🤖 AI Summary
Addressing the challenge of unsupervised long-time-series shapelet learning, this paper proposes NN-STNE—a neural network for end-to-end modeling of shapelet membership probabilities and low-dimensional representation of tactile time-series data. Methodologically, it integrates t-SNE into a hidden layer to mitigate crowding in low-dimensional embeddings; employs L1 regularization for adaptive shapelet length optimization; introduces a Gaussian-kernel-weighted MSE loss to preserve local structure; and leverages K-means initialization to accelerate convergence. The framework constructs a Bag-of-Words tactile representation grounded in shapelet membership. Evaluated on the UCR time-series benchmark and robotic tasks involving electrical component manipulation (e.g., switch actuation), NN-STNE achieves significantly higher clustering accuracy than state-of-the-art unsupervised feature learning methods.
📝 Abstract
This paper introduces NN-STNE, a neural network using t-distributed stochastic neighbor embedding (t-SNE) as a hidden layer to reduce input dimensions by mapping long time-series data into shapelet membership probabilities. A Gaussian kernel-based mean square error preserves local data structure, while K-means initializes shapelet candidates due to the non-convex optimization challenge. Unlike existing methods, our approach uses t-SNE to address crowding in low-dimensional space and applies L1-norm regularization to optimize shapelet length. Evaluations on the UCR dataset and an electrical component manipulation task, like switching on, demonstrate improved clustering accuracy over state-of-the-art feature-learning methods in robotics.