Catastrophic forgetting remains a central challenge in continual learning (CL) with pre-trained models. While existing approaches typically freeze the backbone and fine-tune a small number of parameters to mitigate forgetting, they still rely on iterative error backpropagation and gradient-based optimization, which can be computationally intensive and less suitable for resource-constrained environments. To address this, we propose FoRo, a forward-only, gradient-free continual learning method. FoRo consists of a lightweight prompt tuning strategy and a novel knowledge encoding mechanism, both designed without modifying the pre-trained model. Specifically, prompt embeddings are inserted at the input layer and optimized using the Covariance Matrix Adaptation Evolution Strategy (CMA-ES), which mitigates distribution shifts and extracts high-quality task representations. Subsequently, task-specific knowledge is encoded into a knowledge encoding matrix via nonlinear random projection and recursive least squares, enabling incremental updates to the classifier without revisiting prior data. Experiments show that FoRo significantly reduces average forgetting and improves accuracy. Thanks to forward-only learning, FoRo reduces memory usage and run time while maintaining high knowledge retention across long task sequences. These results suggest that FoRo could serve as a promising direction for exploring continual learning with pre-trained models, especially in real-world multimedia applications where both efficiency and effectiveness are critical.

What started in Indiana at the Crossroads of America in 2019 was well-received by recent AASHTO attendees seeking new and innovative ways to work more efficiently and meet or exceed their goals. For many years, Indiana was the center of the automobile industry. Elwood Haynes built one of the first successful automobiles in 1894 and tested it on the Fourth of July in Kokomo. He wanted to show the world that he could do better. AI and ML News: Why SMBs Shouldn't Be Afraid of Artificial Intelligence (AI) Fast forward 125 years later to 2019 when the Indiana DOT had the same desire as Elwood Haynes.

The Indiana Transportation Department (INDOT) is applying machine learning (ML) to project bundling to maximize cost savings and reduce the time it takes to create bundles from a week to minutes. Transportation engineers typically group infrastructure construction projects into bundles to gain efficiencies in contracting and project management, minimize infrastructure disruption and achieve economies of scale. Bundles are created from similar projects – bridge projects, freeway lighting installations or safety improvements – or from a variety of types of work in a specific location. Bundling is complex, manual, subjective process, with engineers usually working from maps and spreadsheets. INDOT has done manual project bundling for about five years and used that data to test benefits of ML-enabled bundling, in which an algorithm that had been trained on historic data makes bundling suggestions.