Learning is a repetitive process where information and outputs are classified and stored for future tasks. A key part of learning is to generalize from experience.
Ardi Learning includes Machine Learning, which trains, tests and manages learning models online. With the recognition of certain patterns from historical data, users can then make better decisions. It also includes Autonomous Learning which allows machines to go beyond the training model and self-evolve, it helps to breakthrough the limit of existing data and process.
(ML) Algorithms Supports:
Ardi Learning is designed with different modules. Supported by Graphen’s most advanced transfer learning and hyperparameter search technology, it allows users to easily develop high-quality custom machine learning models without writing training code. Ardi’s Data Labeling Service also makes it possible to create high-accuracy tags provided by manual taggers in order to obtain a better machine learning model.
Model Training – Provide convenient functions. Users can choose machine learning model and algorithm and tune parameters flexibly.
Model Deployment – Support user to set the frequency of model execution, deployment and execution.
Model Optimization – Support users to optimize model with flexibility.
Model Evaluation – Support general evaluation criteria to regression and classification including accuracy and recall.
Model Management – Integrated support of importing various features of data, model type, saved model, access and deployment. Support imports of models trained on external platforms. Support automatically generated versions of models.
Example: Autonomous Learning through Imperfect Training Labels
Imperfect learning: Machine Learning theories and algorithms for supervised concept learning developed from imperfect annotations.
Developed methodologies to obtain imperfect annotation – learning from cross-modality information or web links.
Developed algorithms and systems to generate concept models – novel generalized Multiple-Instance Learning algorithm with Uncertain Labeling Density.
Jose is a toy factory director. With the expansion of his business, he recently discovered that the manual quality inspection efficiency is relatively low, while the cost is too high. He wants to find a more cost-effective way of quality inspection.
Ardi platform helps him solve this problem. Through the random forest classification model trained by uploading pipeline data on Ardi, the factory can find defective products faster and more accurately.