# Machine Learning

**URL:** https://discourse.julialang.org/c/domain/ml/24.md?page=51

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

**Page:** 52

---

## [HN comment on Julia vs Swift type checking and auto diff](https://discourse.julialang.org/t/hn-comment-on-julia-vs-swift-type-checking-and-auto-diff/24178)

<div class="topic-metadata">

**Author:** [@datnamer](https://discourse.julialang.org/u/datnamer)\
**Replies:** 24\
**Last updated:** [May 14, 2019, 6:55pm UTC](https://discourse.julialang.org/t/hn-comment-on-julia-vs-swift-type-checking-and-auto-diff/24178 "2019-05-14T18:55:42Z")

</div>

Anyone know if this is accurate? https://news.ycombinator.com/item?id=19897585 “I don’t think static type checking is a matter of tooling, getting Julia to perform well amounts to either a lot of trust in the compiler …

---

## [Gradient Descent Optimizer in Flux.jl](https://discourse.julialang.org/t/gradient-descent-optimizer-in-flux-jl/24169)

<div class="topic-metadata">

**Author:** [@FGerace](https://discourse.julialang.org/u/FGerace)\
**Replies:** 1\
**Last updated:** [May 13, 2019, 5:12pm UTC](https://discourse.julialang.org/t/gradient-descent-optimizer-in-flux-jl/24169 "2019-05-13T17:12:11Z")

</div>

Hello, I’ve recently started using Flux.jl. I’ve seen that Flux provides different kinds of optimizers, such as Descent, ADAM, etc etc… I was just wondering: does the Descent optimizer perform full batch gradient desce…

---

## [Separable convolution](https://discourse.julialang.org/t/separable-convolution/24141)

<div class="topic-metadata">

**Author:** [@Tomas\_Pevny](https://discourse.julialang.org/u/Tomas_Pevny)\
**Replies:** 7\
**Last updated:** [May 13, 2019, 6:15am UTC](https://discourse.julialang.org/t/separable-convolution/24141 "2019-05-13T06:15:27Z")

</div>

Hi All, I would like to run a separate over individual filters. Specifically imagine I have two matrices and I want to convolve each of them with its own filter. And I wonder if it possible to write it as a single convo…

---

## [Do you actually use FluxML to get work done?](https://discourse.julialang.org/t/do-you-actually-use-fluxml-to-get-work-done/23304)

<div class="topic-metadata">

**Author:** [@PhilipB](https://discourse.julialang.org/u/PhilipB)\
**Replies:** 16\
**Last updated:** [May 9, 2019, 9:21pm UTC](https://discourse.julialang.org/t/do-you-actually-use-fluxml-to-get-work-done/23304 "2019-05-09T21:21:08Z")

</div>

Hello. I am a student and I am studying tensor methods for deep learning. I decided to try Julia with FluxML yesterday, and so far I met 2 problems of type “FluxML tracked array doesn’t compile when used with this standa…

---

## [Flux networks as arguments in functions](https://discourse.julialang.org/t/flux-networks-as-arguments-in-functions/23988)

<div class="topic-metadata">

**Author:** [@charlesll](https://discourse.julialang.org/u/charlesll)\
**Replies:** 6\
**Last updated:** [May 8, 2019, 10:22pm UTC](https://discourse.julialang.org/t/flux-networks-as-arguments-in-functions/23988 "2019-05-08T22:22:25Z")

</div>

Hey, I recently opened an issue on Flux as there is something that needs clarification I think about using Flux with Julia’s functions. To summarize my issue: I have a project where I have various equations embedded in…

---

## [Training a network to optimize output value of another network](https://discourse.julialang.org/t/training-a-network-to-optimize-output-value-of-another-network/23857)

<div class="topic-metadata">

**Author:** [@HenriDeh](https://discourse.julialang.org/u/HenriDeh)\
**Replies:** 4\
**Last updated:** [May 5, 2019, 3:35pm UTC](https://discourse.julialang.org/t/training-a-network-to-optimize-output-value-of-another-network/23857 "2019-05-05T15:35:40Z")

</div>

I am trying to set-up a Deep Deterministic Policy Gradient algorithm (Reinforcement Learning) with Flux. Let me give a context-free explanation of what I am trying to achieve. Say I have two networks A and C. Network A …

---

## [Reservoir computing or echo state networks in Flux.jl or Knet.jl](https://discourse.julialang.org/t/reservoir-computing-or-echo-state-networks-in-flux-jl-or-knet-jl/16466)

<div class="topic-metadata">

**Author:** [@John\_Gibson](https://discourse.julialang.org/u/John_Gibson)\
**Replies:** 5\
**Last updated:** [April 27, 2019, 7:59pm UTC](https://discourse.julialang.org/t/reservoir-computing-or-echo-state-networks-in-flux-jl-or-knet-jl/16466 "2019-04-27T19:59:24Z")

</div>

Can any of the Julia machine learning frameworks do reservoir computing or echo state computing? Are there examples how to configure Flux.jl or Knet.jl to do either of these on a simple problem? As far as I understand i…

---

## [Implementing multiple skip connections in Flux.ml?](https://discourse.julialang.org/t/implementing-multiple-skip-connections-in-flux-ml/23378)

<div class="topic-metadata">

**Author:** [@dopamine](https://discourse.julialang.org/u/dopamine)\
**Replies:** 0\
**Last updated:** [April 21, 2019, 6:09pm UTC](https://discourse.julialang.org/t/implementing-multiple-skip-connections-in-flux-ml/23378 "2019-04-21T18:09:42Z")

</div>

Hello, I’m trying to implement u-net (see here for architecture) like architecture using Flux.ml, which requires multiple skip connections. What is the most efficient way to implement multiple skip connections? I’ve see…

---

## [Julia Implementation of Transformer Neural Network Model](https://discourse.julialang.org/t/julia-implementation-of-transformer-neural-network-model/23282)

<div class="topic-metadata">

**Author:** [@mjb](https://discourse.julialang.org/u/mjb)\
**Replies:** 3\
**Last updated:** [April 19, 2019, 6:08pm UTC](https://discourse.julialang.org/t/julia-implementation-of-transformer-neural-network-model/23282 "2019-04-19T18:08:49Z")

</div>

I developed an implementation of a Transformer model using Flux 0.8.2. The implementation is based on The Annotated Transformer If this is of interest, I am happy to contribute it (I don’t know how or where the best pla…

---

## [Julia motivation for machine learning](https://discourse.julialang.org/t/julia-motivation-for-machine-learning/23259)

<div class="topic-metadata">

**Author:** [@tk3369](https://discourse.julialang.org/u/tk3369)\
**Replies:** 6\
**Last updated:** [April 18, 2019, 1:29pm UTC](https://discourse.julialang.org/t/julia-motivation-for-machine-learning/23259 "2019-04-18T13:29:30Z")

</div>

Out of curiosity, in the domain of machine learning and NLP - what would be the reasons to choose julia over python? What would be the reasons to choose python over julia? What are some of the gaps in julia as c…

---

## [Is it possible to selectively update the params in Flux.jl?](https://discourse.julialang.org/t/is-it-possible-to-selectively-update-the-params-in-flux-jl/23198)

<div class="topic-metadata">

**Author:** [@xiaodai](https://discourse.julialang.org/u/xiaodai)\
**Replies:** 9\
**Last updated:** [April 17, 2019, 10:54am UTC](https://discourse.julialang.org/t/is-it-possible-to-selectively-update-the-params-in-flux-jl/23198 "2019-04-17T10:54:49Z")

</div>

I can’t seem to update the weights selectively. See below example where x is a set of 3 weights and I want to only update, say the first component. But can’t seem to do so. Is there a way to do that? x = param(rand(3)) …

---

## [Backward pass for sampling](https://discourse.julialang.org/t/backward-pass-for-sampling/23071)

<div class="topic-metadata">

**Author:** [@Roger-luo](https://discourse.julialang.org/u/Roger-luo)\
**Replies:** 2\
**Last updated:** [April 14, 2019, 5:13pm UTC](https://discourse.julialang.org/t/backward-pass-for-sampling/23071 "2019-04-14T17:13:13Z")

</div>

I’m wondering if people have implement this before somewhere, like Turing? but I’m not very familiar with Turing and how it is integrated with ADs, so maybe someone could point me to reference: Given a parameterized dis…

---

## [Naive Bayes text classification: what package in Julia?](https://discourse.julialang.org/t/naive-bayes-text-classification-what-package-in-julia/23042)

<div class="topic-metadata">

**Author:** [@axelle](https://discourse.julialang.org/u/axelle)\
**Replies:** 1\
**Last updated:** [April 11, 2019, 5:33pm UTC](https://discourse.julialang.org/t/naive-bayes-text-classification-what-package-in-julia/23042 "2019-04-11T17:33:57Z")

</div>

Good afternoon, I would like to use Julia for naive Bayes text classification. I have many investment fund names that i would like to regroup under the same master parent which can generally be recognized by looking at …

---

## [\`back!\` was already used error in Flux](https://discourse.julialang.org/t/back-was-already-used-error-in-flux/22709)

<div class="topic-metadata">

**Author:** [@Dipak\_Narayan](https://discourse.julialang.org/u/Dipak_Narayan)\
**Replies:** 0\
**Last updated:** [April 3, 2019, 5:45pm UTC](https://discourse.julialang.org/t/back-was-already-used-error-in-flux/22709 "2019-04-03T17:45:18Z")

</div>

What I’m trying to do is pass a linear combination of images (lets say a1img1 + a2img2 + a3\*img3) through a neural network and I want the network to choose the best a1, a2, a3 for classification. To do this I’ve made a …

---

## [Is it possible to make \`Normal(mu, sigma)\` work for when \`mu\` and \`sigma\` are tracked arrays?](https://discourse.julialang.org/t/is-it-possible-to-make-normal-mu-sigma-work-for-when-mu-and-sigma-are-tracked-arrays/22616)

<div class="topic-metadata">

**Author:** [@xiaodai](https://discourse.julialang.org/u/xiaodai)\
**Replies:** 0\
**Last updated:** [April 1, 2019, 11:11pm UTC](https://discourse.julialang.org/t/is-it-possible-to-make-normal-mu-sigma-work-for-when-mu-and-sigma-are-tracked-arrays/22616 "2019-04-01T23:11:45Z")

</div>

I am trying to make some code where the parameters defined the distribution. This is done in variational auto-encoders (VAE). But I am doing something simpler, I am just trying to implement a maximum likelihood estimator…

---

## [Flux params() not working for composed layers](https://discourse.julialang.org/t/flux-params-not-working-for-composed-layers/22480)

<div class="topic-metadata">

**Author:** [@aldopareja](https://discourse.julialang.org/u/aldopareja)\
**Replies:** 2\
**Last updated:** [April 1, 2019, 6:08pm UTC](https://discourse.julialang.org/t/flux-params-not-working-for-composed-layers/22480 "2019-04-01T18:08:24Z")

</div>

I’m trying to create a multi-headed attention layer, which is composed several “heads”. Each head has its own parameters (three matrices) and the “wrapper” layer has one output parameter (matrix). This is the code to do …

---

## [Optimizing parameters of ODE in DiffEqFlux: Not implemented: convert tracked Tracker.TrackedReal{Float64} to tracked Float64](https://discourse.julialang.org/t/optimizing-parameters-of-ode-in-diffeqflux-not-implemented-convert-tracked-tracker-trackedreal-float64-to-tracked-float64/22560)

<div class="topic-metadata">

**Author:** [@mzhenirovskyy](https://discourse.julialang.org/u/mzhenirovskyy)\
**Replies:** 4\
**Last updated:** [March 31, 2019, 8:33pm UTC](https://discourse.julialang.org/t/optimizing-parameters-of-ode-in-diffeqflux-not-implemented-convert-tracked-tracker-trackedreal-float64-to-tracked-float64/22560 "2019-03-31T20:33:58Z")

</div>

My model: using Random using Flux, DiffEqFlux, DifferentialEquations function model(dZ, Z, params, t) W0 = reshape(params\[1:12\], (1, 12)) b0 = reshape(params\[13:24\], (1, 12)) W1 = reshape(params\[25:36\], (1…

---

## [Flux.jl: Different set of optimisation parameters per layer](https://discourse.julialang.org/t/flux-jl-different-set-of-optimisation-parameters-per-layer/22518)

<div class="topic-metadata">

**Author:** [@rederekt](https://discourse.julialang.org/u/rederekt)\
**Replies:** 1\
**Last updated:** [March 30, 2019, 8:59pm UTC](https://discourse.julialang.org/t/flux-jl-different-set-of-optimisation-parameters-per-layer/22518 "2019-03-30T20:59:49Z")

</div>

Hi, Just copying over my question from the GitHub issues to hopefully get more visibility here. In Flux.jl, I’d like to use a separate set of optimisation parameters for each hidden layer in a single Chain. For example…

---

## [Flux.jl: Changing the value of activations from hidden nodes during backprop](https://discourse.julialang.org/t/flux-jl-changing-the-value-of-activations-from-hidden-nodes-during-backprop/22525)

<div class="topic-metadata">

**Author:** [@v-i-s-h](https://discourse.julialang.org/u/v-i-s-h)\
**Replies:** 0\
**Last updated:** [March 30, 2019, 3:46am UTC](https://discourse.julialang.org/t/flux-jl-changing-the-value-of-activations-from-hidden-nodes-during-backprop/22525 "2019-03-30T03:46:00Z")

</div>

Hi, I have a research project where I want to change the value of activations during backpropagation. It works like this: First I do the forward pass the traditional way and compute node activations at all layers. Then …

---

## [Neural ODE in DiffEqFlux that is not a time series](https://discourse.julialang.org/t/neural-ode-in-diffeqflux-that-is-not-a-time-series/22395)

<div class="topic-metadata">

**Author:** [@JackDevine](https://discourse.julialang.org/u/JackDevine)\
**Replies:** 6\
**Last updated:** [March 30, 2019, 12:07am UTC](https://discourse.julialang.org/t/neural-ode-in-diffeqflux-that-is-not-a-time-series/22395 "2019-03-30T00:07:16Z")

</div>

I am trying to use DiffEqFlux.jl to make a model that takes in a vector of length 200 and passes it through a neural ODE layer to output a vector of length 200. I tried to do so by tweaking the time series example in th…

---

## [Flux.jl Training loop where data sequentially depends on model output](https://discourse.julialang.org/t/flux-jl-training-loop-where-data-sequentially-depends-on-model-output/22510)

<div class="topic-metadata">

**Author:** [@Ares\_Fisher](https://discourse.julialang.org/u/Ares_Fisher)\
**Replies:** 1\
**Last updated:** [March 29, 2019, 7:22pm UTC](https://discourse.julialang.org/t/flux-jl-training-loop-where-data-sequentially-depends-on-model-output/22510 "2019-03-29T19:22:54Z")

</div>

Hey all! I’m interested in writing some “closed loop” generative models that receive their prediction error for the previous input as their next input. Is there a way to do this in Flux.jl’s Train! loop? Thanks!

---

## [Flux parallel execution](https://discourse.julialang.org/t/flux-parallel-execution/12931)

<div class="topic-metadata">

**Author:** [@kkmann](https://discourse.julialang.org/u/kkmann)\
**Replies:** 3\
**Last updated:** [March 29, 2019, 3:48pm UTC](https://discourse.julialang.org/t/flux-parallel-execution/12931 "2019-03-29T15:48:33Z")

</div>

Hi, so, this might be a very dumb question but how do I train model in flux.jl in parallel on my cpu?

---

## [Flux.train failing with CUDA GPU](https://discourse.julialang.org/t/flux-train-failing-with-cuda-gpu/22003)

<div class="topic-metadata">

**Author:** [@bjenkinsgit](https://discourse.julialang.org/u/bjenkinsgit)\
**Replies:** 5\
**Last updated:** [March 27, 2019, 1:57pm UTC](https://discourse.julialang.org/t/flux-train-failing-with-cuda-gpu/22003 "2019-03-27T13:57:30Z")

</div>

I’m testing some julia Deep Learning code and at the point of invoking Flux.train, I get the following: ReadOnlyMemoryError() specifically, I’m invoking the train function call like so: @time for i in 1:100 Flux.tra…

---

## [PyTorch and Julia](https://discourse.julialang.org/t/pytorch-and-julia/22211)

<div class="topic-metadata">

**Author:** [@dmolina](https://discourse.julialang.org/u/dmolina)\
**Replies:** 12\
**Last updated:** [March 27, 2019, 2:46am UTC](https://discourse.julialang.org/t/pytorch-and-julia/22211 "2019-03-27T02:46:04Z")

</div>

I was thinking about something related, but not about numpy, it’s about pytorch. As many folks know that PyTorch is implementing something seems to be very similar to Julia beneath the Python interface to enable JIT comp…

---

## [Shape of data for sequence learning in Flux.jl?](https://discourse.julialang.org/t/shape-of-data-for-sequence-learning-in-flux-jl/20811)

<div class="topic-metadata">

**Author:** [@clemej](https://discourse.julialang.org/u/clemej)\
**Replies:** 1\
**Last updated:** [March 26, 2019, 5:43pm UTC](https://discourse.julialang.org/t/shape-of-data-for-sequence-learning-in-flux-jl/20811 "2019-03-26T17:43:12Z")

</div>

I’m teaching myself Julia and porting some Python/Keras code over to FluxML, and I’m having trouble shaping the data in Julia. The python keras code is (basically): model = Sequential() model.add(LSTM(9, input\_shape=(1…

---

## [\[ANN\] Flux v0.8](https://discourse.julialang.org/t/ann-flux-v0-8/22305)

<div class="topic-metadata">

**Author:** [@dhairyagandhi96](https://discourse.julialang.org/u/dhairyagandhi96)\
**Replies:** 1\
**Last updated:** [March 25, 2019, 3:31pm UTC](https://discourse.julialang.org/t/ann-flux-v0-8/22305 "2019-03-25T15:31:30Z")

</div>

We are pleased to announce that Flux v0.8 has been released and is available for download now. Here are some of the major highlights from this release and the full list is available in the NEWS.md: New ConvTranspose la…

---

## [Shuffeld minibatches of a large datasets](https://discourse.julialang.org/t/shuffeld-minibatches-of-a-large-datasets/21774)

<div class="topic-metadata">

**Author:** [@Alexander-Barth](https://discourse.julialang.org/u/Alexander-Barth)\
**Replies:** 3\
**Last updated:** [March 25, 2019, 2:39pm UTC](https://discourse.julialang.org/t/shuffeld-minibatches-of-a-large-datasets/21774 "2019-03-25T14:39:35Z")

</div>

I would like to train a neural network (Knet or Flux, maybe I test both) on a large date set (larger than the available memory) representing a serie of images. In python tensorflow, I would use the Dataset API (tf.data.…

---

## [Training dataset with happy and sad faces?](https://discourse.julialang.org/t/training-dataset-with-happy-and-sad-faces/22207)

<div class="topic-metadata">

**Author:** [@johnh](https://discourse.julialang.org/u/johnh)\
**Replies:** 2\
**Last updated:** [March 22, 2019, 3:59pm UTC](https://discourse.julialang.org/t/training-dataset-with-happy-and-sad-faces/22207 "2019-03-22T15:59:05Z")

</div>

Are there any datasets for download which have faces tagged with emotion - eg happy/sad? I have an idea for a demonstrator system using an Nvidia Jetson board. The camera takes a picture of your face. an ML framework th…

---

## [OpenAL GPT](https://discourse.julialang.org/t/openal-gpt/22197)

<div class="topic-metadata">

**Author:** [@yash](https://discourse.julialang.org/u/yash)\
**Replies:** 0\
**Last updated:** [March 22, 2019, 1:07pm UTC](https://discourse.julialang.org/t/openal-gpt/22197 "2019-03-22T13:07:27Z")

</div>

Is there any implementation of OpenAL GPT model in julia?

---

## [FluxML Basic Custom Layer with Custom Loss Function](https://discourse.julialang.org/t/fluxml-basic-custom-layer-with-custom-loss-function/22036)

<div class="topic-metadata">

**Author:** [@LudiWin](https://discourse.julialang.org/u/LudiWin)\
**Replies:** 3\
**Last updated:** [March 19, 2019, 2:01pm UTC](https://discourse.julialang.org/t/fluxml-basic-custom-layer-with-custom-loss-function/22036 "2019-03-19T14:01:45Z")

</div>

Hi, I want to write a custom layer and a custom loss function and thought I’d start out with the basic linear regression problem from the tutorial of Flux and do the SGD from scratch for a custom layer in a Chain. For …

[Previous page](https://discourse.julialang.org/c/domain/ml/24.md?page=50)

[Next page](https://discourse.julialang.org/c/domain/ml/24.md?page=52)