Questions tagged [neural-networks]
Questions involving defining, training, executing, importing and exporting neural networks.
354 questions
0
votes
0answers
22 views
Unable to replicate Predict/NeuralNetwork example from Help Pages
I am trying to replicate an example right out of the help manual (ref/method/NeuralNetwork). This example compares the convolutional NN versus a fully-connected NN....
-1
votes
1answer
78 views
Neural network to classify gender from username
Does anyone know of a trained net that can classify gender from username from any website? It would be nice if there is something in Wolfram Neural Net Repository or something that can be easily ...
0
votes
1answer
36 views
AttentionLayer what is it and why would it be used
Could anyone elaborate on the new Neural Network layer in Mathematica Verison 12:
https://reference.wolfram.com/language/ref/AttentionLayer.html
AttentionLayer
The documentation is rather sparse on ...
2
votes
0answers
48 views
Scalar vs vector of size 1 in neural networks
I've been playing with Mathematica's neural network functions, and I keep getting stuck on the same annoying problem. Often Mathematica declares that the output of my neural network is a vector of ...
10
votes
0answers
87 views
+100
Implementing Contextual Bandits in version 12
The multi-armed bandit problem (MAB) is a classic in Reinforcement learning. Now that mathematica version 12 finally has support for Reinforcement learning:
Here's a nice "mnist example" for training ...
8
votes
1answer
173 views
Is Mathematica's implementation of convolutional neural networks really inefficient?
I'm surprised how slow ConvolutionLayer works with TargetDevice->"CPU". For instance this code:
...
1
vote
1answer
63 views
Random sampling of a discrete distribution in NetGraph
Within a NetGraph I would like to use the output of a softmax layer, interpreted as a probability distribution, to sample a "next token". Since this is done inside an RNN structure, I cannot do this ...
2
votes
0answers
50 views
Problem with Mathematica 12.0.0.0 (NetTrain on a UNet with a NVidia Quadro P5200 GPU)
I am trying to train a UNet on my laptop (Lenovo P72) and I obtain the following error when I use NetTrain on it:
CompiledFunction::cfta: Argument {Mean[LibraryFunctionError[LIBRARY_FUNCTION_ERROR,6]]...
3
votes
0answers
38 views
Can I extract symbolic expression of neural network loss function?
Once we create a neural network with NetChain in Mathematica, can we extract the loss function in the symbolic form for Mathematica to play with symbolic manipulations using Mathematica's built-in ...
3
votes
1answer
63 views
How do I create my own activation function in NetChain in Mathematica?
I want to have my own activation function, say a quadratic function, instead of Tanh, or Sigmoid etc. in Mathematica. How do I insert it in NetChain?
e.g., how will the below code look for a new ...
0
votes
0answers
47 views
TargetDevice->“GPU” takes no effect in prediction of Classify and ClassifierMeasurements?
I can train the model with GPU via method->"NeuralNetwork" in Classify.
I'm on Mathematica 12.0 and the problem also happens in 11.3.
Is this a bug since 11.1 to 12.0?
In training, GPU is faster ...
8
votes
1answer
278 views
Bug in Mathematica 12 neural networks?
I recently downloaded Mathematica 12, and I'm experiencing a curious failure in neural networks that had earlier worked. This failure I will demonstrate with the NN in "Classifying Data with Neural ...
4
votes
0answers
49 views
How to extend ImageCases?
I was very excited to see ImageCases being documented in version 12. But I immediately desired a mechanism for adding more entities to ...
1
vote
0answers
38 views
5
votes
0answers
39 views
Compute neural network loss over sliding windows?
I am trying to use the structural similarity index, which is defined in terms of a computation over sliding windows, as a neural network loss function.
Obviously one can get a sliding mean using a <...
6
votes
1answer
112 views
5
votes
1answer
101 views
0
votes
0answers
46 views
Compatibility Nvidia and Mathematica 11.3 [duplicate]
I am wondering about the compatibility between Mathematica 11.3 (Windows or Linux) and the Nvidia Quadro P5200 provided in the ThinkPad P72.
Can I be sure that Mathematica will be able to detect and ...
1
vote
1answer
56 views
How can I have two neural network layers act as the input to one layer?
When using NetGraph[], I am try to specify the connections between 3 layers like so:
NetGraph[{...some layers...},{...{2,3}->4}]
But I receive the following ...
2
votes
1answer
51 views
Where to find information about models, data and labels used by Image/AudioIdentify?
I am thinking about using Mathematica 12's ImageIdentify and AudioIdentify functions for an interdisciplinary research project. The social science part of it requires to describe a little bit the ...
1
vote
0answers
14 views
Is there an inverse type function to SequenceLastLayer[]?
I am working on a text based autoencoder, and was happy to see I can feed a variable length text sequence into my network and use SequenceLastLayer[] after a GatedRecurrentLayer[] to only get the ...
1
vote
0answers
59 views
Is Wolfram Cloud going to support GPU training? [duplicate]
Is Wolfram Cloud going to support GPU training?
Is there any other possibility for training a neural net (in WL) on a cloud like AWS, Floydhub, or so?
I think this concerns lots of people.
2
votes
1answer
75 views
3D Deconvolutional Layer
I noticed that one of the mxnet APIs (Gluon) includes a 3D Deconvolutional Layer. However, mathematica currently only allows up to 2D deconvolutional layers (see here). Can one somehow implement a 3D ...
1
vote
0answers
50 views
Neural Network stick in an unlearned state
Background: I want to apply a net to mapping from xyY color space to Munsell. The mapping in freely downloadable as a 5000+ entries table. I transformed the table data so that the Munsell Hue is given ...
0
votes
0answers
62 views
Learning weights during training
Let's assume I have a net so the loss function is composed from 3 components: $C_1, C_2, C_3$, and each component is multiply by its weight $(w1, w2, w3)$.
The weights should be a function of $\beta$,...
4
votes
0answers
71 views
Visualizing and Understanding CNN with Mathematica
The famous 2013 paper from Zeiler and Fergus "Visualizing and Understanding Convolutional Networks" proposes a method to understand the behavior of CNN using one (or more) DeConv networks coupled with ...
3
votes
1answer
86 views
NetPortGradient returns “floating-point overflow” for this simple network
Consider this simple network that computes the following function:
$$ \left\{ \begin{matrix} xy & \mbox{if} \;x,y>0 \\ 0 & \mbox{otherwise} \end{matrix} \right.$$
The net looks like:
<...
47
votes
5answers
7k views
Mathematica command that allows it to read my intentions
In version 2.0 of Mathematica there was a command, ReadProgrammerIntentions[ ] which allowed it rapidly compose code that could solve problems simple to complex, ...
0
votes
0answers
21 views
Replace part of a net with NetReplacePart with another net with different number of outputs
I have a net with a nested sub-net like
...
4
votes
0answers
120 views
Transfer learning a bounding box detector
I'd like to add a new class to one of the standard object detection architectures. Consider the YOLO v2 detector from the Neural Net Repo.
...
4
votes
1answer
222 views
Creating a generator function for a large data set for a neural network
After reading documentation on Training on Large Data Sets, I am still at a loss on how to create a generator function, to read a batch of 1000 of this
...
4
votes
1answer
96 views
How to improve spelling in the Deep Speech NetModel?
There is only one NetModel for speech recognition. As is stated in the network repository, models trained with CTC loss have difficulties in producing correct spelling:
...
2
votes
1answer
27 views
Can I get the “Image Size”, etc. from a NetEncoder in an existing net from the Repository?
I want to get the details about the input used to train a net in the Net Repository, in my case "LeNet Trained on MNIST Data". I can get its decoder by:
...
0
votes
1answer
195 views
NetGraph/Model for BERT?
Does anyone know of a way to get the BERT model working for word embeddings in Mathematica? Doesn’t seem to be available on GitHub or the Wolfram Net Repo...
1
vote
1answer
27 views
Error NetTrain::invencin: Invalid input when training a neural network [closed]
I've build a Deep neural network with various variables:
...
9
votes
1answer
146 views
How do I implement masked Softmax?
Consider:
input = {1, 2, 3};
mask = {1, 0, 1}
Classic Softmax:
SoftmaxLayer[]@input
{0.0900306, 0.244728, 0.665241}
...
2
votes
1answer
73 views
How to provide an RNN various type of inputs?
I'm trying to build an RNN in Mathematica.
My window is only 2 but the variables are of different types:
...
4
votes
0answers
37 views
How to implement external memory for the “memory augmented neural network”?
I am trying to implement the memory-augmented neural network using MMA 11.3.
The memory-augmented neural network in based on the paper here. There are implementations using ...
0
votes
1answer
96 views
Robust face detection
FindFaces doesn't work when people are far away from the camera. I'm looking for a single accurate method of detecting one or more people in an images, at all ...
1
vote
1answer
68 views
12
votes
0answers
309 views
Matlab Neural Network Toolbox vs Wolfram Neural Network Framework [closed]
Can anyone who has worked with both Matlab and Mathematica provide some comparison of how the Matlab Neural Network Toolbox compare to that of Mathematica.
Mathematica as far as I understand is a ...
1
vote
0answers
49 views
Loading machine learning library without enabling my discrete Nvidia GPU? [closed]
I've asked a similar question
[here].
I cannot use some of the machine learning modules without enabling my discrete Nvidia GPU.
My setup:
Ubuntu 16.04 on a Dell XPS-15 9560 Intel i7 CPU with 8-...
1
vote
0answers
48 views
NetTrain with GPU crashed
I tried to run NetTrain with TargetDevice->"GPU", and Mathematica crashed. I got the following message:
When I executed ...
2
votes
0answers
48 views
“Learning” to parameterize Hermitian matrices with a basis using neural networks
I'm interested in using neural networks to "learn" how to write down Hermitian matrices, specifically those which are defined over spaces with a specific tensor product structure. The simplest example ...
8
votes
0answers
98 views
Slow down training many neural networks (Bug?)
I am working on something in which I need to train a neural network many times with slightly different training data. I noticed that after many training runs the input rate had dropped significantly.
...
1
vote
0answers
54 views
customized complex loss function to a neural network
My general question is how can I add a customized and complex loss function to neural network?
I have to address this issue ASAP, because I did not find any good examples that can do it. In the loss ...
3
votes
2answers
167 views
3
votes
1answer
54 views
Specify input ports with NetGraph for network within a network
I have a network with two input ports ("a" and "b").
I want to use this net within a bigger network. The problem is that I cannot find a syntax to specify to which input port the previous layers of ...
1
vote
1answer
63 views
Neural network not training due to function not loading
I'm trying to train a neural network on Mathematica 11.3,
but NetTrain won't work and gives the follwoing error:
...
10
votes
1answer
285 views
Creating an indicator function in a neural network
I am working on implementing the loss function of the Yolo version 2 object detection network and building from the construction notebook in the Wolfram Neural Network repository Yolo Version 2 so I ...
