Increasing a ML model accuracy is not enough, we must also increase its trustworthiness. This is an important step for building resilient AI systems for safety-critical applications such as automotive, finance, and healthcare. For that purpose, we propose a multi-agent system that combines both machine and human agents. In this system, a checker agent calculates a trust score of each instance (which penalizes overconfidence and overcautiousness in predictions) using an agreement-based method and ranks it; then an improver agent filters the anomalous instances based on a human rule-based procedure (which is considered safe), gets the human labels, applies geometric data augmentation, and retrains with the augmented data using transfer learning. We evaluate the system on corrupted versions of the MNIST and FashionMNIST datasets. We get an improvement in accuracy and trust score with just few additional labels compared to a baseline approach.
Machine Learning (ML) is more than just training models, the whole workflow must be considered. Once deployed, a ML model needs to be watched and constantly supervised and debugged to guarantee its validity and robustness in unexpected situations. Debugging in ML aims to identify (and address) the model weaknesses in not trivial contexts. Several techniques have been proposed to identify different types of model weaknesses, such as bias in classification, model decay, adversarial attacks, etc., yet there is not a generic framework that allows them to work in a collaborative, modular, portable, iterative way and, more importantly, flexible enough to allow both human- and machine-driven techniques. In this paper, we propose a novel containerized directed graph framework to support and accelerate end-to-end ML workflow management, supervision, and debugging. The framework allows defining and deploying ML workflows in containers, tracking their metadata, checking their behavior in production, and improving the models by using both learned and human-provided knowledge. We demonstrate these capabilities by integrating in the framework two hybrid systems to detect data drift distribution which identify the samples that are far from the latent space of the original distribution, ask for human intervention, and whether retrain the model or wrap it with a filter to remove the noise of corrupted data at inference time. We test these systems on MNIST-C, CIFAR-10-C, and FashionMNIST-C datasets, obtaining promising accuracy results with the help of human involvement.