# Machine Learning

**URL:** https://discourse.julialang.org/c/domain/ml/24.md?page=26

[Latest](https://discourse.julialang.org/latest.md) · [Categories](https://discourse.julialang.org/categories.md) · [Tags](https://discourse.julialang.org/tags.md)

**Page:** 27

---

## [Data augmentation for audio](https://discourse.julialang.org/t/data-augmentation-for-audio/31117)

<div class="topic-metadata">

**Author:** [@baggepinnen](https://discourse.julialang.org/u/baggepinnen)\
**Replies:** 5\
**Last updated:** [February 25, 2022, 5:59am UTC](https://discourse.julialang.org/t/data-augmentation-for-audio/31117 "2022-02-25T05:59:32Z")

</div>

Hey! We are likely to embark on an adventure to create something very similar to @Evizero’s excellent package Augmentor.jl, but for environmental acoustic recordings. I think the overall idea of Augmentor.jl looks solid…

---

## [How to implement a Gaussian NN with covariance matrix?](https://discourse.julialang.org/t/how-to-implement-a-gaussian-nn-with-covariance-matrix/76704)

<div class="topic-metadata">

**Author:** [@HenriDeh](https://discourse.julialang.org/u/HenriDeh)\
**Replies:** 7\
**Last updated:** [February 23, 2022, 11:28am UTC](https://discourse.julialang.org/t/how-to-implement-a-gaussian-nn-with-covariance-matrix/76704 "2022-02-23T11:28:56Z")

</div>

Hello, I’m editing this question to make it clearer since I got no solution still :frowning:. I need to implement a NN that outputs a covariance matrix. Say we have N variables, I need to be able to compute a symetric …

---

## [What does MLJ.save really save?](https://discourse.julialang.org/t/what-does-mlj-save-really-save/74606)

<div class="topic-metadata">

**Author:** [@wc4wc4wc4](https://discourse.julialang.org/u/wc4wc4wc4)\
**Replies:** 12\
**Last updated:** [February 23, 2022, 8:31am UTC](https://discourse.julialang.org/t/what-does-mlj-save-really-save/74606 "2022-02-23T08:31:17Z")

</div>

I am running a MLJ pipeline on a quite large dataset and I want to be able to save the model once, and then use it in later sessions without having to retrain it. I have tried following the documentation regarding saving…

---

## [Segmentation fault when applying PCA to a big dataset on an HPC](https://discourse.julialang.org/t/segmentation-fault-when-applying-pca-to-a-big-dataset-on-an-hpc/76900)

<div class="topic-metadata">

**Author:** [@Ivan](https://discourse.julialang.org/u/Ivan)\
**Replies:** 2\
**Last updated:** [February 22, 2022, 7:15pm UTC](https://discourse.julialang.org/t/segmentation-fault-when-applying-pca-to-a-big-dataset-on-an-hpc/76900 "2022-02-22T19:15:11Z")

</div>

I understand the following may be due to the fact that I am using an HPC remote service. If that’s the case, I would appreciate comments on where to ask this question best. I have a big array (2765 x 12212224), and I wa…

---

## [Boruta algorithm](https://discourse.julialang.org/t/boruta-algorithm/76821)

<div class="topic-metadata">

**Author:** [@Jaidy](https://discourse.julialang.org/u/Jaidy)\
**Replies:** 4\
**Last updated:** [February 21, 2022, 11:46am UTC](https://discourse.julialang.org/t/boruta-algorithm/76821 "2022-02-21T11:46:00Z")

</div>

Hi everyone, I am looking to use Boruta algorithm to select features for my project. First I looked up for a Julia implementation of Baruta algorithm, but could not find any. Then I found Python implementation and tried…

---

## [Flux cpu() type stability](https://discourse.julialang.org/t/flux-cpu-type-stability/76610)

<div class="topic-metadata">

**Author:** [@RobertGregg](https://discourse.julialang.org/u/RobertGregg)\
**Replies:** 3\
**Last updated:** [February 18, 2022, 11:02pm UTC](https://discourse.julialang.org/t/flux-cpu-type-stability/76610 "2022-02-18T23:02:37Z")

</div>

While reviewing a loss function I wrote, I noticed that the cpu() function from Flux produced type instability: julia\> using Flux julia\> @code\_warntype cpu(rand(10)) MethodInstance for Flux.cpu(::Vector{Float64}) fro…

---

## [Iterated Random Search Methods?](https://discourse.julialang.org/t/iterated-random-search-methods/76636)

<div class="topic-metadata">

**Author:** [@Satvik](https://discourse.julialang.org/u/Satvik)\
**Replies:** 2\
**Last updated:** [February 17, 2022, 5:16pm UTC](https://discourse.julialang.org/t/iterated-random-search-methods/76636 "2022-02-17T17:16:46Z")

</div>

Hi, I’m trying to optimize a non-differentiable function with a mixture of discrete and continuous parameters. The function takes a long time to run, but we can run many batches of it at a time on separate machines, so …

---

## [How to force Flux to use FiniteDiff](https://discourse.julialang.org/t/how-to-force-flux-to-use-finitediff/76324)

<div class="topic-metadata">

**Author:** [@BMval](https://discourse.julialang.org/u/BMval)\
**Replies:** 16\
**Last updated:** [February 16, 2022, 10:48pm UTC](https://discourse.julialang.org/t/how-to-force-flux-to-use-finitediff/76324 "2022-02-16T22:48:47Z")

</div>

I have loss() function. FiniteDiff.finite\_difference\_derivative - works great with the this loss() ForwardDiff.derivative - doesn’t work at all. As result Flux.train!(loss,Flux.params(NNODE),data,ADAM(0.001)) doesn’t …

---

## [Neural ODE minibatch error with multi-dimension input condition](https://discourse.julialang.org/t/neural-ode-minibatch-error-with-multi-dimension-input-condition/76121)

<div class="topic-metadata">

**Author:** [@lindblot](https://discourse.julialang.org/u/lindblot)\
**Replies:** 5\
**Last updated:** [February 16, 2022, 9:07pm UTC](https://discourse.julialang.org/t/neural-ode-minibatch-error-with-multi-dimension-input-condition/76121 "2022-02-16T21:07:11Z")

</div>

Hello - I’m trying to add minibatching to a problem discussed in https://discourse.julialang.org/t/neuralode-layer-strategy/71297/4. I’ve successfully refactored the code to predict either a single u0 or matrix of u0s, …

---

## [Problems on training LSTM](https://discourse.julialang.org/t/problems-on-training-lstm/76471)

<div class="topic-metadata">

**Author:** [@Louis612](https://discourse.julialang.org/u/Louis612)\
**Replies:** 2\
**Last updated:** [February 15, 2022, 11:02pm UTC](https://discourse.julialang.org/t/problems-on-training-lstm/76471 "2022-02-15T23:02:44Z")

</div>

I’m trying to use LSTM to forecast the trend of Covid-19. Code is below: using MAT using Plots using Measures using Flux using DifferentialEquations using DiffEqFlux using LaTeXStrings #data importing vars = matread("")…

---

## [AD with respect to Flux parameters](https://discourse.julialang.org/t/ad-with-respect-to-flux-parameters/76468)

<div class="topic-metadata">

**Author:** [@raktim](https://discourse.julialang.org/u/raktim)\
**Replies:** 1\
**Last updated:** [February 15, 2022, 5:11am UTC](https://discourse.julialang.org/t/ad-with-respect-to-flux-parameters/76468 "2022-02-15T05:11:17Z")

</div>

I am trying to compute gradient of a function with respect to Flux parameters. Example: using Flux, ForwardDiff function f(x::Vector) return \[x\[2\]; -x\[1\] + (1 - x\[1\]^2) \* x\[2\]\] end # Function approximation n = 2…

---

## [Problems regarding an implementation of a word2word Autoencoder](https://discourse.julialang.org/t/problems-regarding-an-implementation-of-a-word2word-autoencoder/76255)

<div class="topic-metadata">

**Author:** [@Arcumenn](https://discourse.julialang.org/u/Arcumenn)\
**Replies:** 7\
**Last updated:** [February 14, 2022, 6:14pm UTC](https://discourse.julialang.org/t/problems-regarding-an-implementation-of-a-word2word-autoencoder/76255 "2022-02-14T18:14:25Z")

</div>

Hello there! I am currently trying to implement a word-2-word Autoencoder with the Flux.jl package. It went mostly well, but I’m running into some problems now and I hope someone might be able to nudge me into the righ…

---

## [Problems on Flux.destructure()](https://discourse.julialang.org/t/problems-on-flux-destructure/76398)

<div class="topic-metadata">

**Author:** [@Louis612](https://discourse.julialang.org/u/Louis612)\
**Replies:** 14\
**Last updated:** [February 14, 2022, 9:44am UTC](https://discourse.julialang.org/t/problems-on-flux-destructure/76398 "2022-02-14T09:44:36Z")

</div>

As the figure shown above, I build a LSTM using the code "Chain(LSTM(1,1),LSTM(1,1)) " Then I want to build the network with specific parameters p. code re(p) but it failed and showed above. How to deal with this i…

---

## [Turing, Zygote, and Interpolations](https://discourse.julialang.org/t/turing-zygote-and-interpolations/76371)

<div class="topic-metadata">

**Author:** [@gideonsimpson](https://discourse.julialang.org/u/gideonsimpson)\
**Replies:** 1\
**Last updated:** [February 14, 2022, 3:16am UTC](https://discourse.julialang.org/t/turing-zygote-and-interpolations/76371 "2022-02-14T03:16:07Z")

</div>

This is a follow up to a previous post I did, in which I thought everything was resolved, but it is not. I want to use Turing with both FFTW and some type of interpolation method. I am trying something of the form: usi…

---

## [Errors when learning a function of a Jacobian. What am I doing wrong?](https://discourse.julialang.org/t/errors-when-learning-a-function-of-a-jacobian-what-am-i-doing-wrong/76352)

<div class="topic-metadata">

**Author:** [@ynsch](https://discourse.julialang.org/u/ynsch)\
**Replies:** 0\
**Last updated:** [February 13, 2022, 2:23pm UTC](https://discourse.julialang.org/t/errors-when-learning-a-function-of-a-jacobian-what-am-i-doing-wrong/76352 "2022-02-13T14:23:20Z")

</div>

Hello, I am working on a simple example with Flux and Zygote, mainly translating this notebook to Julia. The loss function contains the divergence (trace of the Jacobian) and Zygote gives an error when updating the mod…

---

## [Why my NN overfitting?](https://discourse.julialang.org/t/why-my-nn-overfitting/76333)

<div class="topic-metadata">

**Author:** [@sherlock](https://discourse.julialang.org/u/sherlock)\
**Replies:** 1\
**Last updated:** [February 13, 2022, 9:27am UTC](https://discourse.julialang.org/t/why-my-nn-overfitting/76333 "2022-02-13T09:27:58Z")

</div>

Here is the Julia code, with 2 hidden layers, (784, 40) and (40, 10). Using relu as activation function. using Flux using MLDatasets using ImageCore using Distributions p = Binomial(1, 0.5) one\_hot\_encoding(x) = uniqu…

---

## [DiffEqFlux does not expose sciml\_train](https://discourse.julialang.org/t/diffeqflux-does-not-expose-sciml-train/76172)

<div class="topic-metadata">

**Author:** [@AlexG](https://discourse.julialang.org/u/AlexG)\
**Replies:** 2\
**Last updated:** [February 10, 2022, 4:47pm UTC](https://discourse.julialang.org/t/diffeqflux-does-not-expose-sciml-train/76172 "2022-02-10T16:47:44Z")

</div>

I’m trying to get DiffEqFlux.jl 1.4.5 running on Julia 1.6.5 (Mac x86). When running the Lotka-Volterra example in the docs verbatim, it throws the following error: UndefVarError: sciml\_train not defined Stacktrace: \[…

---

## [Machine Learning using Julia - Aim/Idealogy of Flux.jl to for simplicity over compexity for programmers](https://discourse.julialang.org/t/machine-learning-using-julia-aim-idealogy-of-flux-jl-to-for-simplicity-over-compexity-for-programmers/75602)

<div class="topic-metadata">

**Author:** [@arcAman07](https://discourse.julialang.org/u/arcAman07)\
**Replies:** 11\
**Last updated:** [February 8, 2022, 1:51pm UTC](https://discourse.julialang.org/t/machine-learning-using-julia-aim-idealogy-of-flux-jl-to-for-simplicity-over-compexity-for-programmers/75602 "2022-02-08T13:51:27Z")

</div>

I had one question while learning Flux.jl library, what is the reason behind the minimal complexity/implementation ( less things done behind the hood ) by Flux.jl? Why isn’t Flux aiming to be something like tensorfl…

---

## [How to use Python library in Julia?](https://discourse.julialang.org/t/how-to-use-python-library-in-julia/75644)

<div class="topic-metadata">

**Author:** [@YitaoCai](https://discourse.julialang.org/u/YitaoCai)\
**Replies:** 9\
**Last updated:** [February 5, 2022, 5:42pm UTC](https://discourse.julialang.org/t/how-to-use-python-library-in-julia/75644 "2022-02-05T17:42:24Z")

</div>

I was wondering how can I use this package https://github.com/DeepGraphLearning/ConfGF in Julia by using PyCall.jl or PythonCall.jl? this is how to use this package in python. I want to do it in Julia, is it feasible? …

---

## [Pytorch post on taking inspiration from julia](https://discourse.julialang.org/t/pytorch-post-on-taking-inspiration-from-julia/75791)

<div class="topic-metadata">

**Author:** [@bjarthur](https://discourse.julialang.org/u/bjarthur)\
**Replies:** 0\
**Last updated:** [February 4, 2022, 1:03pm UTC](https://discourse.julialang.org/t/pytorch-post-on-taking-inspiration-from-julia/75791 "2022-02-04T13:03:55Z")

</div>

---

## [Parameter Estimation of HPV Disease Model ( DiffEqFlux ) (beginner issue)](https://discourse.julialang.org/t/parameter-estimation-of-hpv-disease-model-diffeqflux-beginner-issue/75758)

<div class="topic-metadata">

**Author:** [@Mevlut\_Cayli](https://discourse.julialang.org/u/Mevlut_Cayli)\
**Replies:** 1\
**Last updated:** [February 4, 2022, 12:41am UTC](https://discourse.julialang.org/t/parameter-estimation-of-hpv-disease-model-diffeqflux-beginner-issue/75758 "2022-02-04T00:41:18Z")

</div>

Hi, I 'm trying to estimate HPV transmission model parameters with DiffeqFlux but seems like it gets stuck in local optima and I’m not sure about using mini batching because my observation data just has length of 6 in …

---

## [Function for creating a neural network with n hidden layers in Flux](https://discourse.julialang.org/t/function-for-creating-a-neural-network-with-n-hidden-layers-in-flux/75589)

<div class="topic-metadata">

**Author:** [@Handam](https://discourse.julialang.org/u/Handam)\
**Replies:** 1\
**Last updated:** [February 1, 2022, 6:08pm UTC](https://discourse.julialang.org/t/function-for-creating-a-neural-network-with-n-hidden-layers-in-flux/75589 "2022-02-01T18:08:18Z")

</div>

Hi, I want to create a function, where I can set the number of hidden layers with an argument. One option that kinda works is using Flux function make\_nn(length\_in = 80, length\_out = 1, nodes = 128, hid\_lay = 3, act\_fu…

---

## [Passing gradient back from remote server](https://discourse.julialang.org/t/passing-gradient-back-from-remote-server/41150)

<div class="topic-metadata">

**Author:** [@Jimmie\_Butler](https://discourse.julialang.org/u/Jimmie_Butler)\
**Replies:** 0\
**Last updated:** [June 10, 2020, 5:27pm UTC](https://discourse.julialang.org/t/passing-gradient-back-from-remote-server/41150 "2020-06-10T17:27:14Z")

</div>

I’m trying to pass a gradient through a remote call with Zygote. The server seems to be passing the gradient back properly, the server decodes and returns it from the adjoint, but the final gradient being returned is c…

---

## [Hessian Vector Products on GPU using ForwardDiff, Zygote, and Flux](https://discourse.julialang.org/t/hessian-vector-products-on-gpu-using-forwarddiff-zygote-and-flux/75306)

<div class="topic-metadata">

**Author:** [@RS-Coop](https://discourse.julialang.org/u/RS-Coop)\
**Replies:** 13\
**Last updated:** [January 31, 2022, 2:28pm UTC](https://discourse.julialang.org/t/hessian-vector-products-on-gpu-using-forwarddiff-zygote-and-flux/75306 "2022-01-31T14:28:10Z")

</div>

Hello, I am trying to compute the Hessian vector product of a Flux model (with some vector) on the GPU. This is giving me some errors, which I think are realted specifically to ForwardDiff. Any help would be greatly app…

---

## [Optimizing performance of 2D nonlinear diffusion UDE](https://discourse.julialang.org/t/optimizing-performance-of-2d-nonlinear-diffusion-ude/72113)

<div class="topic-metadata">

**Author:** [@JordiBolibar](https://discourse.julialang.org/u/JordiBolibar)\
**Replies:** 54\
**Last updated:** [January 31, 2022, 11:47am UTC](https://discourse.julialang.org/t/optimizing-performance-of-2d-nonlinear-diffusion-ude/72113 "2022-01-31T11:47:17Z")

</div>

I have a UDE based on a 2D nonlinear ice flow diffusion PDE. It works correctly both for the forward and backwards pass, but despite using a semi-implicit solver the memory usage for backpropagation is just too big for e…

---

## [Julia as a neural interpreter? Could this be a game-change for Julia and ML/AI?](https://discourse.julialang.org/t/julia-as-a-neural-interpreter-could-this-be-a-game-change-for-julia-and-ml-ai/75386)

<div class="topic-metadata">

**Author:** [@Palli](https://discourse.julialang.org/u/Palli)\
**Replies:** 0\
**Last updated:** [January 28, 2022, 5:56pm UTC](https://discourse.julialang.org/t/julia-as-a-neural-interpreter-could-this-be-a-game-change-for-julia-and-ml-ai/75386 "2022-01-28T17:56:07Z")

</div>

I’ve sometimes posted ML stuff I find intriguing to offtopic. This seems relevant enough to post here. @ChrisRackauckas claims Julia doesn’t have an advantage for “standard” ML (unlike SciML). This may not be “standard”…

---

## [NeuralPDE to simulate Burguer 1D - not a satisfactorial solution](https://discourse.julialang.org/t/neuralpde-to-simulate-burguer-1d-not-a-satisfactorial-solution/75229)

<div class="topic-metadata">

**Author:** [@BuddhiLW](https://discourse.julialang.org/u/BuddhiLW)\
**Replies:** 6\
**Last updated:** [January 26, 2022, 1:56pm UTC](https://discourse.julialang.org/t/neuralpde-to-simulate-burguer-1d-not-a-satisfactorial-solution/75229 "2022-01-26T13:56:02Z")

</div>

I’m writting a research in which I heavily used NeuralPDE.jl, and I would like to showcase solution vs analytical, or classical simulations vs provided by the package. In one of these instances, I’m getting an answer I …

---

## [Backpropagating through orthogonal projection operator](https://discourse.julialang.org/t/backpropagating-through-orthogonal-projection-operator/63083)

<div class="topic-metadata">

**Author:** [@rkube](https://discourse.julialang.org/u/rkube)\
**Replies:** 2\
**Last updated:** [January 25, 2022, 5:32pm UTC](https://discourse.julialang.org/t/backpropagating-through-orthogonal-projection-operator/63083 "2022-01-25T17:32:34Z")

</div>

Hi, I have a neural network whose loss functions requires to calculate orthogonal projections onto the output of the network. The code runs fine on the CPU, albeit slow. Porting it to the GPU is difficult and I’m stuck.…

---

## [Missing line from deffeqflux blog post](https://discourse.julialang.org/t/missing-line-from-deffeqflux-blog-post/75141)

<div class="topic-metadata">

**Author:** [@pl-ai](https://discourse.julialang.org/u/pl-ai)\
**Replies:** 1\
**Last updated:** [January 25, 2022, 8:10am UTC](https://discourse.julialang.org/t/missing-line-from-deffeqflux-blog-post/75141 "2022-01-25T08:10:13Z")

</div>

In the blog post linked from the diffeqflux package (DiffEqFlux.jl – A Julia Library for Neural Differential Equations), there is this section down the page: Notice that we are not learning a solution to the ODE. Instea…

---

## [Training Deep Neural Network using Data Parallel?](https://discourse.julialang.org/t/training-deep-neural-network-using-data-parallel/75079)

<div class="topic-metadata">

**Author:** [@rkube](https://discourse.julialang.org/u/rkube)\
**Replies:** 5\
**Last updated:** [January 24, 2022, 9:08pm UTC](https://discourse.julialang.org/t/training-deep-neural-network-using-data-parallel/75079 "2022-01-24T21:08:19Z")

</div>

Hi, I need to train some simpler networks on large datasets and would like to use multiple GPUs for training. Is there a tutorial or writeup on how to do this in Julia? I know about this one in torch: Multi-GPU Examples…

[Previous page](https://discourse.julialang.org/c/domain/ml/24.md?page=25)

[Next page](https://discourse.julialang.org/c/domain/ml/24.md?page=27)