Tranlations made by googe translator
From today you can enjoy the new version (1.3) of LearningML. In this version, new functionalities have been added aimed at exploring Machine Learning algorithms to understand them better, since until now they were a black box over which the user had no control.
But it has also incorporated the recognition of numerical data, that is, of sets of numbers organized tabularly (as in spreadsheets). This will allow the creation of Machine Learning models from datasets obtained by reading experiments with sensors, or from known datasets that are openly available on the network, such as the iris dataset or the boston house dataset.
You can access this new version from the “Version 1.3 (beta)” button on the main page, which links to the following url: https://beta.learningml.org/editor.
The novelties introduced in this version are the following:
At the code level
- Separation of feature extraction and ML algorithm.
- Update lml-editor to Angular 12.
- Reorganization of the lml-scratch repos to sync with the develop branch of scratch.
- Implementation of a message protocol to request ML services from lml-scratch to lml-editor (lml-message-protocol).
- Simplification of lml-scratch code thanks to lml-message-protocol.
At the level of functionalities:
- Advanced mode
- Selection of different Machine Learning algorithms. For now, the algorithms are offered: KNN and neural network.
- Adjustment of the typical training parameters of each algorithm.
- Definition (optional) of a percentage of training data for the calculation of a confusion matrix with which to evaluate the precision of the constructed model.
- Visualization of a confusion matrix to evaluate the built model.
- Visualization of the learning process for the neural network.
- Recognition of numerical data sets.
- Visualization of algorithm decision limits for two-dimensional numerical models.
Little by little I will be adding new entries and practical activities to explain in detail these new functionalities that have been designed to gain more insight into the operation of Machine Learning algorithms. For now, you can see in the LearningML manual how they are used, although no pedagogical explanation is made in this manual.
It is important that you keep in mind that it is a beta version, that is, it may still have a fault than another. In fact, if you find an error, I appreciate that you let me know by sending a message through the contact form on this website. When this version is sufficiently debugged, it will become the stable version and work will begin on the new beta.
And that’s it! I hope you enjoy the new features in this version.