dotnet / machinelearning-samples

Samples for ML.NET, an open source and cross-platform machine learning framework for .NET.

Note: We'd love to hear your thoughts about MLOps. Let us know in this survey.

ML.NET Samples

ML.NET is a cross-platform open-source machine learning framework that makes machine learning accessible to .NET developers.

In this GitHub repo, we provide samples which will help you get started with ML.NET and how to infuse ML into existing and new .NET apps.

Note: Please open issues related to ML.NET framework in the Machine Learning repository. Please create the issue in this repo only if you face issues with the samples in this repository.

There are two types of samples/apps in the repo:

• Getting Started : ML.NET code focused samples for each ML task or area, usually implemented as simple console apps.

• End-End apps : End-user sample web and desktop apps infused with Machine Learning models based on ML.NET.

The official ML.NET samples are divided in multiple categories depending on the scenario and machine learning problem/task, accessible through the following tables:

Binary classification
Sentiment Analysis C#     F#	Spam Detection C#     F#	Credit Card Fraud Detection (Binary Classification) C#    F#
Heart Disease Prediction C#
Multi-class classification
Issues Classification C#  F#	Iris Flowers Classification C#    F#	MNIST C#
Recommendation
Product Recommendation C#	Movie Recommender (Matrix Factorization) C#	Movie Recommender (Field Aware Factorization Machines) C#
Regression
Price Prediction C#     F#	Sales Forecasting (Regression) C#	Demand Prediction C#    F#
Time Series Forecasting
Sales Forecasting (Time Series) C#
Anomaly Detection
Sales Spike Detection  C#       C#	Power Anomaly Detection C#	Credit Card Fraud Detection (Anomaly Detection) C#
Clustering
Customer Segmentation C#     F#	IRIS Flowers Clustering C#     F#
Ranking
Rank Search Engine Results C#
Computer Vision
Image Classification Training (High-Level API)  C# F#	Image Classification Predictions (Pretrained TensorFlow model scoring)  C#   F#        C#	Image Classification Training (TensorFlow Featurizer Estimator)  C#   F#
Object Detection (ONNX model scoring)  C#       C#

Cross Cutting Scenarios
Scalable Model on WebAPI C#	Scalable Model on Razor web app C#	Scalable Model on Azure Functions C#
Scalable Model on Blazor web app C#	Large Datasets C#	Loading data with DatabaseLoader C#
Loading data with LoadFromEnumerable C#	Model Explainability C#	Export to ONNX C#

Automate ML.NET models generation (Preview state)

The previous samples show you how to use the ML.NET API 1.0 (GA since May 2019).

However, we're also working on simplifying ML.NET usage with additional technologies that automate the creation of the model for you so you don't need to write the code by yourself to train a model, you simply need to provide your datasets. The "best" model and the code for running it will be generated for you.

These additional technologies for automating model generation are in PREVIEW state and currently only support Binary-Classification, Multiclass Classification and Regression. In upcoming versions we'll be supporting additional ML Tasks such as Recommendations, Anomaly Detection, Clustering, etc..

CLI samples: (Preview state)

The ML.NET CLI (command-line interface) is a tool you can run on any command-prompt (Windows, Mac or Linux) for generating good quality ML.NET models based on training datasets you provide. In addition, it also generates sample C# code to run/score that model plus the C# code that was used to create/train it so you can research what algorithm and settings it is using.

CLI (Command Line Interface) samples
Binary Classification sample
MultiClass Classification sample
Regression sample

AutoML API samples: (Preview state)

THESE SAMPLES USE THE 0.1.x VERSION OF THE AUTOML API. WHILE THESE APIS STILL WORK IN VERSION 0.2.x WE RECOMMEND USING THE NEW APIS INTRODUCED IN 0.2.x AND LATER. FOR 0.2.x SAMPLES, SEE ML.NET 2.0 Samples.

ML.NET AutoML API is basically a set of libraries packaged as a NuGet package you can use from your .NET code. AutoML eliminates the task of selecting different algorithms, hyperparameters. AutoML will intelligently generate many combinations of algorithms and hyperparameters and will find high quality models for you.

AutoML API samples
Binary Classification sample
MultiClass Classification sample
Ranking sample
Regression sample
Advanced experiment sample

---

Additional ML.NET Community Samples

In addition to the ML.NET samples provided by Microsoft, we're also highlighting samples created by the community showcased in this separated page: ML.NET Community Samples

Those Community Samples are not maintained by Microsoft but by their owners. If you have created any cool ML.NET sample, please, add its info into this REQUEST issue and we'll publish its information in the mentioned page, eventually.

Translations of Samples:

• Chinese Simplified

Learn more

See ML.NET Guide for detailed information on tutorials, ML basics, etc.

API reference

Check out the ML.NET API Reference to see the breadth of APIs available.

Contributing

We welcome contributions! Please review our contribution guide.

Community

Please join our community on Gitter

This project has adopted the code of conduct defined by the Contributor Covenant to clarify expected behavior in our community. For more information, see the .NET Foundation Code of Conduct.

License

ML.NET Samples are licensed under the MIT license.

machinelearning-samples's Issues

StopWordRemover in conjunction with FeaturizeText

Hi

Looking through the text related samples (such as SentimentAnalysis), I observed that non has used stop word remover to preprocess the text before transforming and training.

May I request for samples on implementing stop word remover for text analysis?

[Request for the Community] Add your ML.NET sample to the Community-samples page

Do you have any interesting ML.NET sample that you'd like to share with the community?
If so, you can respond to this issue with info about your ML.NET samples.

Eventually, when reviewed, it'll be showcased in the COMMUNITY-SAMPLES page:

Thanks for your contribution! :)

How to get a predicted group list output?

is there a way to output grouped items using the prediction algorithm and clustering?
here is an example i have a list of 50 employees and i want to put them into groups of 10 by predicting which employees best suited to work together in a group and then output those 5 groups with 10 employees each.

the elements that will be used to make the prediction will be (

• experience from 1 - 10 years experience,
• easy to work with rated by yes or no options
• Visual learner vs non visual learner
• and if the employee is a quick learner or not

thank you in advance :)

,

Review samples for correct data science approach and ML.NET API usage

Status

Folder	Sample	Data Science Review	API Review
C#\getting_started	BinaryClassification_ CreditCardFraudDetection	OK (#96)	OK (v0.8)
C#\getting_started	BinaryClassification_ SentimentAnalysis	OK	OK (v0.8)
C#\getting_started	Clustering_ CustomerSegmentation	OK (#95)	OK (v0.8)
C#\getting_started	Clustering_ Iris	OK (#109)	OK (v0.8)
C#\getting_started	MulticlassClassification_ Iris	OK	OK (v0.8)
C#\getting_started	Regression_ BikeSharingDemand	OK	OK (v0.8)
C#\getting_started	Regression_ TaxiFarePrediction	OK (#95)	OK (v0.8)
C#\getting_started	DeepLearning_ ImageClassification_ TensorFlow	Pixel data preprocessing needed or not? Also, info/github page for the used inception model should be included	OK (v0.8) #155
C#\getting_started	DeepLearning_ TensorFlowEstimator	Same as above	Still 0.7
C#\getting_started	MatrixFactorization_ MovieRecommendation	OK	Still 0.7
C#\end-to-end-apps	MulticlassClassification- GitHubLabeler	OK (#96)	OK (v0.8)
C#\end-to-end-apps	Recommendation- MovieRecommender	OK	Still 0.7
C#\end-to-end-apps	Regression- SalesForecast	OK	OK (v0.8)
C#\getting_started	AnomalyDetection- Sales	OK	OK (v0.11)

DS Review

BinaryClassification_CreditCardFraudDetection

• Data preprocessing: OK (MeanVar normalization)
• Feature engineering: Not needed (input data are PCA dimensions)
• Learner: Ok
• Training: Ok
• Scoring: Ok
• Metrics: Ok
  • Accuracy: 99.9%
  • Auc: 97.5%
  • F1Score: 77.7%

BinaryClassification_SentimentAnalysis

• Data preprocessing: Not needed
• Feature engineering: Ok (Text -> Feature vector using TextTransform)
• Learner: Ok
• Training: Ok (single training on training data)
• Scoring: Ok
• Metrics: Ok
  • Accuracy: 72%
  • Auc: 97%
  • F1Score: 78%

Clustering_CustomerSegmentation

• Data preprocessing: Ok (Join and pivot tables using Linq)
• Feature engineering: Ok (Counting offers per customers)
• Learner: Ok
• Training: Why PcaEstimator seed is 42 (magic number)?
• Scoring: Ok
• Metrics: Ok
  • AvgMinScore: 2.3
  • Dbi: 2.5

Clustering_Iris

• Data loading: Ok (Can load all the numeric values as vector instead of individually and then concatenating them. But keeping it as is for educational purposes.)
• Data preprocessing: Ok (not needed)
• Feature engineering: Ok (not needed)
• Learner: Ok
• Training: Ok
• Scoring: Ok
• Metrics:
  • AvgMinScore: 0.564
  • DBI: 0.955

MulticlassClassification_Iris

• Data loading: Ok (Can load all the numeric values as vector instead of individually and then concatenating them. But keeping it as is for educational purposes.)
• Data preprocessing: Ok (not needed)
• Feature engineering: Ok (not needed)
• Learner: Ok
• Training: Ok
• Scoring: Ok
• Metrics: Ok (Accuracy is 1 because of small test set)

Regression_BikeSharingDemand

Ok.

Regression_TaxiFarePrediction

• Data loading: Ok.
• Data preprocessing: Ok (not needed)
• Feature engineering: Ok (Categorical transform for text columns)
• Learner: Ok
• Training: Ok
• Scoring: Ok
• Metrics: LossFn needs to be removed from outputted metrics. It's the same as L2 b/c no custom loss function is defined.
  • R2 Score: 0.7
  • RMS loss: 5.97
  • Absolute loss: .99

MatrixFactorization_MovieRecommendation

MF using MFTrainer. Evaluation done as regressions.

MulticlassClassification-GitHubLabeler

• Data loading: Ok.
• Data preprocessing: Ok (not needed)
• Feature engineering: Ok (Categorical transform for text columns)
• Learner: Ok
• Training: Ok
• Scoring: Ok
• Metrics: Ok
  • MicroAcuracy (Avg): 71.8%
  • MacroAccuracy (Avg): 50.3%
  • LogLoss: 1.076
  • LogLossReduction: 56.2

Regression-SalesForecast (eShopDashboardML)

• Data loading: Ok.
• Data preprocessing: Ok (not needed)
• Feature engineering: Ok (Categorical transform for text columns)
• Learner: Ok
• Training: Ok
• Scoring: Ok
• Metrics: Ok
  • Product model:
    • L1 Loss: 96.5
    • L2 Loss: 74493.7
    • RMS: 96.5
    • R-squared: 56.6%
  • Country model:
    • L1 Loss: 0.446
    • L2 Loss: 0.386
    • RMS: 0.446
    • R-squared: 45.3%

AnomalyDetection-Sales

• Data loading: Ok.
• Data preprocessing: Ok (not needed)
• Learner: Ok
• Training: Ok
• Scoring: Ok
• Metrics: there is no evaluation in timeseries- spike detection and change point detection algorithm

eShopDashboardML duplicates forecast requests

when plotting forecasts, the forecast controller is called twice with no reason

LightGBM expects double like "0.2" but ML.net is using the local CultureInfo

When using LightGBM, ML.net is passing parameters with the local Culture (in my case "0,2") instead of en or us Culture.

This leads to an Error when calling the Train method

in WrappedLightGbmInterface.cs

Clustering_Iris sample code - column SepalWidth is defined twice

In the TextLoader.Arguments column SepalWidth is defined twice in line 34 (second definition is way to the right):

new TextLoader.Column("SepalWidth", DataKind.R4, 2),                                     new TextLoader.Column("SepalWidth", DataKind.R4, 2),

Is there any guide to choose the Algorithms

For a new member to machinelearning, I found there are multiple algorithms for the same scenario like Regression, there are Sdca Regression, FastTreeTweedie Regression, Fast Tree and etc.

How to choose which algorithms to use?

Or, we train the model with all the available algorithms and to choose the best one per to the Evaluate result.

Is there any doc to describe the algorithms, their advantages and disadvantages and when to use them?

F# samples problems in Visual Studio 2017 15.7.6

On opening samples\fsharp\getting-started\GettingStarted.sln a message box is shown:

Internal error: Critical capabilities changes were detected without any change
made to the project 'Regression_TaxiFarePrediction.fsproj', and forces the
project to be reloaded.  Potential capabilities involved: 'FSharp'.

After that the solution and its projects are loaded with the following oddities:

• In the solution explorer projects miss their Dependencies.
• Building gives errors like:

    1>------ Build started: Project: Regression_TaxiFarePrediction, Configuration: Debug Any CPU ------
    1>C:\Program Files\dotnet\sdk\2.1.302\Sdks\Microsoft.NET.Sdk\targets\Microsoft.PackageDependencyResolution.targets(198,5):
    error : Assets file 'C:\-\ML\machinelearning-samples\samples\fsharp\getting-started\Regression_TaxiFarePrediction\obj\project.assets.json' not found.
    Run a NuGet package restore to generate this file.
    1>Done building project "Regression_TaxiFarePrediction.fsproj" -- FAILED.

Restoring and building from the command line using dotnet works fine.

Adding F# samples

I'm looking at adding some initial F# samples

Should the pivot be
samples\CSharp\getting-started\BinaryClassification_SentimentAnalysis
samples\FSharp\getting-started\BinaryClassification_SentimentAnalysis
or
samples\getting-started\BinaryClassification_SentimentAnalysis\CSharp
samples\getting-started\BinaryClassification_SentimentAnalysis\FSharp

I don't mind either way. The former is probably better

• PRO: you choose a language, then you get the sample set

• PRO: if community provide samples they are likely to be in one language

• CON: it means the sample sets are more likely will diverge over time

• PRO: if samples are missing for one language it doesn't really matter, you don't really notice

Any ONNX scoring examples?

We can find how to use ONNX scoring API from https://github.com/dotnet/machinelearning/blob/76dd9235f348aee4fdf7e62a9edf27510f14769b/test/Microsoft.ML.OnnxTransformTest/OnnxTransformTests.cs. However, is there any ONNX scoring examples with some explanation?

In GitHubLabeler project, TextLoader should be made `useHeader` is true

machinelearning-samples/samples/end-to-end-apps/github-labeler/GitHubLabeler/Predictor.cs

Line 25 in 2bf33da

pipeline.Add(new TextLoader(DataPath).CreateFrom<GitHubIssue>());

should be:

pipeline.Add(new TextLoader(DataPath).CreateFrom<GitHubIssue>(useHeader: true));

if useHeader=false that means in corefx-issues-train.tsv data file , a first row Area label will be as a new category and total number of category is 23, not 22.

F# samples with build errors

@mariuszwojcik - When possible for you, can you take a look to the F# samples? I'm getting build errors in most of the F# samples, since this week.
Thanks,

typo

https://github.com/dotnet/machinelearning-samples/tree/master/samples/examples
Reggression -> Regression

Samples with Object Pooling of PredictionFunction objects when scoring in ASP.NET Core apps

Related to this issue at ML.NET: dotnet/machinelearning#1789 (comment) and the discussions here: dotnet/machinelearning#1718

All the Web based end-user apps using ASP.NET Core should follow a similar approach:

• eShopDashboard - Sales Forecast
• Movies recommender
• Any other ASP.NETCore MVC, Razor or Web API sample

F# samples need to be migrated to the new "Dynamic API available since v0.6"

All the F# samples in this repo are still using the deprecated LearningPipeline API and need to be migrated to the new "Dynamic API", like we're doing with the C# samples.

Contributions from the community will be very much welcomed! 👍

Visual Studio 2017 (15.7) build error

Use Visual Studio 2017 (15.7) to open GettingStarted.sln, compile, and quote the following error: Error NU1100: Cannot parse.
Netcoreapp,version=v2.0 's "microsoft.ml (>= 0.2.0)" Using dotnet restore is also the same error

F:\GitHub\machinelearning-samples\samples\getting-started>dotnet restore
正在还原 F:\GitHub\machinelearning-samples\samples\getting-started\Clustering_Iris\Clustering_Iris.csproj 的包...
正在还原 F:\GitHub\machinelearning-samples\samples\getting-started\BinaryClassification_SentimentAnalysis\BinaryClassification_SentimentAnalysis.csproj 的包...
正在还原 F:\GitHub\machinelearning-samples\samples\getting-started\MulticlassClassification_Iris\MulticlassClassification_Iris.csproj 的包...
正在还原 F:\GitHub\machinelearning-samples\samples\getting-started\Regression_TaxiFarePrediction\Regression_TaxiFarePrediction.csproj 的包...
F:\GitHub\machinelearning-samples\samples\getting-started\Regression_TaxiFarePrediction\Regression_TaxiFarePrediction.csproj : error NU1100: 无法解析 .NETCoreApp,Version=v2.0 的“Microsoft.ML (>= 0.2.0)”。 [F:\GitHub\machinelearning-samples\samples\getting-started\GettingStarted.sln]
F:\GitHub\machinelearning-samples\samples\getting-started\MulticlassClassification_Iris\MulticlassClassification_Iris.csproj : error NU1100: 无法解析 .NETCoreApp,Version=v2.0 的“Microsoft.ML (>= 0.2.0)”。 [F:\GitHub\machinelearning-samples\samples\getting-started\GettingStarted.sln]
F:\GitHub\machinelearning-samples\samples\getting-started\BinaryClassification_SentimentAnalysis\BinaryClassification_SentimentAnalysis.csproj : error NU1100: 无法解析 .NETCoreApp,Version=v2.0 的“Microsoft.ML (>= 0.2.0)”。 [F:\GitHub\machinelearning-samples\samples\getting-started\GettingStarted.sln]
F:\GitHub\machinelearning-samples\samples\getting-started\Clustering_Iris\Clustering_Iris.csproj : error NU1100: 无法解析 .NETCoreApp,Version=v2.0 的“Microsoft.ML (>= 0.2.0)”。 [F:\GitHub\machinelearning-samples\samples\getting-started\GettingStarted.sln]
F:\GitHub\machinelearning-samples\samples\getting-started\MulticlassClassification_Iris\MulticlassClassification_Iris.csproj 的还原在 175.21 ms 内失败。
F:\GitHub\machinelearning-samples\samples\getting-started\Clustering_Iris\Clustering_Iris.csproj 的还原在 175.21 ms 内失败。
F:\GitHub\machinelearning-samples\samples\getting-started\Regression_TaxiFarePrediction\Regression_TaxiFarePrediction.csproj 的还原在 175.21 ms 内失败。
F:\GitHub\machinelearning-samples\samples\getting-started\BinaryClassification_SentimentAnalysis\BinaryClassification_SentimentAnalysis.csproj 的还原在 175.21 ms 内失败。

Why sentiment "Muppets songs were so cool" is negative

I tried with sentiment "Muppets songs were so cool"
and output was negative
training data is same given in dataset

StochasticDualCoordinateAscentClassifier Issue With 86 bit

when i use StochasticDualCoordinateAscentClassifier Issue With 86 bit

it through the following exception
'A call to PInvoke function 'Microsoft.ML.CpuMath!Microsoft.ML.Runtime.Internal.CpuMath.Thunk::SumSqU' has unbalanced the stack. This is likely because the managed PInvoke signature does not match the unmanaged target signature. Check that the calling convention and parameters of the PInvoke signature match the target unmanaged signature.'

but when used it with 64 it work

unfortunately i need to integrate it with 32 bit application

any advice pleas e

this sample of my code

         `   pipeline.Add(CollectionDataSource.Create(noramalTagsTrainingData));

            pipeline.Add(new Dictionarizer("Label"));
            // Transform any text feature to numeric values
            pipeline.Add(new CategoricalOneHotVectorizer(
               "fontColor",
               "tagText",
               "firstWord"
               ));

pipeline.Add(new ColumnConcatenator(
"Features",
"fontSize",
"isBold",
"isItalic",
"isUnderLine",
"containsDot",
"containsQuestionMark",
"isAllCaps",
"tagText",
"firstWord"
));

            pipeline.Add(new StochasticDualCoordinateAscentClassifier
            {
                MaxIterations = 100,
                NumThreads = 7,
                LossFunction = new SmoothedHingeLossSDCAClassificationLossFunction()
            });

pipeline.Add(new PredictedLabelColumnOriginalValueConverter() { PredictedLabelColumn = "PredictedLabel" });

            // The pipeline is trained on the dataset that has been loaded and transformed.
            var model = pipeline.Train<NormalTagsModelFeatures, NormalTagsPrediction>();`

it through the exception at last line

Confidence Interval for CV metrics should be calculated using t-distribution quantiles

In samples\csharp\common\ConsoleHelpers.cs , the 95% confidence interval upper bound for CV metrics is calculated using 1.96, which is the 97.5% quantile of the standard normal distribution. We should instead use the 97.5% quantile of the Student's t distribution with (n - 1) degrees of freedom, where n is the number of folds, because we use the sample standard deviation.

        public static double CalculateConfidenceInterval95(IEnumerable<double> values)
        {
            double confidenceInterval95 = 1.96 * CalculateStandardDeviation(values) / Math.Sqrt(values.Count());
            return confidenceInterval95;
        }

Moreover, the standard deviation calculation should use the formula for the sample instead of the population, i.e. use (n - 1) instead of n in the denominator.

        public static double CalculateStandardDeviation (IEnumerable<double> values)
        {
            double average = values.Average();
            double sumOfSquaresOfDifferences = values.Select(val => (val - average) * (val - average)).Sum();
            double standardDeviation = Math.Sqrt(sumOfSquaresOfDifferences / values.Count());
            return standardDeviation;
        }

@shmoradims @CESARDELATORRE @justinormont can we use either Accord.NET or Math.NET for the t-distribution inverse CDF function? Any potential licensing issues with either? Accord.NET is GNU Lesser GPL 2.1, Math.NET Numerics is MIT/X11.

Translate into Chinese

I plan to translate machinelearning-samples into Chinese and introduce ML.NET to Chinese programmers.

I created a repo and translated the first README.md.

Can I create a PR to add link in original English page or upload README.zh-cn.md ?

creditcard.csv data loading error in VB.NET sample

In CreditCardFraudDetection.Trainer project (VB.NET version), under input folder the creditcard.csv file is generated using the unzip feature in code. Works fine.
In ConsoleHelpers.vb file, line 100 we get an error "Parsing failed with an exception: Stream reading encountered exception". Detailed Inner Exception as below.

Could not open file 'D:\Source\Repos\machinelearning-samples\samples\visualbasic\getting-started\BinaryClassification_CreditCardFraudDetection\CreditCardFraudDetection.Trainer\bin\Debug\netcoreapp2.1......\assets'. Error is: Access to the path 'D:\Source\Repos\machinelearning-samples\samples\visualbasic\getting-started\BinaryClassification_CreditCardFraudDetection\CreditCardFraudDetection.Trainer\assets' is denied.

I'm not a local admin in this PC. creditcard.csv is created from the given zip file by the program running in my user id. I gave full control for the entire root folder. But still no use. How to get rid of "System.UnauthorizedAccessException"? I understand this is an OS issue not a program issue. But the program has created this file and the program itself is unable to read it. Please help.

Need example how to load using IDataView using IEnumerable

I need an sample how to load IDataView from an IEnumerable, all examples are using a text csv file, i need to load the IDataView from a context object or a database table. Please help providing an sample.

BikeSharingDemand is still using ML.NET 0.2.0

Just to mention that the sample

samples\examples\Reggression_BikeSharingDemands\BikeSharingDemand

is still using ML.NET 0.2.0.

Sample Requirement for Field Aware Factorization Machines

For Movie Recommender, it describe that using Field Aware Factorization Machines to add more features like UserId, ProductId, Ratings, Product Description, Product Price etc, but for the demo project, is just analyse the useid and movieid.

Is there any complete demo to take more features in Recommendation demo?

Namesapces in Samples do not exist

The following Namespaces for the C# samples do not exist.
Where can I find them?

using Microsoft.ML.Core.Data;
using Microsoft.ML.Transforms.Categorical;
using static Microsoft.ML.Transforms.Normalizers.NormalizingEstimator;

TaxiFarePrediction sample gets Error MS3491 File Path/Name are too long

Severity Code Description Project File Line Suppression State
Error MSB3491 Could not write lines to file "obj\Debug\netcoreapp2.0\Regression_TaxiFarePrediction.csproj.FileListAbsolute.txt". The specified path, file name, or both are too long. The fully qualified file name must be less than 260 characters, and the directory name must be less than 248 characters. Regression_TaxiFarePrediction D:\Program Files (x86)\Microsoft Visual Studio\2017\Professional\MSBuild\15.0\Bin\Microsoft.Common.CurrentVersion.targets 5070

Bounding boxes while classifying images with ML.NET?

Hi there,

Does ML.NET support extracting bounding boxes for predictions when classifying images?

Redundant Contributing section in CONTRIBUTING.md

CONTRIBUTING.md has a Contributing section within itself with a broken link pointing to another contributing.md .

❌

README.md already has a contributing section with proper link to the actual CONTRIBUTING.md

✔️

I propose removing the redundant section with the wrong link from CONTRIBUTING.md

Custom model for image classification

In the sample DeepLearning_ImageClassification_TensorFlow I try to create my own model folowing this:
https://www.tensorflow.org/hub/tutorials/image_retraining

But when I execute the sample I get:

EXCEPTION
#########
Input column 'input' does not exist in the model
Parameter name: inputs

F# samples need to be migrated to ML.NET v0.7 (minor migration from v0.6)

We released v0.7 today (Nov. 6th) and there are a few minor breaking changes compared to v0.6 in some Estimators naming, most of all.
For now, I'm setting a special "Directory.Build.props" file targeting v0.6 for the F# samples until they are migrated to v0.7.

C# samples are most of them already migrated to v0.7, so contributors for F# can check it out in regards naming, etc.

[Discussion] Profusion of subfolders and approaches for samples structure

https://github.com/dotnet/machinelearning-samples/tree/master/samples
has three subfolders, each with few examples.

Sorting of the folders is alphabetical, not in order of difficutly.

Suggestion: Flatten this hierarchy as follows.

101_BinaryClassification_SentimentAnalysis
102_Clustering_Iris
103_MulticlassClassification_Iris
201_Regression_TaxiFarePrediction
202_BinaryClasification_Titanic
203_Regression_BikeSharingDemands
301_eShopDashboardML
302_GitHubLabeler

[BUG] CreditCard-Fraud-Detection - If “creditcard.csv” already exists, error when trying to create it from the compressed/zip file

If “creditcard.csv” already exists, error when trying to create it from the compressed/zip file in the following code.

machinelearning-samples/samples/csharp/getting-started/BinaryClassification_CreditCardFraudDetection/CreditCardFraudDetection.Common/ConsoleHelpers.cs

Line 137 in 8ccb265

public static void UnZipDataSet(string zipDataSet, string destinationFile)

    public static void UnZipDataSet(string zipDataSet, string destinationFile)
    {
        if (!File.Exists(destinationFile))
        {
            var destinationDirectory = Path.GetDirectoryName(destinationFile);
            ZipFile.ExtractToDirectory(zipDataSet, $"{destinationDirectory}");
        }
    }

Weird, the file does exist but the API says it does not.. Need to investigate this issue.

3 F# samples fail due to unhandled exceptions

Hopefully I miss something obvious, please point me to the right direction.
3 F# samples fail due to Unhandled System.ArgumentOutOfRangeException.

Namely,

BinaryClassification_SentimentAnalysis

It works fine.

Clustering_Iris, MulticlassClassification_Iris

They fail:

> dotnet run
=============== Training model ===============
Automatically adding a MinMax normalization transform, use 'norm=Warn' or 'norm=No' to turn this behavior off.

Unhandled Exception: System.Reflection.TargetInvocationException: Exception has been thrown by the target of an invocation. ---> System.ArgumentOutOfRangeException:
 Training feature column 'Features' must be a known-size vector of R4, but has type: Vec<R8, 4>.
Parameter name: data

Regression_TaxiFarePrediction

It fails:

> dotnet run
=============== Training model ===============

Unhandled Exception: System.Reflection.TargetInvocationException: Exception has been thrown by the target of an invocation. ---> System.ArgumentOutOfRangeException:
 Column 'Features' has invalid source types: All source columns must have the same type. Source types: 'Vec<R4, 3>, Vec<R4, 7>, R8, R8, Vec<R4, 6>'.
Parameter name: Column

Move at least F# samples to Paket

Paket is default package manager for almost every F# project. If you skip few Xamarin once.

Instead of going into discussion that which one is better, I just like to propose paket for at least F# samples purely based on popularity within community.

If C# samples or ML.Net like to add paket support I am very much happy to help in that. But for now, for this repo it would be great if at least F# samples moved to paket.

cc/ @dsyme @CESARDELATORRE @mariuszwojcik

ml 0.6

please convert these samples to the new 0.6 api

Broken image

Broken image at page:
https://github.com/dotnet/machinelearning-samples/blob/master/samples/csharp/getting-started/BinaryClassification_SentimentAnalysis/README.md
in Solution paragraph
current link:
https://github.com/dotnet/machinelearning-samples/raw/master/samples/getting-started/shared_content/modelpipeline.png

Path for Model Load and Save do not work out of the box on Mac

I was testing Github Labeler and Recommendation samples on a Mac. The file load and save path don' work out of the box. The relative path is incorrect on macOS/Linux.
This may be applicable to all samples in the repo.

LocalEnvironment could load it due to low access level

LocalEnvironment could load it due to low access level

any help please

Database example

Purpose

There are requests for ML.Net to support database integration to retrieve data. While we do not have database support, we do provide a way to create an IDataView from an IEnumerable, which provides an open-ended way to integrate data into an ML.Net pipeline.

Therefore I am proposing that we provide an example of how to integrate a database as a datasource into an ML.Net pipeline using ML.Net's interface to create an IDataView from an IEnumerable. A working example would give users a starting point for their own integration. Having an example will allow us to understand the performance and limitations with using a database as a datasource.

Entity Framework

Rather than provide a database specific implementation (like ODBC or SQL), this example will use the Entity Framework. The Entity Framework would cover a number of databases and is cross plat as it is supported in .net core. This also provides a common framework that .net developers will most likely be familiar with.

Entity Framework will also give configuration options that we can experiment with to determine what provides the best performance when using ML.Net. An example of this is eager loading vs lazy loading.

More information about the Entity Framework can be found here:
https://docs.microsoft.com/en-us/ef/

References

Here are the issues from ML.Net requesting/asking about database support:
dotnet/machinelearning#1130
dotnet/machinelearning#107
dotnet/machinelearning#96

How to use Feature selection

I want to use FeatureSelectorByMutualInformation feature selection, can you please provide me a sample code snippet how to use it with a dataset, i could not able find in the samples provided.

[Enhancement] Use NameOf for References to columns in all samples

Hi!

I noticed in the eShopForecastModelsTrainer that the code uses NameOf(model.property) for all of the column references.

I think this is a great idea and I'm wondering if that should be an encouraged pattern by updating all of the existing samples to follow that pattern.

Error when trying to build solution: Multiple assemblies with equivalent identity have been imported

Hi! I'm unable to build the solution successfully, here's what I did:

• downloaded the repo in zip format
• unzipped to a local folder
• opened the solution with VS Studio 2017 (v15.5.2; .Net framework 4.7.03056; Windows 10; 64 bit)
• built solution

Immediately errors appeared. Here's the first one:

Severity Code Description Project File Line Suppression State
Error CS1703 Multiple assemblies with equivalent identity have been imported: 'C:\Users\hello.nuget\packages\microsoft.netcore.app\2.0.0\ref\netcoreapp2.0\System.Reflection.Emit.Lightweight.dll' and 'C:\Users\hello.nuget\packages\system.reflection.emit.lightweight\4.3.0\ref\netstandard1.0\System.Reflection.Emit.Lightweight.dll'. Remove one of the duplicate references. BinaryClassification_SentimentAnalysis C:\Projects\machinelearning-samples-master\samples\getting-started\BinaryClassification_SentimentAnalysis\CSC 1 Active

Please advise, thank you!

TensorFlow demo does not compile

Hi Team

TensorFlow demo does not compile properly because it requires ML.NET 0.6 Preview version which is not available in NuGet (even if I select 'Include Prelease' option)

Please have a look at it.

Thanks

Add VB code examples

I want to help VB developers add the "machine learning" capability to their programs. So, I'll translate existing C# .NET Core console examples into VB.

Cannot execute tensorflow scoring samples

I have a problem triying to execute tensorflow scoring inception sample. It throws an exception: cannot instantiate transform.ImageLoader

Get cluster items

Sorry for my English :)

I am using sample Clustering Iris Data. But is there anyway except cluster Id also get items which were put in the same cluster?

MovieRecommender

Addd movie recommender from the presentation https://www.youtube.com/watch?v=zXn10vy8F6E&feature=youtu.be&t=1175?

Relative paths to Data and MLModels folders are incorrect for most sample apps

Not sure if it's just "my machine" or because I'm on Windows (10) but the paths to "Data" and "MLModels" appear to be going 3 level up too many.

e.g.
private static string BaseModelsPath = @"../../../../MLModels";
s/b:
private static string BaseModelsPath = @"../MLModels";

Exception in Product-Recommendation-Matrix Factorization sample

Exception in Product Recommendation-Matrix Factorization
This sample needs to be fixed. It is currently throwing the following exception when running:

System.FormatException
HResult=0x80131537
Message=Parsing failed with an exception: Stream reading encountered exception
Source=Microsoft.ML.Data
StackTrace:
at Microsoft.ML.Runtime.Data.TextLoader.Cursor.d__33.MoveNext()
at Microsoft.ML.Runtime.Data.TextLoader.Cursor.MoveNextCore()
at Microsoft.ML.Runtime.Data.RootCursorBase.MoveNext()
at Microsoft.ML.Runtime.Recommender.Internal.SafeTrainingAndModelBuffer.ConstructLabeledNodesFrom(IChannel ch, ICursor cursor, ValueGetter1 labGetter, ValueGetter1 rowGetter, ValueGetter1 colGetter, Int32 rowCount, Int32 colCount) at Microsoft.ML.Runtime.Recommender.Internal.SafeTrainingAndModelBuffer.Train(IChannel ch, Int32 rowCount, Int32 colCount, ICursor cursor, ValueGetter1 labGetter, ValueGetter1 rowGetter, ValueGetter1 colGetter)
at Microsoft.ML.Trainers.MatrixFactorizationTrainer.TrainCore(IChannel ch, RoleMappedData data, RoleMappedData validData)
at Microsoft.ML.Trainers.MatrixFactorizationTrainer.Train(IDataView trainData, IDataView validationData)
at ProductRecommender.Program.Main(String[] args) in D:\GitRepos\machinelearning-samples-master\samples\csharp\getting-started\MatrixFactorization_ProductRecommendation\ProductRecommender\Program.cs:line 60

Inner Exception 1:
FormatException: Stream reading encountered exception

Inner Exception 2:
IOException: Could not open file './Data/Amazon0302.txt'. Error is: Could not find a part of the path 'D:\GitRepos\machinelearning-samples-master\samples\csharp\getting-started\MatrixFactorization_ProductRecommendation\ProductRecommender\bin\Debug\netcoreapp2.1\Data\Amazon0302.txt'.

Inner Exception 3:
DirectoryNotFoundException: Could not find a part of the path 'XXXXXXXX\samples\csharp\getting-started\MatrixFactorization_ProductRecommendation\ProductRecommender\bin\Debug\netcoreapp2.1\Data\Amazon0302.txt'.

It also needs to be updated from version 0.8.0-preview-27128-6 to version 0.8