Questions tagged [neural-networks]

Questions involving defining, training, executing, importing and exporting neural networks.

Filter by
Sorted by
Tagged with
3
votes
0answers
110 views

Can you train ImageClassify model with multiple GPUs correctly?

After some testing, I still cann't train with two GPU in Windows 10. But I can train with Mathematica 11.3 in Ubuntu 16.04 with two graphics card GTX 1080 The problem is, If I use two card to train, ...
6
votes
1answer
95 views

Using LearningRateMultipliers in a recurrent neural network

I'm trying to use LearningRateMultipliers in a recurrent neural network, but I haven't found the right way yet. An example of the code is as follows: ...
15
votes
2answers
964 views

How can I add noise to a neural network input?

A common "trick" in neural network training is to train an autoencoder using a "degraded" sample as input and the original sample as target output. The layer type ...
3
votes
1answer
83 views

How to NetTrain only N batches?

I'd like to train a neural network for only a specified number of batches, that is for only part of an epoch (or TrainingRound). I've tried using the ...
0
votes
0answers
79 views

Basic Feed Forward Network

I am trying to build a simple neural network, with an input layer of 250 nodes, 2 'hidden' layers each of 100 nodes, and an output layer with 1 node. I want each node to take in the linear ...
3
votes
1answer
136 views

How to modify NetTrain's error calculation

I'm training a neural network at the moment on an image segmentation problem (that is, I have an images and masks, and I am trying to train the network to generate its own masks from other images. I'...
3
votes
1answer
82 views

How to build a Neural Network with only categorical variables?

I've built mixed data NN, but I'm trying to build an NN that only has categorical variables. In my problem, I cannot represent my categorical variables using the commonly used ...
5
votes
2answers
257 views

How to make similar objects have similar embedding weights?

In the read world, there are many many phones(aa ae ah ao ar aw ax ay b ch d dh ...). But for brief,I only use three phones(aa ah ae) to illustrate my question.And it only need training Data,no Test ...
3
votes
1answer
99 views

How to reconstruct the result of FeatureExtraction?

For the below code for Mathematica 11.3, I got {-0.153792, 1., 1.} as output. ...
0
votes
1answer
88 views

Problem of Low Accuracy of NeuralNetwork Method in Classify MNIST 11.3

The accuracy is extremely low... even after changing a lot of options of Method... Other methods' result is fine and normal, 0.80 accuracy like LogisticRegression. ...
2
votes
0answers
77 views

Why is the output of the Embedding layer so non-linear?

I have 2 nets, i.e. net1 (i.e. LinearLayer) and net2 (i.e. ...
2
votes
1answer
86 views

Custom NetEncoder

I've been trying to train an RNN on some data that I pre-processed externally and saved to disk. My dataset is too large to fit in memory and is additionally varying in shape. Looking for a way to do ...
5
votes
1answer
94 views

Why would a symmetric image have an asymmetric PoolingLayer-Gradient Image?

I have a symmetric image img. For simplicity, I would use the below image as my example: ...
4
votes
0answers
84 views

In neural network, how to force the NetTrain to use full batch gradient descent instead SGD?

In a neural network, I want to use full batch gradient descent instead of SGD. Currently, I am using code like this (you can see that I forced BatchSize to be the ...
4
votes
1answer
111 views

How to apply the layer of sigmoid function in neural network?

I saw an article How to build a simple neural network in 9 lines of Python code and want to reconstruct the result in Mathematica. I tried the below codes:- ...
2
votes
1answer
108 views

CloudDeploy broken with neural net (new)

over the last few weeks I routinely deployed neural nets to the cloud for computation. Today, even the simplest example does not work anymore: ...
11
votes
1answer
941 views

Neural Networks in Predict function

I'm trying to reproduce the result giving by a Predict function with a Neural Network as the chosen method. So, I'm training from this set: ...
3
votes
1answer
128 views

Extract neural network model from ClassifierFunction

I'm very new to Neural Networks and Mathematica 11 (I used MMA8 some time ago) and I'm trying to learn about NN using Mathematica. I'm using MMA 11.2 in a Rapsberry Pi (yes, I know that doing NN in a ...
3
votes
0answers
45 views

Layer states of NetBidirectionalOperator

I want to define a NetBidirectionalOperator version of this network, how do I access states of forward and backward net with ...
11
votes
2answers
585 views

How to combine Neural Network and Audio like Classify?

I want to combine Audio object and Neural Network ,first thing I remember is Classify ...
2
votes
1answer
100 views

What does NetMapOperator do?

Recently, I have become addicted to Mathematica's support for neural networks. As I learn those commands that are prefixed with "Net", I can't tell the difference between using ...
2
votes
0answers
99 views

Does Mathematica auto update something causes error load old-version model?

I've trained a NN model by NetTrain with Mathematica 11.3.0, and I can load the model correctly to predict. My model is saved by .mx format. After few months, I cann't load the model correctly. ...
4
votes
1answer
104 views

Deep Learning SummationLayer does not support varying dimensions on Linux server

I am trying a recurrent network in Linux server using MMA 11.3. The network structure is as follows: ...
4
votes
0answers
107 views

How to explain overfitting?

SeedRandom[0]; X = RandomInteger[{1, 5000}, {100000, 50}]; Y = RandomInteger[{0, 1}, 100000]; For net1 and ...
2
votes
1answer
101 views

NetTrain & SoftmaxLayer: Expected a vector of indices

Is SoftmaxLayer broken in version 11.3.0.0? I know, it's probably me. Here's a MWE of the issue. Note that this works fine if e.g. Exp is substituted for SoftmaxLayer[].
2
votes
1answer
103 views

How to convert Image to class index in NetEncoder

In neural network segmentation task, the training data usually consists of pairs of images. In each training sample, the input image is the image we want to segment, and the target image contains the ...
0
votes
1answer
166 views

How do I change activation function of LinearLayer[] in NetGraph

I want to analize a network such that all neurons of its hidden layers output LogisticSigmoid[w.x+b], which is one of the most common ouputs analized in a neuron. ...
1
vote
1answer
65 views

Issue fine-tuning a simple embedding network

I'm trying to learn an embedding on Cifar-10 by finetuning ResNet-50 with a few added layers: ...
44
votes
4answers
3k views

Generative Adversarial Network

Generative adversarial networks (GAN) is regarded as one of "the most interesting idea in the last ten years in machine learning" by Yann LeCun. It can be used to generate photo-realistic images that ...
3
votes
1answer
118 views

Associations & generating training sets

Solution: Pi is allowed in training sets, but not in batches generated by the batch generating function preamble: I want to generate my own batches of training data for training a deep CNN. ref/...
4
votes
1answer
264 views

Detecting a pattern in a time series with a Neural Network

I new to working with Neural Networks and I want to detect a pattern in a time series and return some measure of the certainty/uncertainty when a pattern has been detected: ...
9
votes
3answers
745 views

Example of NetChain running in MXNet cause Error

Cross post here Problem remains in Mathematica 11.2 First defining the net ...
8
votes
1answer
215 views

Dilation on 3D Images?

I need to apply a dilated convolution to a 3D image, but dilation values greater than 1 are not currently supported in v11.3: ...
4
votes
2answers
186 views

Quality of ImageRestyle

There is an application of image restyle here. It is an example in following: But if I just use ImageRestyle of Mathematica <...
5
votes
1answer
232 views

Error Import a MXNet model Export by Mathematica in 11.2 and 11.3

the model was trained by net = NetModel["Inception V1 Trained on ImageNet Competition Data"] ...
4
votes
0answers
144 views

YOLOv3 Netmodel?

I see that YOLOv2 is on the Wolfram net repo, but has anyone ported the new YOLO version 3 architecture into Mathematica?
3
votes
0answers
108 views

Can someone help me to understand sequence-to-sequence transformation using (Wolfram-specific) recurrent neural nets?

I've been learning about recurrent neural networks lately and I think I'm starting to get the basic idea of how they work. I'm particularly interested in the sequence transformation capabilities of ...
10
votes
0answers
444 views

How to plug holes in an Image (with partial convolutions)?

NVIDIA recently published a new way to reconstruct images with holes. They use a new type of convolutional layer which is described in the paper Image Inpainting for Irregular Holes Using Partial ...
3
votes
1answer
165 views

Tutorial on net surgery?

I'm trying to replace the "fc1000" layer in a NetModel, what am I missing here: ...
11
votes
1answer
396 views

How to import Neural Networks from ONNX?

ONNX is a open format to represent deep learning models. So I want to import neural networks from other frameworks via ONNX. And the Mathematica 11.3 supports ...
3
votes
1answer
130 views

How exactly does NetTrain use a ValidationSet (can we reuse it as TestSet)?

Typically when you train a model you have three sets of data: training, validation, and testing - and they can’t be mixed, i.e. you can’t use images from one in the other. I want to know if the ...
9
votes
1answer
213 views

NetEncoder for images of “Varying” dimension?

I'm looking for someway to do what this syntax "should" do: NetEncoder[{"Image", Automatic, ColorSpace -> "RGB"}] /@ imgs; The documentation states that ...
1
vote
1answer
99 views

Specify NetEncode's tokenization model?

NetEncoder[{"Tokens"}] uses WordData indices to tokenize an input string. But is there a way to tell ...
18
votes
4answers
917 views

How to use Mathematica to train a network Using out of core classification?

I see there is doc about how to train a network Using out of core image classification and this question.But the object is only image. I want to use a binary file as data(Sequence to Sequence case),...
3
votes
1answer
66 views

Validation failed for ConvolutionLayer message?

I have a simple fully convolutional network and I'm trying to figure out why I get this error NetChain::valfail: Validation failed for ConvolutionLayer: kernel size 3*3 plus padding size 0*0 cannot ...
12
votes
2answers
1k views

How to implement neural style transfer?

In the paper A Neural Algorithm of Artistic Style, Leon A. Gatys et all. describe a process of extracting the content and style vectors from an image separately. They use this technique to generate ...
3
votes
1answer
130 views

how to use CatenateLayer to connect nodes within a NetChain

EDIT: Looking closely at my network I have a similar problem of providing inputs to the CropLayer function. The output from the pooling layer (node 30) and node 87 (activation) should feed to the ...
3
votes
1answer
39 views

How to replace all activations in a NetGraph?

I'm trying to replace all activation layers of a certain type, there must be a simpler way to do this? ...
86
votes
1answer
12k views

Crack CAPTCHA using deep learning

Deep convolutional neural networks are very good at computer vision related tasks. Optical Character Recognition (OCR) is one important branch of computer vision. In fact, the convolution neural ...
2
votes
0answers
59 views

Unable to create ClassifierMeasurements for NetGraph

I am trying to check the accuracy of a neural network by using ClassifierMeasurements. I take the MNIST as example:- Code for setting ...