Learning invariant features using inertial priors

We address the technical challenges involved in combining key features from several theories of the visual cortex in a single coherent model. The resulting model is a hierarchical Bayesian network factored into modular component networks embedding variable-order Markov models. Each component network has an associated receptive field corresponding to components residing in the level directly below it in the hierarchy. The variable-order Markov models account for features that are invariant to naturally occurring transformations in their inputs. These invariant features give rise to increasingly stable, persistent representations as we ascend the hierarchy. The receptive fields of proximate components on the same level overlap to restore selectivity that might otherwise be lost to invariance.      

Learning overcomplete representations

In an overcomplete basis, the number of basis vectors is greater than the dimensionality of the input, and the representation of an input is not a unique combination of basis vectors. Overcomplete representations have been advocated because they have greater robustness in the presence of noise, can be sparser, and can have greater flexibility in matching structure in the data. Overcomplete codes have also been proposed as a model of some of the response properties of neurons in primary visual cortex. Previous work has focused on finding the best representation of a signal using a fixed overcomplete basis (or dictionary). We present an algorithm for learning an overcomplete basis by viewing it as probabilistic model of the observed data. We show that overcomplete bases can yield a better approximation of the underlying statistical distribution of the data and can thus lead to greater coding efficiency. This can be viewed as a generalization of the technique of independent component analysis and provides a method for Bayesian reconstruction of signals in the presence of noise and for blind source separation when there are more sources than mixtures.     

Learning representations from dendrograms

Machine Learning - We propose unsupervised representation learning and feature extraction from dendrograms. The commonly used Minimax distance measures correspond to building a dendrogram with...     

Learning control in robotic systems

The concept of learning and training machines, and some early methodologies were introduced about two decades ago. Robotic systems, undoubtedly, can be developed to a more advanced and intelligent stage. The realization of the learning capability, analogous to the human learning and thinking process, is a desired primary function. A betterment process has been investigated in the literature for applications of learning control to robotic systems. The existing schemes are a type of one-step learning process. A neighboring (2m+1)-step learning control scheme for robotic systems is presented in this paper. For each process, a betterment algorithm which chooses a generalized momentum as an output function, is executed. Also, it is associated with a conceptual learning process by adding a self-teaching knowledge base for speeding up the convergence, so that the learning capability of the resulting robotic systems can be enhanced.     

预测状态表示综述*

预测状态表示是描述离散时间有限状态的动态系统的新方法。使用动作—观测值序列的预测向量表示系统状态在将来时刻发生的概率,能解决现有动态系统决策过程中计算复杂的问题。综述了预测状态表示的基本原理,介绍了预测状态表示的建模过程和规划算法,对已有的建模方法和规划方法进行总结分析和比较,指出了该研究领域的发展方向,最后提出了研究面临的挑战。     

Learning social representations with deep autoencoder for recommender system

World Wide Web - With the development of online social media, it attracts increasingly attentions to utilize social information for recommender systems. Based on the intuition that users are...     

Continuous state representations for AR Systems

In this paper we consider in a behavioral setting the subclass of dis-crete-time, linear, finite-dimensional systems, which can be represented by autoregressive (AR) equations. It will be shown that, with respect to the convergence of all coefficients in an AR representation, there exist continuously dependent input-state-output (i/s/o) representations, under the condition that some specified degree remains constant. This continuous i/s/o representation is given by the Fuhrmann realization.     

Learning spatial concepts from RatSLAM representations

RatSLAM is a biologically-inspired visual SLAM and navigation system that has been shown to be effective indoors and outdoors on real robots. The spatial representation at the core of RatSLAM, the experience map, forms in a distributed fashion as the robot learns the environment. The activity in RatSLAM’s experience map possesses some geometric properties, but still does not represent the world in a human readable form. A new system, dubbed RatChat, has been introduced to enable meaningful communication with the robot. The intention is to use the “language games” paradigm to build spatial concepts that can be used as the basis for communication. This paper describes the first step in the language game experiments, showing the potential for meaningful categorization of the spatial representations in RatSLAM.     

Learning context-free grammars using tabular representations

We present a novel algorithm using new hypothesis representations for learning context-free grammars from a finite set of positive and negative examples. We propose an efficient hypothesis representation method which consists of a table-like data structure similar to the parse table used in efficient parsing algorithms for context-free grammars such as Cocke-Younger-Kasami algorithm. By employing this representation method, the problem of learning context-free grammars from examples can be reduced to the problem of partitioning the set of nonterminals. We use genetic algorithms for solving this partitioning problem. Further, we incorporate partially structured examples to improve the efficiency of our learning algorithm, where a structured example is represented by a string with some parentheses inserted to indicate the shape of the derivation tree of the unknown grammar. We demonstrate some experimental results using these algorithms and theoretically analyse the completeness of the search space using the tabular method for context-free grammars.     

On state representations of time-varying nonlinear systems

This work considers a nonlinear time-varying system described by a state representation, with input u and state x. A given set of functions v, which is not necessarily the original input u of the system, is the (new) input candidate. The main result provides necessary and sufficient conditions for the existence of a local classical state space representation with input v. These conditions rely on integrability tests that are based on a derived flag. As a byproduct, one obtains a sufficient condition of differential flatness of nonlinear systems.     

Canonical state space representations for multivariable systems

The problem of state space representations for multiple-input, multiple-output linear time-invariant systems is considered. The known definitions of the first and second canonical forma for scalar systems are used to facilitate the derivation of explicit canonical expressions for multivariable systems.     

On state representations of nonlinear implicit systems

The parameter identification problem in systems governed by partial differential equations is investigated. Stochastic approximation algorithms are applied for identifying a class of distributed systems driven by random inputs and observed through noisy measurements. No restrictions about the probability distributions are imposed. These algorithms converge with probability one, and are suitable for on-line applications. The proposed identification method assumes that a previous system classification has been performed, such that the model to be identified is known up to a set of space-varying parameters, where extraneous terms may be included. The very real case of noisy measurements taken at a limited number of discrete points Located in the spatial domain is considered.     

Learning discriminative representations for semantical crossmodal retrieval

Heterogeneous gap among different modalities emerges as one of the critical issues in multimedia retrieval areas. Unlike traditional unimodal cases, where raw features are extracted and directly measured, the heterogeneous nature of crossmodal tasks requires the intrinsic semantic representation to be compared in a unified framework. Based on a flexible “feature up-lifting and down projecting” mechanism, this paper studies the learning of crossmodal semantic features that can be retrieved across different modalities. Two effective methods are proposed to mine semantic correlations. One is for traditional handcrafted features, and the other is based on deep neural network. We treat them respectively as normal and deep version of our proposed shared discriminative semantic representation learning (SDSRL) framework. We evaluate both of these two methods on two public multimodal datasets for crossmodal and unimodal retrieval tasks. The experimental results demonstrate that our proposed methods outperform the compared baselines and achieve state-of-the-art performance in most scenarios.     

Learning robustly stable open-loop motions for robotic manipulation

Robotic arms have been shown to be able to perform cyclic tasks with an open-loop stable controller. However, model errors make it hard to predict in simulation what cycle the real arm will perform. This makes it difficult to accurately perform pick and place tasks using an open-loop stable controller. This paper presents an approach to make open-loop controllers follow the desired cycles more accurately. First, we check if the desired cycle is robustly open-loop stable, meaning that it is stable even when the model is not accurate. A novel robustness test using linear matrix inequalities is introduced for this purpose. Second, using repetitive control we learn the open loop controller that tracks the desired cycle. Hardware experiments show that using this method, the accuracy of the task execution is improved to a precision of 2.5 cm, which suffices for many pick and place tasks.     

机器人机械手的离散时间学习控制*

本文为机器人机械手提出了一种基于离散时间的重复学习控制法，这种学习控制法利用机器人动力学模型的部分知识，从它的特性和实用观点看，这种控制法比现有的其它学习控制法更有吸引力，本文还给出了学习控制法的收敛性证明和计算机仿真结果。     

Learning the external and internal priors for multispectral and hyperspectral image fusion

Recently, multispectral image(MSI) and hyperspectral image(HSI) fusion has been a popular topic in high-resolution HSI acquisition. This fusion leads to a challenging underdetermined problem, which image priors are used to regularize, aiming at improving fusion accuracy. To fully exploit HSI priors, this paper proposes two kinds of priors, i.e., external priors and internal priors, to regularize the fusion problem.An external prior represents the general image characteristics and is learned from...     

Learning viewpoint invariant perceptual representations from cluttered images

In order to perform object recognition, it is necessary to form perceptual representations that are sufficiently specific to distinguish between objects, but that are also sufficiently flexible to generalize across changes in location, rotation, and scale. A standard method for learning perceptual representations that are invariant to viewpoint is to form temporal associations across image sequences showing object transformations. However, this method requires that individual stimuli be presented in isolation and is therefore unlikely to succeed in real-world applications where multiple objects can co-occur in the visual input. This paper proposes a simple modification to the learning method that can overcome this limitation and results in more robust learning of invariant representations.     

Learning semantic representations of objects and their parts

Recently, large scale image annotation datasets have been collected with millions of images and thousands of possible annotations. Latent variable models, or embedding methods, that simultaneously learn semantic representations of object labels and image representations can provide tractable solutions on such tasks. In this work, we are interested in jointly learning representations both for the objects in an image, and the parts of those objects, because such deeper semantic representations could bring a leap forward in image retrieval or browsing. Despite the size of these datasets, the amount of annotated data for objects and parts can be costly and may not be available. In this paper, we propose to bypass this cost with a method able to learn to jointly label objects and parts without requiring exhaustively labeled data. We design a model architecture that can be trained under a proxy supervision obtained by combining standard image annotation (from ImageNet) with semantic part-based within-label relations (from WordNet). The model itself is designed to model both object image to object label similarities, and object label to object part label similarities in a single joint system. Experiments conducted on our combined data and a precisely annotated evaluation set demonstrate the usefulness of our approach.