Online Group-Structured Dictionary Learning Supplementary Material
نویسندگان
چکیده
Lemma 1. If ρ = 0, then Mt = M0 +M′t (∀t ≥ 1). When ρ > 0, then Mt = M ′ t (∀t ≥ 1). Proof. 1. Case ρ = 0: Since γt = 1 (∀t ≥ 1), thus Mt = M0 + ∑t i=1 Ni. We also have that ( i t 0 = 1 (∀i ≥ 1), and therefore M′t = ∑t i=1 Ni, which completes the proof. 2. Case ρ > 0: The proof proceeds by induction. • t = 1: In this case γ1 = 0, M1 = 0 × M0 + N1 = N1 and M′1 = N1, which proves that M1 = M ′ 1. • t > 1: Using the definitions of Mt and M ′ t, and exploiting the fact that Mt−1 = M ′ t−1 by induction, after some calculation we have that:
منابع مشابه
Online Dictionary Learning with Group Structure Inducing Norms
• Sparse coding. • Structured sparsity (e.g., disjunct groups, trees): increased performance in several applications. • Our goal: develop a dictionary learning method, which – enables general overlapping group structures, – is online: fast, memory efficient, adaptive, – applies non-convex sparsity inducing regularization: ∗ fewer measurements, ∗ weaker conditions on the dictionary, ∗ robust (w....
متن کاملSupplementary Material: Analysis-Synthesis Dictionary Learning for Universality-Particularity Representation based Classification
In the supplementary material, we give more analysis and experimental results for the proposed ASDL-UP. In the following sections, we will present the algorithm of dictionary atom pair updating in universal ASDL, control experiments to evaluate different components of ASDL-UP, analysissynthesis dictionary learned from face images, recognition on the dataset of Scene-15 (S. Lazebnik 2007), and t...
متن کاملSupplementary Material: Jointly Learning Non-negative Projection and Dictionary with Discriminative Graph Constraints for Classification
The JNPDL model is well motivated by the current drawbacks of dictionary learning approaches, while each constraints are also well designed (the novel discriminative graph constraints are proposed, and all constrains are designed to be easily optimized). Aiming to bridge the gap between features and dictionary, we do think the proposed idea of learning a projection for features jointly with the...
متن کاملLearning a Dictionary of Shape Epitomes with Applications to Image Labeling: Supplementary Material
In the main paper, we have proposed a new representation for image shape: a dictionary of shape epitomes, which explicitly include hidden variables to encode shift and rotation. We have also applied this representation for semantic image labeling. In this supplementary material, we further show the versatility of shape epitomes by employing them for edge detection and local appearance modeling.
متن کاملPart-based Visual Tracking with Online Latent Structural Learning: Supplementary Material
In this supplementary material, we provide more related work, more details of structured output tracking and online latent structural learning for visual tracking, and more experimental results (both qualitative and quantitative). The experimental results includes demonstration videos, more quantitative results of our tracking algorithm with and without parts in Tab. 2 and Fig. 1, more quantita...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2011