Think of Provider as InheretedWidget, with less boilerplate. Infact, Provider is built ontop of InheretedWidget. It just

These must-have tools for .NET developers will set you free from monotonous tasks and optimize your performance. Indeed, you don’t have to install all these tools at once, you can adjust them according to your project needs.,These must-have tools for .NET developers will set you free from monotonous tasks and optimize your performance. Indeed, you don’t have to install all these tools at once, you can adjust them according to your project needs.,There’s no need for data cleaning — all data types are numeric, and there are no missing data. The train/test split is the next step. The column quality is the target variable, and it can be either good or bad. To get the same split, please set the value of random_state to 42:,on RiverPod’s website, they call themselves “Provider, but different.” which couldn’t ring more true. Provider, even if it cuts out a lot of boilerplate, still has some that can be reduced further. Also, Provider depends on BuildContext — Which, I think in a lot of cases is really great (it forces you to use the widget tree) but sometimes, like app lifecycle, it’s just not practical to get BuildContext where you need it.,And now we’re ready to train the model. XGBoost classifier will do the job, so make sure to install it first (pip install xgboost). Once again, the value of random_state is set to 42 for reproducibility:,No dependency on BuildContext: This is also a great one, for reasons mentioned already. Sometimes, you just can’t get the BuildContext where you need it. This might be the onAppClose listener, that you set up right in the main function.,SHAP (SHapley Additive exPlanations) is a game-theoretic approach to explain the output of any machine learning model. It connects optimal credit allocation with local explanations using the classic Shapley values from game theory and their related extensions. (source: https://github.com/slundberg/shap),But what are Shapely values? Put simply, they are measures of contributions each predictor (feature) has in a machine learning model. This is the least fancy definition on the web, guaranteed, but I reckon it’s easy enough to understand.,Today you’ll learn how to explain machine learning models to the general population. We’ll use three different plots for interpretation — one for a single prediction, one for a single variable, and one for the entire dataset.,If your app is already using Provider, you’re 100% good to keep using it. It’s a very good package for what it does and there’s no reason to do a full migration to another solution.,To explain the model through SHAP, we first need to install the library. You can do it by executing pip install shap from the Terminal. We can then import it, make an explainer based on the XGBoost model, and finally calculate the SHAP values:,Pipelines are used in Ktor as an extension mechanism to plug functionality in at the right place. For example, a Ktor application defines five main phases: Setup, Monitoring, Features, Call and Fallback. The routing feature defines its own nested pipeline inside the application’s call phase.,Less boilerplate than Provider: RiverPod does a great job trimming the fat of Provider, allowing you to just register a single top level store and then be done, rather than having to provide each provider individually. (Some of you might be wincing at the thought of everything in one place — Don’t worry! You can scope your pods if necessary.),Compile Safety: This is, by far, the best innovation to come to state management. If your code compiles, you’re good. No longer are we going to wonder why we can’t find our provider in the tree. This is an innovation that is huge and will save you so, so much time.,Nobody likes a black-box model. With sophisticated algorithms and a fair amount of data preparation, building good models is easy, but what’s going on inside? That’s where Explainable AI and SHAP come into place.