Our current research in recommender systems focuses on several open problems with deep theoretical foundations and concrete practical applications. We're investigating:

• Utilization of deep neural networks to reduce the cold-start problem in recommender systems
• Construction of transformers for shopping basket prediction
• Reinforcement learning for optimization of long-term metrics such as user satisfaction
• Transfer learning methods for integrating new recommendation databases
• Application of AutoML for optimizing the architecture and hyperparameters of recommender systems

This research explores automated methods for designing optimal neural network architectures that balance accuracy and computational efficiency. The focus is on developing novel neural architecture search (NAS) algorithms that can discover compact yet powerful models suitable for deployment in resource-constrained environments.

• Designing differentiable architecture search methods for specific domains
• Developing multi-objective optimization approaches that balance accuracy, latency, energy consumption, and memory usage
• Creating adaptive neural networks that can dynamically adjust their architecture during inference
• Integration of domain knowledge into the architecture search process
• Applications in edge computing and embedded systems