Compare MLlib and Weka 
- Users report that MLlib excels in scalability, particularly for large datasets, making it a preferred choice for mid-market companies looking to leverage big data analytics.
- Reviewers mention that Weka offers a more user-friendly interface, which is beneficial for beginners in machine learning, allowing them to easily navigate through its various features.
- G2 users highlight that MLlib's integration with Apache Spark provides powerful distributed computing capabilities, which is a significant advantage for organizations already using Spark for data processing.
- Users on G2 indicate that Weka's extensive collection of built-in algorithms and visualization tools makes it easier for users to experiment with different models and understand their performance.
- Reviewers say that while MLlib has a steeper learning curve, its performance in handling complex machine learning tasks is often praised, especially in production environments.
- Users report that Weka's strong community support and comprehensive documentation help users troubleshoot issues more effectively, enhancing the overall user experience.
MLlib now works on the new DataFrame API and thus is very easy to use.
It is rigid with some of the algorithms, specially with advanced one like neural network. For instance, you are unable to change activation functions of a neural network. You can either use Sigmoid for all the layers, or tanh which is not really making...
Using Naive Bayes theorems and other clustering algorithms it was very easy to get a visualization and classification
- Graphics are not of the best quality out of the box - Does not have implemented some alternative algorithms
MLlib now works on the new DataFrame API and thus is very easy to use.
Using Naive Bayes theorems and other clustering algorithms it was very easy to get a visualization and classification
It is rigid with some of the algorithms, specially with advanced one like neural network. For instance, you are unable to change activation functions of a neural network. You can either use Sigmoid for all the layers, or tanh which is not really making...
- Graphics are not of the best quality out of the box - Does not have implemented some alternative algorithms
