# Machine Learning

**URL:** https://discourse.julialang.org/c/domain/ml/24.md?page=38

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

**Page:** 39

---

## [Are there good resources for learning deep learning in Julia?](https://discourse.julialang.org/t/are-there-good-resources-for-learning-deep-learning-in-julia/48355)

<div class="topic-metadata">

**Author:** [@greatpet](https://discourse.julialang.org/u/greatpet)\
**Replies:** 3\
**Last updated:** [December 30, 2020, 5:08pm UTC](https://discourse.julialang.org/t/are-there-good-resources-for-learning-deep-learning-in-julia/48355 "2020-12-30T17:08:17Z")

</div>

I want to learn some machine learning / deep learning in my spare time, to understand algorithms like Google’s AlphaGo. I recently learned Julia and I’m wondering if there are good pedagogical introductions to ML without…

---

## [So I think I've learned the basics of Julia, where do I learn some more advanced content? (specifcally image feature extraction)](https://discourse.julialang.org/t/so-i-think-ive-learned-the-basics-of-julia-where-do-i-learn-some-more-advanced-content-specifcally-image-feature-extraction/34108)

<div class="topic-metadata">

**Author:** [@samantp](https://discourse.julialang.org/u/samantp)\
**Replies:** 3\
**Last updated:** [December 30, 2020, 2:01pm UTC](https://discourse.julialang.org/t/so-i-think-ive-learned-the-basics-of-julia-where-do-i-learn-some-more-advanced-content-specifcally-image-feature-extraction/34108 "2020-12-30T14:01:00Z")

</div>

I am learning to use Julia for my new research (MRI image feature extraction and classification). The language I am most comfortable with is MATLAB but I also have some basic understanding of python. What I don’t yet hav…

---

## [Google's MuZero](https://discourse.julialang.org/t/googles-muzero/52364)

<div class="topic-metadata">

**Author:** [@StatisticalMouse](https://discourse.julialang.org/u/StatisticalMouse)\
**Replies:** 0\
**Last updated:** [December 25, 2020, 10:42am UTC](https://discourse.julialang.org/t/googles-muzero/52364 "2020-12-25T10:42:56Z")

</div>

This looks like a big deal. It uses model-based reinforcement learning and can learn to play both chess and Pac-Man. Abstract from the Science article: " Constructing agents with planning capabilities has long been one …

---

## [Zygote fails with gradients of destructured/restructured RNNs](https://discourse.julialang.org/t/zygote-fails-with-gradients-of-destructured-restructured-rnns/52065)

<div class="topic-metadata">

**Author:** [@noumea](https://discourse.julialang.org/u/noumea)\
**Replies:** 0\
**Last updated:** [December 18, 2020, 6:59pm UTC](https://discourse.julialang.org/t/zygote-fails-with-gradients-of-destructured-restructured-rnns/52065 "2020-12-18T18:59:24Z")

</div>

I am trying to optimize a NN with an Optim.jl optimizer by using destructure/restructure on the model, and Zygote to obtain the gradient of the loss function. I noticed that Zygote does not seem to be able to calculate t…

---

## [Do I have to implement Flux.testmode! for my own models?](https://discourse.julialang.org/t/do-i-have-to-implement-flux-testmode-for-my-own-models/52038)

<div class="topic-metadata">

**Author:** [@maxfreu](https://discourse.julialang.org/u/maxfreu)\
**Replies:** 1\
**Last updated:** [December 18, 2020, 6:00pm UTC](https://discourse.julialang.org/t/do-i-have-to-implement-flux-testmode-for-my-own-models/52038 "2020-12-18T18:00:58Z")

</div>

Hi, I implemented a U-Net and would like to switch between train and test mode for the BN layers. struct Unet encoder # tuple of chains decoder # tuple of UpBlocks segmentation\_head # normal conv layer end …

---

## [BatchNorm gradient is inconsistent with PyTorch in training mode](https://discourse.julialang.org/t/batchnorm-gradient-is-inconsistent-with-pytorch-in-training-mode/40389)

<div class="topic-metadata">

**Author:** [@yiyuezhuo](https://discourse.julialang.org/u/yiyuezhuo)\
**Replies:** 2\
**Last updated:** [December 18, 2020, 11:12am UTC](https://discourse.julialang.org/t/batchnorm-gradient-is-inconsistent-with-pytorch-in-training-mode/40389 "2020-12-18T11:12:29Z")

</div>

I found Flux’s BatchNorm output in training/test mode and gradient in test mode is consistent with PyTorch but gradient in training mode is not. It’s kind of strange since all parameters look the same. Example: using F…

---

## [MLJ and ClusterManagers](https://discourse.julialang.org/t/mlj-and-clustermanagers/51540)

<div class="topic-metadata">

**Author:** [@pfarndt](https://discourse.julialang.org/u/pfarndt)\
**Replies:** 4\
**Last updated:** [December 17, 2020, 8:16pm UTC](https://discourse.julialang.org/t/mlj-and-clustermanagers/51540 "2020-12-17T20:16:11Z")

</div>

I want do some machine learning using MLJ on a cluster of 200+ nodes. To hook up my workers to the main process I want to use an ElasticManager from the ClusterManagers package (this way I can also utilize workers which …

---

## [A question on the DiffEqFlux module](https://discourse.julialang.org/t/a-question-on-the-diffeqflux-module/51987)

<div class="topic-metadata">

**Author:** [@anC](https://discourse.julialang.org/u/anC)\
**Replies:** 0\
**Last updated:** [December 17, 2020, 3:51pm UTC](https://discourse.julialang.org/t/a-question-on-the-diffeqflux-module/51987 "2020-12-17T15:51:56Z")

</div>

First of all, I’d like to say thanks and congratulations to Dr. Rackauckas and the team responsible for the DiffEqFlux module. This is an excellent module and I can see some really useful applications of this work. I hav…

---

## [Weights from neural network as interaction score between pairs](https://discourse.julialang.org/t/weights-from-neural-network-as-interaction-score-between-pairs/51921)

<div class="topic-metadata">

**Author:** [@alperyilmaz](https://discourse.julialang.org/u/alperyilmaz)\
**Replies:** 0\
**Last updated:** [December 16, 2020, 9:05am UTC](https://discourse.julialang.org/t/weights-from-neural-network-as-interaction-score-between-pairs/51921 "2020-12-16T09:05:52Z")

</div>

The question is using genes as example but I thought “Machine Learning” section is more suitable for this question. Because the concept is independent from genes, might be applicable to other domains. Suppose we have ma…

---

## [Learning non-linear spatio-temporal dynamics with convolutional Neural ODEs](https://discourse.julialang.org/t/learning-non-linear-spatio-temporal-dynamics-with-convolutional-neural-odes/51448)

<div class="topic-metadata">

**Author:** [@johnh](https://discourse.julialang.org/u/johnh)\
**Replies:** 1\
**Last updated:** [December 10, 2020, 4:42am UTC](https://discourse.julialang.org/t/learning-non-linear-spatio-temporal-dynamics-with-convolutional-neural-odes/51448 "2020-12-10T04:42:27Z")

</div>

I work for Dell, but am not involved. @ChrisRackauckas please tell us this is all written in Julia… https://ml4physicalsciences.github.io/2020/files/NeurIPS\_ML4PS\_2020\_125.pdf

---

## [A differentiable and slicable 2D data structure for parameters?](https://discourse.julialang.org/t/a-differentiable-and-slicable-2d-data-structure-for-parameters/51405)

<div class="topic-metadata">

**Author:** [@kmaile](https://discourse.julialang.org/u/kmaile)\
**Replies:** 0\
**Last updated:** [December 7, 2020, 3:28pm UTC](https://discourse.julialang.org/t/a-differentiable-and-slicable-2d-data-structure-for-parameters/51405 "2020-12-07T15:28:36Z")

</div>

I am implementing neural architecture search methods in Julia. One class of them uses trainable architecture weights to learn the best option for each layer of the network. Each repeated cell within a network has N layer…

---

## [Mixed input layer using Flux gives "ERROR: Mutating arrays is not supported"](https://discourse.julialang.org/t/mixed-input-layer-using-flux-gives-error-mutating-arrays-is-not-supported/41595)

<div class="topic-metadata">

**Author:** [@rsteckel](https://discourse.julialang.org/u/rsteckel)\
**Replies:** 1\
**Last updated:** [December 7, 2020, 5:08am UTC](https://discourse.julialang.org/t/mixed-input-layer-using-flux-gives-error-mutating-arrays-is-not-supported/41595 "2020-12-07T05:08:45Z")

</div>

I would like to combine input types in the input layer of my Flux model. For example, using the Embedded example layer (found on this forum), I would like to embed the first column of my input data and use a Dense layer…

---

## [Flux: "cannot broadcast array to have fewer dimensions"](https://discourse.julialang.org/t/flux-cannot-broadcast-array-to-have-fewer-dimensions/39821)

<div class="topic-metadata">

**Author:** [@dunefox](https://discourse.julialang.org/u/dunefox)\
**Replies:** 4\
**Last updated:** [December 6, 2020, 7:06pm UTC](https://discourse.julialang.org/t/flux-cannot-broadcast-array-to-have-fewer-dimensions/39821 "2020-12-06T19:06:23Z")

</div>

I’m trying to get into Flux and Julia, so I defined two simple models. The first one works, the second one throws an error that I don’t quite understand yet. I’m sure it’s relatively simple to solve, I just don’t see it. …

---

## [How to apply Transfer Learning with Flux](https://discourse.julialang.org/t/how-to-apply-transfer-learning-with-flux/51250)

<div class="topic-metadata">

**Author:** [@Mariana](https://discourse.julialang.org/u/Mariana)\
**Replies:** 6\
**Last updated:** [December 4, 2020, 10:49pm UTC](https://discourse.julialang.org/t/how-to-apply-transfer-learning-with-flux/51250 "2020-12-04T22:49:51Z")

</div>

I have a trained neural network, which is already giving good results, and I’d like to use the parameters previously obtained for the model as the starting parameters to train the ANN on a new task (basically, it’s a tra…

---

## [Flux Zygote Gradient: Understanding Mutating arrays is not supported](https://discourse.julialang.org/t/flux-zygote-gradient-understanding-mutating-arrays-is-not-supported/46324)

<div class="topic-metadata">

**Author:** [@rakeshvar](https://discourse.julialang.org/u/rakeshvar)\
**Replies:** 21\
**Last updated:** [December 3, 2020, 5:21am UTC](https://discourse.julialang.org/t/flux-zygote-gradient-understanding-mutating-arrays-is-not-supported/46324 "2020-12-03T05:21:49Z")

</div>

Namaste, I am trying to implement CTC loss in Flux/Zygote. And running into Error: Mutating arrays is not supported. So I cooked up the toy example below to gain better understanding. Given an input x, I perform some s…

---

## [Can Julia efficiently make use of 20+ cores for transforming hundreds of millions of rows for machine learning?](https://discourse.julialang.org/t/can-julia-efficiently-make-use-of-20-cores-for-transforming-hundreds-of-millions-of-rows-for-machine-learning/50968)

<div class="topic-metadata">

**Author:** [@Kevin\_Shen](https://discourse.julialang.org/u/Kevin_Shen)\
**Replies:** 27\
**Last updated:** [December 1, 2020, 3:15pm UTC](https://discourse.julialang.org/t/can-julia-efficiently-make-use-of-20-cores-for-transforming-hundreds-of-millions-of-rows-for-machine-learning/50968 "2020-12-01T15:15:30Z")

</div>

I’ve seen similar questions posted but I wanted to ask my question to see if I can get more precise answers. I train machine learning models on datasets with hundreds of millions of rows (10^8). On this order of magnitu…

---

## [Partitioning Around Medoids (PAM) in Julia](https://discourse.julialang.org/t/partitioning-around-medoids-pam-in-julia/50879)

<div class="topic-metadata">

**Author:** [@mthelm85](https://discourse.julialang.org/u/mthelm85)\
**Replies:** 10\
**Last updated:** [November 30, 2020, 4:45pm UTC](https://discourse.julialang.org/t/partitioning-around-medoids-pam-in-julia/50879 "2020-11-30T16:45:20Z")

</div>

I’m unable to find any package that implements this algorithm. I’m going to attempt to implement the algorithm myself but I’m not sure if I have the know-how to get it over the finish line, to be perfectly honest. Would …

---

## [Best way to learn differentiable computing?](https://discourse.julialang.org/t/best-way-to-learn-differentiable-computing/50955)

<div class="topic-metadata">

**Author:** [@StatisticalMouse](https://discourse.julialang.org/u/StatisticalMouse)\
**Replies:** 1\
**Last updated:** [November 30, 2020, 3:52pm UTC](https://discourse.julialang.org/t/best-way-to-learn-differentiable-computing/50955 "2020-11-30T15:52:28Z")

</div>

What would be the best resource to get started? Also, what would be the ‘state of the art’ in this area right now?

---

## [K-Medoids in Julia - Results Quality](https://discourse.julialang.org/t/k-medoids-in-julia-results-quality/50687)

<div class="topic-metadata">

**Author:** [@mthelm85](https://discourse.julialang.org/u/mthelm85)\
**Replies:** 7\
**Last updated:** [November 27, 2020, 9:54pm UTC](https://discourse.julialang.org/t/k-medoids-in-julia-results-quality/50687 "2020-11-27T21:54:05Z")

</div>

I was reading the Wikipedia page on k-medoids clustering and in the Software section, it’s stated that: Julia contains a k-medoid implementation of the k-means style algorithm (faster, but much worse result quality) i…

---

## [Using gradients from struct with Zygote](https://discourse.julialang.org/t/using-gradients-from-struct-with-zygote/50766)

<div class="topic-metadata">

**Author:** [@theogf](https://discourse.julialang.org/u/theogf)\
**Replies:** 2\
**Last updated:** [November 26, 2020, 12:35pm UTC](https://discourse.julialang.org/t/using-gradients-from-struct-with-zygote/50766 "2020-11-26T12:35:15Z")

</div>

Hi! Let’s say I have some nested struct (in my case the nesting is arbitrarily complicated) struct Foo a::Vector end struct Bar b::Foo c::Float64 end c = Bar(Foo(\[2.0\]), 1.0) When using g = Zygote.gradient(c) do x …

---

## [MLJ needs your testimony](https://discourse.julialang.org/t/mlj-needs-your-testimony/50784)

<div class="topic-metadata">

**Author:** [@TheRoniOne](https://discourse.julialang.org/u/TheRoniOne)\
**Replies:** 0\
**Last updated:** [November 25, 2020, 11:54pm UTC](https://discourse.julialang.org/t/mlj-needs-your-testimony/50784 "2020-11-25T23:54:25Z")

</div>

In The Alan Turing Institute they are reviewing the funding of MLJ and they are asking for evidence that MLJ is being used, filling a need and having an impact. In order for the project to continue seeing support, they n…

---

## [Correct sensealg for an ensemble problem of stochastic UODEs with diagonal noise](https://discourse.julialang.org/t/correct-sensealg-for-an-ensemble-problem-of-stochastic-uodes-with-diagonal-noise/50490)

<div class="topic-metadata">

**Author:** [@Libbum](https://discourse.julialang.org/u/Libbum)\
**Replies:** 17\
**Last updated:** [November 23, 2020, 10:03am UTC](https://discourse.julialang.org/t/correct-sensealg-for-an-ensemble-problem-of-stochastic-uodes-with-diagonal-noise/50490 "2020-11-23T10:03:04Z")

</div>

Hi all, I can’t seem to get my stochastic UODE running. Here’s the setup and generation of synthetic data: using Plots, DiffEqFlux, Flux, Optim, LinearAlgebra, DifferentialEquations using Statistics using DiffEqSensiti…

---

## [Flux differentiation error](https://discourse.julialang.org/t/flux-differentiation-error/49811)

<div class="topic-metadata">

**Author:** [@Honza9723](https://discourse.julialang.org/u/Honza9723)\
**Replies:** 19\
**Last updated:** [November 19, 2020, 1:20pm UTC](https://discourse.julialang.org/t/flux-differentiation-error/49811 "2020-11-19T13:20:51Z")

</div>

Dear All, I am working on a solution of simple differential equation y(x) - y’(x) = 0 using neural network. I am using a two-layer softplus network for parameterization of y(x). To get y’(x), I am using simple Flux.gra…

---

## [Some structures cannot be differentiated automatically?](https://discourse.julialang.org/t/some-structures-cannot-be-differentiated-automatically/50358)

<div class="topic-metadata">

**Author:** [@yeruoforever](https://discourse.julialang.org/u/yeruoforever)\
**Replies:** 0\
**Last updated:** [November 18, 2020, 5:46am UTC](https://discourse.julialang.org/t/some-structures-cannot-be-differentiated-automatically/50358 "2020-11-18T05:46:43Z")

</div>

Dear All, I am a student learning neural network. I implemented a small graph neural network model using Flux and MetaGraphs, but there was a problem that puzzled me. I posted the problematic code below, hope my friends…

---

## [DifferentialEquations - Derivatives in ODE function/ nesting AD](https://discourse.julialang.org/t/differentialequations-derivatives-in-ode-function-nesting-ad/50100)

<div class="topic-metadata">

**Author:** [@Andreas\_Schlaginhauf](https://discourse.julialang.org/u/Andreas_Schlaginhauf)\
**Replies:** 6\
**Last updated:** [November 17, 2020, 2:57pm UTC](https://discourse.julialang.org/t/differentialequations-derivatives-in-ode-function-nesting-ad/50100 "2020-11-17T14:57:21Z")

</div>

If the ODE function of a neural ODE includes the gradient f’(x) of a neural network f(x), we need to calculate second order gradients of f(x) in the backward pass. For me this led to all sorts of problems when the AD met…

---

## [Pixelwise loss weights in Flux?](https://discourse.julialang.org/t/pixelwise-loss-weights-in-flux/49332)

<div class="topic-metadata">

**Author:** [@tlnagy](https://discourse.julialang.org/u/tlnagy)\
**Replies:** 2\
**Last updated:** [November 14, 2020, 7:48am UTC](https://discourse.julialang.org/t/pixelwise-loss-weights-in-flux/49332 "2020-11-14T07:48:49Z")

</div>

Hi all, I’m interested in implementing a pixel-wise loss function in Flux where I would weight up pixels that separate segments from each other to force the network to learn those pixels preferentially. That is, I want …

---

## [Using Julia to extract information from a ballot](https://discourse.julialang.org/t/using-julia-to-extract-information-from-a-ballot/47239)

<div class="topic-metadata">

**Author:** [@Nosferican](https://discourse.julialang.org/u/Nosferican)\
**Replies:** 7\
**Last updated:** [November 12, 2020, 10:14am UTC](https://discourse.julialang.org/t/using-julia-to-extract-information-from-a-ballot/47239 "2020-11-12T10:14:13Z")

</div>

I would like to identify tools in the Julia ecosystem to parse a model ballot. For example, from the following PDF, obtain Candidature Candidate’s name Party Affiliation

---

## [PyTorch DataLoader equivalent for training large models with Flux](https://discourse.julialang.org/t/pytorch-dataloader-equivalent-for-training-large-models-with-flux/30763)

<div class="topic-metadata">

**Author:** [@holylorenzo](https://discourse.julialang.org/u/holylorenzo)\
**Replies:** 16\
**Last updated:** [November 8, 2020, 8:58am UTC](https://discourse.julialang.org/t/pytorch-dataloader-equivalent-for-training-large-models-with-flux/30763 "2020-11-08T08:58:03Z")

</div>

I’m trying to train a large computer vision model I built in Flux. Unfortunately the dataset does not fit into memory. Does something similar to PyTorch’s DataLoader exist as a Julia package? Is there an ongoing effort …

---

## [Gradient of gradient](https://discourse.julialang.org/t/gradient-of-gradient/49660)

<div class="topic-metadata">

**Author:** [@jlmaccal](https://discourse.julialang.org/u/jlmaccal)\
**Replies:** 9\
**Last updated:** [November 6, 2020, 9:44pm UTC](https://discourse.julialang.org/t/gradient-of-gradient/49660 "2020-11-06T21:44:06Z")

</div>

I’m implementing a model that requires using the gradients of a feedforward network with respect to its inputs as part of the loss function. I then need to train the network by differentiating the loss with respect to th…

---

## [Calculate gradient of UDE model (without DiffEqFlux.sciml\_train)](https://discourse.julialang.org/t/calculate-gradient-of-ude-model-without-diffeqflux-sciml-train/49488)

<div class="topic-metadata">

**Author:** [@bad\_at\_math](https://discourse.julialang.org/u/bad_at_math)\
**Replies:** 3\
**Last updated:** [November 6, 2020, 5:12pm UTC](https://discourse.julialang.org/t/calculate-gradient-of-ude-model-without-diffeqflux-sciml-train/49488 "2020-11-06T17:12:01Z")

</div>

Hello everyone, I’m interested in using the SciML ecosystem to train universal differential equations models in Julia. However, I don’t want to have to rely on the sciml\_train function – I’m interested in explicitly work…

[Previous page](https://discourse.julialang.org/c/domain/ml/24.md?page=37)

[Next page](https://discourse.julialang.org/c/domain/ml/24.md?page=39)