# Machine Learning

**URL:** https://discourse.julialang.org/c/domain/ml/24.md?page=53

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

**Page:** 54

---

## [Knet issue](https://discourse.julialang.org/t/knet-issue/18355)

<div class="topic-metadata">

**Author:** [@microlifecc](https://discourse.julialang.org/u/microlifecc)\
**Replies:** 3\
**Last updated:** [January 22, 2019, 1:35am UTC](https://discourse.julialang.org/t/knet-issue/18355 "2019-01-22T01:35:30Z")

</div>

Hi Guys! I am trying to go through the tutorial here. But I am getting the error: Cannot find libknet8, please rerun Pkg.build("Knet"). Stacktrace: \[1\] im2col!(::Array{Float64,4}, ::Array{Float64,4}, ::Array{Float64…

---

## [Flux.jl - Best way to do efficient division by a parameter.](https://discourse.julialang.org/t/flux-jl-best-way-to-do-efficient-division-by-a-parameter/19773)

<div class="topic-metadata">

**Author:** [@JJB](https://discourse.julialang.org/u/JJB)\
**Replies:** 0\
**Last updated:** [January 18, 2019, 8:16am UTC](https://discourse.julialang.org/t/flux-jl-best-way-to-do-efficient-division-by-a-parameter/19773 "2019-01-18T08:16:22Z")

</div>

The following function evaluation: function OneDiv(x) return 1.0 ./ x end @code\_warntype OneDiv(Flux.param(\[0.1\])) Compiles to: Body::Any 1 ─ %1 = (Core.tuple)(1.0, x)::Tuple{Float64,TrackedArray{…,Array{Float64,…

---

## [Question about Flux.jl data inputs](https://discourse.julialang.org/t/question-about-flux-jl-data-inputs/19630)

<div class="topic-metadata">

**Author:** [@zjroth](https://discourse.julialang.org/u/zjroth)\
**Replies:** 1\
**Last updated:** [January 15, 2019, 1:27pm UTC](https://discourse.julialang.org/t/question-about-flux-jl-data-inputs/19630 "2019-01-15T13:27:01Z")

</div>

I’m playing around with Flux.jl, and I came across an error that I have some questions about. With a Dense layer, I can provide a matrix input in which each column represents a single input value. Additionally, the num…

---

## [TensorFlow test failing - my config?](https://discourse.julialang.org/t/tensorflow-test-failing-my-config/19463)

<div class="topic-metadata">

**Author:** [@Colin\_Beckingham](https://discourse.julialang.org/u/Colin_Beckingham)\
**Replies:** 1\
**Last updated:** [January 12, 2019, 11:00am UTC](https://discourse.julialang.org/t/tensorflow-test-failing-my-config/19463 "2019-01-12T11:00:58Z")

</div>

I updated my custom libtensorflow.so with bazel and I think everything went well, but test TF is failing in Julia so I wonder if the following is correct? pip is updated, Python has latest. Error on test begins with: E…

---

## [Type safe and expressive "tensor" code](https://discourse.julialang.org/t/type-safe-and-expressive-tensor-code/19338)

<div class="topic-metadata">

**Author:** [@datnamer](https://discourse.julialang.org/u/datnamer)\
**Replies:** 3\
**Last updated:** [January 11, 2019, 1:27am UTC](https://discourse.julialang.org/t/type-safe-and-expressive-tensor-code/19338 "2019-01-11T01:27:19Z")

</div>

There’s been some buzz lately about more expressive, and type safe array code for deep learning. One specific initiative has been the use of arrays with named dimensions: GitHub - ctongfei/nexus: Experimental tensor-ty…

---

## [Flux seq2seq](https://discourse.julialang.org/t/flux-seq2seq/13783)

<div class="topic-metadata">

**Author:** [@merckxiaan](https://discourse.julialang.org/u/merckxiaan)\
**Replies:** 12\
**Last updated:** [January 5, 2019, 12:52pm UTC](https://discourse.julialang.org/t/flux-seq2seq/13783 "2019-01-05T12:52:55Z")

</div>

I don’t know whether this is the right place to ask but I’m trying to code a seq2seq model in Flux and I’ve got a couple of questions. The encoder creates the hidden state for the decoder, how can I pass this state to t…

---

## [Julia code for the book Reinforcement Learning An Introduction(2nd)](https://discourse.julialang.org/t/julia-code-for-the-book-reinforcement-learning-an-introduction-2nd/19190)

<div class="topic-metadata">

**Author:** [@findmyway](https://discourse.julialang.org/u/findmyway)\
**Replies:** 0\
**Last updated:** [January 2, 2019, 11:46am UTC](https://discourse.julialang.org/t/julia-code-for-the-book-reinforcement-learning-an-introduction-2nd/19190 "2019-01-02T11:46:55Z")

</div>

Recently I reproduced most figures on the book Reinforcement Learning An Introduction in Julia. I hope it is helpful to those who are interested in both Julia and RL. If you also have a background in Python, I’d sugge…

---

## ["Homomorphic machine learning" in Haskell; goal to be more generic and faster than Julia](https://discourse.julialang.org/t/homomorphic-machine-learning-in-haskell-goal-to-be-more-generic-and-faster-than-julia/18802)

<div class="topic-metadata">

**Author:** [@datnamer](https://discourse.julialang.org/u/datnamer)\
**Replies:** 5\
**Last updated:** [December 19, 2018, 12:42am UTC](https://discourse.julialang.org/t/homomorphic-machine-learning-in-haskell-goal-to-be-more-generic-and-faster-than-julia/18802 "2018-12-19T00:42:56Z")

</div>

https://github.com/mikeizbicki/HLearn " HLearn is also a research project. The research goal is to discover the “best possible” interface for machine learning. This involves two competing demands: The library should be …

---

## [Setting Flux custom gradients](https://discourse.julialang.org/t/setting-flux-custom-gradients/18728)

<div class="topic-metadata">

**Author:** [@platawiec](https://discourse.julialang.org/u/platawiec)\
**Replies:** 6\
**Last updated:** [December 18, 2018, 8:24am UTC](https://discourse.julialang.org/t/setting-flux-custom-gradients/18728 "2018-12-18T08:24:24Z")

</div>

I have been trying to make a Spatial Transformer Network in Flux (https://arxiv.org/pdf/1506.02025.pdf) mostly for my own education and because I think it would be slick to make one in Julia via Flux + Interpolations. I…

---

## [Machine Learning Toolset Improvement](https://discourse.julialang.org/t/machine-learning-toolset-improvement/4835)

<div class="topic-metadata">

**Author:** [@naveenjafer](https://discourse.julialang.org/u/naveenjafer)\
**Replies:** 10\
**Last updated:** [December 13, 2018, 7:06pm UTC](https://discourse.julialang.org/t/machine-learning-toolset-improvement/4835 "2018-12-13T19:06:19Z")

</div>

Is there any consolidated set of machine learning packages available for Julia? I have spent some time trying to search for one, but have only come across JuliaML and it has not exactly been updated regularly. I would li…

---

## [Flux gradient clipping](https://discourse.julialang.org/t/flux-gradient-clipping/18544)

<div class="topic-metadata">

**Author:** [@Fart](https://discourse.julialang.org/u/Fart)\
**Replies:** 2\
**Last updated:** [December 11, 2018, 7:21pm UTC](https://discourse.julialang.org/t/flux-gradient-clipping/18544 "2018-12-11T19:21:04Z")

</div>

Does anyone have a working implementation of gradient clipping in Flux? I’ve tried directly modifying the param.grad field but get LoadError: type TrackedArray is immutable

---

## [Another error with defining Flux.jl gradients](https://discourse.julialang.org/t/another-error-with-defining-flux-jl-gradients/17803)

<div class="topic-metadata">

**Author:** [@xiaodai](https://discourse.julialang.org/u/xiaodai)\
**Replies:** 5\
**Last updated:** [December 5, 2018, 4:09pm UTC](https://discourse.julialang.org/t/another-error-with-defining-flux-jl-gradients/17803 "2018-12-05T16:09:50Z")

</div>

Here is the MWE x = rand(12, 1000) y = rand(1000) W = param(\[1 for i=1:12\]) a = param(\[0.5\]) b = param(\[0.5\]) model(x) = begin x1 = \[maximum(x\[:,j\] .\* W) for j in 1:size(x, 2)\] (a .+ (x1 .- StatsBase.mean(x1)).…

---

## [What fileformat to use to load data for high performance computing](https://discourse.julialang.org/t/what-fileformat-to-use-to-load-data-for-high-performance-computing/4786)

<div class="topic-metadata">

**Author:** [@Tomas\_Pevny](https://discourse.julialang.org/u/Tomas_Pevny)\
**Replies:** 37\
**Last updated:** [December 1, 2018, 3:41pm UTC](https://discourse.julialang.org/t/what-fileformat-to-use-to-load-data-for-high-performance-computing/4786 "2018-12-01T15:41:02Z")

</div>

Hi All, I am working on a project, where I want to do some machine learning over large data (about 50Gb) stored in AWS S3. I would like to ask people about their opinions and experiences with the dataformat they use and…

---

## [How to create tracked \`cumsum\` using Flux.jl?](https://discourse.julialang.org/t/how-to-create-tracked-cumsum-using-flux-jl/17772)

<div class="topic-metadata">

**Author:** [@xiaodai](https://discourse.julialang.org/u/xiaodai)\
**Replies:** 12\
**Last updated:** [December 1, 2018, 7:35am UTC](https://discourse.julialang.org/t/how-to-create-tracked-cumsum-using-flux-jl/17772 "2018-12-01T07:35:27Z")

</div>

I wanted to create a cumsum of tracked values. How do I do that? I tried the below which failed W = param(rand(12,1)) cumsum(Tracker.data(W), dims = 1) # works cumsum(W, dims = 1) # this fails I got it to work using m…

---

## [How to histogram Flux predictions?](https://discourse.julialang.org/t/how-to-histogram-flux-predictions/18194)

<div class="topic-metadata">

**Author:** [@jstrube](https://discourse.julialang.org/u/jstrube)\
**Replies:** 3\
**Last updated:** [November 30, 2018, 11:29pm UTC](https://discourse.julialang.org/t/how-to-histogram-flux-predictions/18194 "2018-11-30T23:29:23Z")

</div>

I’m training a simple regression network with Flux. I’m still learning about the optimizers, but it seems to work OK. However, when I try to make a histogram of predictions on a new data set, I’m getting m(x\[:,1:4\]) Tra…

---

## [Registering jacobians of composite types in AutoGrad.jl](https://discourse.julialang.org/t/registering-jacobians-of-composite-types-in-autograd-jl/18149)

<div class="topic-metadata">

**Author:** [@crinders](https://discourse.julialang.org/u/crinders)\
**Replies:** 2\
**Last updated:** [November 30, 2018, 1:59pm UTC](https://discourse.julialang.org/t/registering-jacobians-of-composite-types-in-autograd-jl/18149 "2018-11-30T13:59:29Z")

</div>

I am posting this question here, because AutoGrad.jl is associated with Knet. I have a derivative of the svd. svd returns an object of type SVD, with components U, Vt, and S. I want to register the derivative with Auto…

---

## [Object detection (with ROIs) in Mocha.jl?](https://discourse.julialang.org/t/object-detection-with-rois-in-mocha-jl/18023)

<div class="topic-metadata">

**Author:** [@ianshmean](https://discourse.julialang.org/u/ianshmean)\
**Replies:** 0\
**Last updated:** [November 26, 2018, 6:27pm UTC](https://discourse.julialang.org/t/object-detection-with-rois-in-mocha-jl/18023 "2018-11-26T18:27:48Z")

</div>

I’m trying to get bounding-box style object detection running through Mocha.jl and wondering if anyone is aware of any work on this front? Mocha has tidy image classification examples, but I haven’t found an object detec…

---

## [Multivariate Linear Regression in the Julia](https://discourse.julialang.org/t/multivariate-linear-regression-in-the-julia/17909)

<div class="topic-metadata">

**Author:** [@YonghyunRyu](https://discourse.julialang.org/u/YonghyunRyu)\
**Replies:** 1\
**Last updated:** [November 23, 2018, 3:25pm UTC](https://discourse.julialang.org/t/multivariate-linear-regression-in-the-julia/17909 "2018-11-23T15:25:41Z")

</div>

Hi, there! I am looking for a multivariate linear regression library in the Julia. Could anyone recommend a library for me?

---

## [Reshaping outputs between layers in Flux.jl](https://discourse.julialang.org/t/reshaping-outputs-between-layers-in-flux-jl/17838)

<div class="topic-metadata">

**Author:** [@Rademcaher](https://discourse.julialang.org/u/Rademcaher)\
**Replies:** 1\
**Last updated:** [November 22, 2018, 1:02am UTC](https://discourse.julialang.org/t/reshaping-outputs-between-layers-in-flux-jl/17838 "2018-11-22T01:02:41Z")

</div>

The input to my network are N x 2 arrays and I’d like to implement a Dense layer, followed by a convolution. This requires reshaping the output of the dense layer. The forward pass seems to work fine, but I think there’s…

---

## [Error with defining customer gradients in Flux.jl](https://discourse.julialang.org/t/error-with-defining-customer-gradients-in-flux-jl/17769)

<div class="topic-metadata">

**Author:** [@xiaodai](https://discourse.julialang.org/u/xiaodai)\
**Replies:** 9\
**Last updated:** [November 21, 2018, 7:21am UTC](https://discourse.julialang.org/t/error-with-defining-customer-gradients-in-flux-jl/17769 "2018-11-21T07:21:12Z")

</div>

I am trying to define some simple custom graidents in Flux.jl but am running into issues. Here is an MWE # start random parameters W = param(rand(12)) model(x) = begin x\*W end loss(x, y) = sum((model(x) .- y).^2) …

---

## [Reinforcement learning with A\* and a deep heuristic](https://discourse.julialang.org/t/reinforcement-learning-with-a-and-a-deep-heuristic/17480)

<div class="topic-metadata">

**Author:** [@skariel](https://discourse.julialang.org/u/skariel)\
**Replies:** 0\
**Last updated:** [November 13, 2018, 3:29pm UTC](https://discourse.julialang.org/t/reinforcement-learning-with-a-and-a-deep-heuristic/17480 "2018-11-13T15:29:17Z")

</div>

100% Julia - see paper here: https://github.com/imagry/aleph\_star/blob/master/paper/aleph\_star.pdf and code here: https://github.com/imagry/aleph\_star now the real reason I’m writing this here is since I need help wi…

---

## [Natural Language Processing: where do I start?](https://discourse.julialang.org/t/natural-language-processing-where-do-i-start/15509)

<div class="topic-metadata">

**Author:** [@johann.spies](https://discourse.julialang.org/u/johann.spies)\
**Replies:** 3\
**Last updated:** [November 10, 2018, 10:09am UTC](https://discourse.julialang.org/t/natural-language-processing-where-do-i-start/15509 "2018-11-10T10:09:27Z")

</div>

I do not know enough about machine learning to ask an informed question. I am not even sure whether this a Machine Learning question. Maybe. I will tell you what I want to achieve in the end: In short we want to clas…

---

## [Knet vs Flux etc](https://discourse.julialang.org/t/knet-vs-flux-etc/17057)

<div class="topic-metadata">

**Author:** [@denizyuret](https://discourse.julialang.org/u/denizyuret)\
**Replies:** 10\
**Last updated:** [November 9, 2018, 3:33pm UTC](https://discourse.julialang.org/t/knet-vs-flux-etc/17057 "2018-11-09T15:33:30Z")

</div>

We have a rich collection of AD, GPU and ML tools and it may be useful to have common benchmarks to find performance bottlenecks, share optimization tricks and push them forward. I started fiddling around in Klutz.jl, so…

---

## [Porting a RNN model to Flux from PyTorch](https://discourse.julialang.org/t/porting-a-rnn-model-to-flux-from-pytorch/16487)

<div class="topic-metadata">

**Author:** [@sdwfrost](https://discourse.julialang.org/u/sdwfrost)\
**Replies:** 5\
**Last updated:** [October 29, 2018, 12:04pm UTC](https://discourse.julialang.org/t/porting-a-rnn-model-to-flux-from-pytorch/16487 "2018-10-29T12:04:19Z")

</div>

Hi All, I’m trying to port this example of a recurrent neural network in PyTorch to Flux to help me learn the API. I know that I’m not putting the data together with the loss function in the right way (I’m using the cha…

---

## [Flux.jl RNN performance](https://discourse.julialang.org/t/flux-jl-rnn-performance/16803)

<div class="topic-metadata">

**Author:** [@ak168421](https://discourse.julialang.org/u/ak168421)\
**Replies:** 11\
**Last updated:** [October 28, 2018, 10:18pm UTC](https://discourse.julialang.org/t/flux-jl-rnn-performance/16803 "2018-10-28T22:18:36Z")

</div>

I have been trying out Flux.jl for RNNs and I really like the syntax. I am wondering how I should expect performance to compare to, say PyTorch and whether there are things I should look out for in terms of defining mode…

---

## [Article that explains using Flux on MNIST from the model-zoo example](https://discourse.julialang.org/t/article-that-explains-using-flux-on-mnist-from-the-model-zoo-example/16295)

<div class="topic-metadata">

**Author:** [@microgold](https://discourse.julialang.org/u/microgold)\
**Replies:** 4\
**Last updated:** [October 27, 2018, 5:24am UTC](https://discourse.julialang.org/t/article-that-explains-using-flux-on-mnist-from-the-model-zoo-example/16295 "2018-10-27T05:24:43Z")

</div>

I created an article that attempts to explain the Flux example of a Convolutional Neural Network operating on the MNIST set. The article goes through each line and explains what its doing in Julia. I’d appreciate any i…

---

## [Deep Reinforcement Learning Example that will build on Julia 1.0.1?](https://discourse.julialang.org/t/deep-reinforcement-learning-example-that-will-build-on-julia-1-0-1/16656)

<div class="topic-metadata">

**Author:** [@Jack\_Vice](https://discourse.julialang.org/u/Jack_Vice)\
**Replies:** 6\
**Last updated:** [October 23, 2018, 10:17pm UTC](https://discourse.julialang.org/t/deep-reinforcement-learning-example-that-will-build-on-julia-1-0-1/16656 "2018-10-23T22:17:54Z")

</div>

Hi, Can someone recommend a good Deep Reinforcement Learning example code that will build on Julia 1.0.1. I am looking for Actor-Critic, policy gradient or even deep q learning example that works so I can then start doin…

---

## [\[Solved\]Flux.jl: can it be used to train networks with multiple distinct inputs?](https://discourse.julialang.org/t/solved-flux-jl-can-it-be-used-to-train-networks-with-multiple-distinct-inputs/16498)

<div class="topic-metadata">

**Author:** [@thvhauwe](https://discourse.julialang.org/u/thvhauwe)\
**Replies:** 1\
**Last updated:** [October 22, 2018, 2:57pm UTC](https://discourse.julialang.org/t/solved-flux-jl-can-it-be-used-to-train-networks-with-multiple-distinct-inputs/16498 "2018-10-22T14:57:02Z")

</div>

Similar network in Keras: The Functional API I’m getting errors when running the following basic example code trying to train a network with two inputs and 1 output. The network is made up out of a left branch and a rig…

---

## [ML demo code](https://discourse.julialang.org/t/ml-demo-code/16547)

<div class="topic-metadata">

**Author:** [@microlifecc](https://discourse.julialang.org/u/microlifecc)\
**Replies:** 9\
**Last updated:** [October 21, 2018, 7:09pm UTC](https://discourse.julialang.org/t/ml-demo-code/16547 "2018-10-21T19:09:21Z")

</div>

I am new to Machine Learning in Julia and am going through the tutorial here. The code for the tutorial is here. While trying to run the following line I am getting an error: Metalhead.imagenet\_classes\[rand(1:1000,1,1)…

---

## [DifferentialEquations.jl + Flux.jl or Knet.jl](https://discourse.julialang.org/t/differentialequations-jl-flux-jl-or-knet-jl/8208)

<div class="topic-metadata">

**Author:** [@nbren12](https://discourse.julialang.org/u/nbren12)\
**Replies:** 15\
**Last updated:** [October 18, 2018, 2:50pm UTC](https://discourse.julialang.org/t/differentialequations-jl-flux-jl-or-knet-jl/8208 "2018-10-18T14:50:54Z")

</div>

I am trying to estimate various dynamical systems using neural network models. I am very interested in using the integrators provided by DifferentialEquations.jl to do this, and the docs for that package even have nice e…

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

[Next page](https://discourse.julialang.org/c/domain/ml/24.md?page=54)