Getatlas 2z0gl7v1xe

MOSTLY AI

Stay updated on the status of all your company's support tickets.

Customer portal

The success of a machine learning project heavily depends on selecting the appropriate model for your specific task. Mostly AI provides a range of powerful models, but choosing the right one requires careful consideration of your data characteristics and your desired outcome.

 Determine the primary objective of your machine learning project.  Are you trying to:

  Examine the nature and characteristics of your data:

  Used to predict a continuous output variable:

  Used to classify data into discrete categories:

  Used to group similar data points together:

  More complex models, often used for tasks involving high-dimensional data or non-linear relationships:

  More complex models might achieve higher accuracy but can be prone to overfitting, meaning they perform well on training data but poorly on unseen data.

  Some models, like decision trees, are more interpretable, making it easier to understand how they make predictions.

  More complex models can require significantly more computational resources.

By systematically assessing your problem, data characteristics, and the available models in Mostly AI, you can make an informed decision that sets the stage for building a successful machine learning model.

A comprehensive guide to selecting the appropriate machine learning model based on your specific use case and data.

Choosing the Right Model for Your Problem

Once you've chosen the right model for your machine learning task, it's time to train it on your data and evaluate its performance to ensure it meets your requirements. Mostly AI provides a streamlined process for training and evaluating models, making it easy to assess their effectiveness. 

   Within your project's workspace, go to the "Models" section. 

  Choose the model you've chosen based on your problem and data characteristics (e.g., Linear Regression, Logistic Regression, Decision Tree).

  Set any relevant training parameters for the chosen model:

   Initiate the training process.  Mostly AI will display progress, including training metrics like loss and accuracy.

 Observe training metrics during training to understand how the model is performing.  Look for any signs of overfitting or underfitting. 

  Before training, your data is typically split into three sets:

  Evaluate your model's performance on the test set using appropriate metrics, which depend on your task:

  If your model's performance isn't satisfactory, you can iterate on the process by:

By understanding the training and evaluation process, you can effectively build models that perform well on your data and achieve your machine learning objectives within Mostly AI.

Learn how to train your chosen model on your data and evaluate its performance using various metrics.

Training and Evaluating Models

Hyperparameters are settings that control the learning process of a machine learning model. Unlike model parameters, which are learned from data during training, hyperparameters are set before training and influence how the model learns.  Tuning these hyperparameters carefully can significantly impact model performance and prevent overfitting.

 Controls how rapidly the model updates its weights during training. A high learning rate can lead to instability, while a low learning rate can result in slow training.

  The number of times the model iterates through the entire training dataset. More epochs allow for more learning but can lead to overfitting.

  The number of training examples used in each update during training. Smaller batch sizes can lead to more frequent updates but might be computationally expensive.

  Controls the complexity of the model, preventing it from overfitting. Higher regularization penalties reduce the model's ability to fit the training data too closely.

  In ensemble models like Random Forests, this determines the number of decision trees used in the ensemble.

  Controls the maximum depth of individual trees in Decision Tree models.  Deeper trees can capture complex relationships but also risk overfitting.

  Evaluates model performance over a predefined grid of hyperparameter values. It systematically tries all combinations of values.

 Randomly samples hyperparameter values from a specified distribution. Can be more efficient than grid search for high-dimensional hyperparameter spaces.

Uses past evaluations to guide the search for optimal hyperparameters, often more efficient than grid or random search for complex models.

  Stops training early if the model's performance on a validation set starts to degrade, preventing overfitting.

  Splits the training data into multiple folds and trains and evaluates the model on different combinations of folds to get a more robust evaluation.

Performing Hyperparameter Tuning in Mostly AI

  Go to the "Models" section of your project.

  Select the hyperparameter tuning option for your chosen model.

  Choose a tuning technique (Grid Search, Random Search, Bayesian Optimization).

  Specify the ranges of values to explore for each hyperparameter.

  Initiate the tuning process. The platform will evaluate the model's performance across different hyperparameter combinations.

  Review the tuning results and identify the best performing hyperparameters.

  Re-train your model using the optimal hyperparameters identified through tuning.

By understanding and utilizing various hyperparameter tuning techniques, you can build more robust and effective machine learning models within Mostly AI.

Techniques for optimizing the performance of your model by adjusting hyperparameters and finding the best configuration.