Engineering Papers⌕ Search

Engineering topics

Du, Xiaocong

Publications and source records attributed to Du, Xiaocong.

Evolutionary NAS in Light of Model Stability for Accurate Continual Learning

Continual learning, the capability to learn new knowledge from streaming data without forgetting the previous knowledge, is a critical requirement for dynamic learning systems, especially for emerging edge devices such as self-driving cars and drones. However, continual learning is still facing the catastrophic forgetting problem. Previous work illustrate that model performance on continual learning is not only related to the learning algorithms but also strongly dependent on the inherited model, i.e., the model where continual learning starts. The better stability of the inherited model, the less catastrophic forgetting and thus, the inherited model should be elaborately selected. Inspired by this finding, we develop an evolutionary neural architecture search (ENAS) algorithm that emphasizes the Stability of the inherited model, namely ENAS-S. ENAS-S aims to find optimal architectures for accurate continual learning on edge devices. On CIFAR-10 and CIFAR-100, we present that ENAS-S achieves competitive architectures with lower catastrophic forgetting and smaller model size when learning from a data stream, as compared with handcrafted DNNs.

Du, Xiaocong↗

Noise-Based Selection of Robust Inherited Model for Accurate Continual Learning

There is a growing demand for an intelligent system to continually learn knowledge from a data stream. Continual learning requires both the preservation of previous knowledge (i.e., avoiding catastrophic forgetting) and the acquisition of new knowledge. Different from previous works that focus only on model adaptation (e.g., regularization, network expansion, memory rehearsal, etc.), we propose a novel training scheme named acquisitive learning (AL), which emphasizes both the knowledge inheritance and knowledge acquisition. AL starts from an elaborately selected model with pre-trained knowledge (the inherited model) and then adapts it to new data using segmented training. The selection is achieved by injecting random noise to various inherited models for better model robustness, which promises higher accuracy in further knowledge acquisition. The approach is validated by the visualization of the loss landscape and quantitative roughness measurement. The combination of the selective inherited model and knowledge acquisition reduces catastrophic forgetting by 10X on the CIFAR-100 dataset.

Du, Xiaocong↗

Single-Net Continual Learning with Progressive Segmented Training

There is an increasing need of continual learning in dynamic systems, such as the self-driving vehicle, the surveillance drone, and the robotic system. Such a system requires learning from the data stream, training the model to preserve previous information and adapt to a new task, and generating a single-headed vector for future inference. Different from previous approaches with dynamic structures, this work focuses on a single network and model segmentation to prevent catastrophic forgetting. Leveraging the redundant capacity of a single network, model parameters for each task are separated into two groups: one important group which is frozen to preserve current knowledge, and secondary group to be saved (not pruned) for a future learning. A fixed-size memory containing a small amount of previously seen data is further adopted to assist the training. Without additional regularization, the simple yet effective approach of Progressive Segmented Training (PST) successfully incorporates multiple tasks and achieves the state-of-the-art accuracy in the single-head evaluation on CIFAR-10 and CIFAR-100 datasets. Moreover, the segmented training significantly improves computation efficiency in continual learning at the edge.

Du, Xiaocong↗