spent a lot of time trying to understand the Google's WaveNet work (also used in their DeepVoice model), but still confused about some very basic aspects. I'm referring to this Tensorflow ...
0
votes
0answers
11 views

NAN loss while training a image segmentation model with non-object images

I am currently working on a multi-class image segmentation application. A fraction of dataset contains images whose corresponding ground-truth images do not contain any object (completely black ...
1
vote
1answer
28 views

Mismatching dimensions of input/output in the WaveNet model for text-to-speech generation?

I have been trying to understand the model of how speech generation works, particularly in WaveNet model by Google. I was referring to the original WaveNet paper and this implementation: I find the ...
0
votes
0answers
9 views

Training an LSTM on multiple distinct batches of time series data

I am running a time series simulation on an electricity power grid simulation package and I want to use this data to train an LSTM to predict the stability of the grid over a given time interval. My ...
0
votes
0answers
26 views

Most efficient way to apply many 2d convolutions

I have two $4$ dimensional tensors, where the first two indices correspond to some fixed values in my problem, and the last two specify a 2D distribution. For each fixed value of the first two indices,...
0
votes
0answers
14 views

How to understand the dilated conv1d layers dimensions in this model?

I was trying to see the layers used in a Wavenet model for speech generation and I can't seem to make sense of the output layers printed by the TF model. Model is this: https://github.com/Rayhane-...
0
votes
1answer
24 views

Resume training with 'best model parameters' keras

I have been using the Keras callback EarlyStopping to stop my model once the validation error has stopped decreasing. There's an option ...
0
votes
1answer
23 views

Keras single hidden layer model outputting different results given same input?

I'm trying to train a binary classifier with Keras. This is my model: ...
0
votes
0answers
29 views

Approximating a distribution of functions instead of a single function?

It is easy for a neural network to learn to approximate an analytic function such as f(x) = x^2 or f(x) = x^2 + 1. I am ...
1
vote
0answers
18 views
+100

Parallelization strategies for deep learning

When training large neural networks on large datasets, there are several ways of breaking down the problem across machines and cores within a machine for parallel computation. To my knowledge, one ...
0
votes
0answers
4 views

Validation Error less than training error? custom metrics affected from droputs?

I have a neuronal network trained on some data. My testing loss is less than my training loss. As this question is well answered regarding some points here I ask myself, if a custom metric that is ...
1
vote
1answer
17 views

Conceptual question on Image semantic/ instance segmentation networks

I am trying to understand better how the image semantic/ instance segmentation work. I understand going from the concept of the perceptron that Deep Neural Networks have one or both of the following: ...
0
votes
0answers
43 views

Validation loss diverging when training a simple CNN for text classification

I'm training a CNN for text classification on the IMDb movie reviews dataset. The dataset contains 25000 training and 25000 testing samples of movie reviews, each half positive and half negative. The ...
1
vote
0answers
10 views

Tfp.sts unexpected wights prior behavior

I am trying to build a time series structural model and get the coefficients of the external factor. here is the structure of the model ...
1
vote
1answer
36 views

How to understand network neural network architecture from a research paper

Hello everyone I have the following architecture from the DELP-DAR research paper (https://www.sciencedirect.com/science/article/pii/S0167865519303216) and I dont really understand two things, first ...
1
vote
1answer
25 views

Validation and Test accuracy at random performance, whereas Train accuracy very high

I am trying to build a classifier in TensorFlow2.1 for CIFAR10 using ResNet50 pre-trained over imagenet from keras.application and then stacking a small FNN on top of it: ...
1
vote
1answer
26 views

Combine ReLU with TanH is a good idea?

I have a CNN implementation for the Generator of a GAN, internally, the architecture is using ReLU for non-linearities, but at the output, the paper of the architecture specifies Tanh must be used. ...
3
votes
1answer
97 views

Variational autoencoder with L2-regularisation?

I have built a variational autoencoder (VAE) with Keras in Tenforflow 2.0, based on the following model from Seo et al. (link to paper here). The VAE is used for image reconstruction. Note that the ...
0
votes
0answers
18 views

Should the lambda for L1 norm regularizer inversely be proportional to the number of trainable weights?

Say I want to implement Conv2D in keras and for each Conv2D layer, if I apply 20 filters of [2,3] filter on an input with depth of 10, then there will be 20*(2*3*10+1) = 1220 trainable weights. the ...
2
votes
1answer
56 views

Neural Network Loss Function for Predicted Probability

I am building a neural network (using tensorflow/keras) that attempts to classify into one of 5 categories (0, 1, 2, 3, 4). I have been using sparse categorical cross-entropy as my loss function. This ...
0
votes
0answers
10 views

How to improve neural network training against a large data set of points with varying magnitude

I am currently using TensorFlow and have simply been trying to train a neural network directly against a large continuous data set, e.g. $y = [0.014, 1.545, 10.232, 0.948, ...]$. The loss function in ...
0
votes
1answer
20 views

Finding patterns in binary files using deep learning

I am a newbie in deep learning and wanted to know if the problem I have at hand is a suitable fit for deep learning algorithms. I have thousands of fragments each of about 1000 bytes size (i.e. ...
0
votes
2answers
48 views

Manipulate keras multiple loss

Lets assume that we have a model model_A and we want to build up a backpropagation based on 3 different loss functions. The first loss (...
1
vote
0answers
10 views

Is Determinism important for Hyperparameter Tuning?

When training the Model on GPU, different results are retrieved for the same hyperparameters. This effect can be shut down by using CPU or Tensorflow 2.1. with deterministic settings. The Post on ...
1
vote
0answers
6 views

How to build a sentence classifier with tensorflow, that has two sub bi-lstms one for sentence embedding and other for sentence classification?

I want to build a model that takes a document, creates sentence embedding for each sentence using a bi-LSTM, then use the sequence of sentences embedding as input to another bi-LSTM that outputs a ...
0
votes
0answers
24 views

How do I apply Min Max scaling for numerical forecast when both dependent and independent volumes are increasing over time?

I'm want to build a numerical regression to forecast. From my initial analysis, it shows linear models (glm) out performs the typical decision tree models (xgboost, ranger...etc). I hypothesized that ...
1
vote
1answer
41 views

Masked Autoencoder MADE implementation in TensorFlow vs Pytorch

I am following the course CS294-158 [1] and got stuck with the first exercise that requests to implement the MADE paper (see here [2]). My implementation in TensorFlow [3] achieves results that are ...
1
vote
0answers
24 views

Keras Neural Network: use loss from one output in the loss of the other output [closed]

I would like to use the loss from one of my NN auxiliary outputs as part of the loss for the other output. $L($total$) = L($main output$) - \lambda L($auxiliary output$)$ I am unsure how to access ...
0
votes
0answers
6 views

Where the embeddings should be implemented in the RNN model?

Hi All (it's my first question here so welcome everyone), I wrote simple RNN model in tensor flow and I cannot figure out where the embeddings should be inserted inside, please find my code and below ...
0
votes
1answer
25 views

Building a neural network with two training paths in Keras

I am trying to build a NN in Keras with two different output paths where the first path informs the second. The first path passes its loss to the end of the second path, like so: Pass through layer A ...

1
2 3 4 5
12