A basic problem in machine learning is to find a mapping $f$ from a low dimensional latent space to a high dimensional observation space. Equipped with the representation power of non-linearity, a learner can easily find a mapping which perfectly fits all the observations. However such a mapping is often not considered as good as it is not simple enough and over-fits. How to define simplicity? This paper tries to make such a formal definition of the amount of information imposed by a non-linear mapping. This definition is based on information geometry and is independent of observations, nor specific parametrizations. We prove these basic properties and discuss relationships with parametric and non-parametric embeddings.
Motion planning is challenging when it comes to the case of imperfect state information. Decision should be made based on belief state which evolves according to the noise from the system dynamics and sensor measurement. In this paper, we propose the QV-Tree Search algorithm which combines the state-of-art offline and online approximation methods for POMDP. Instead of full node expansions in the tree search, only probable future observations are considered through forward sampling. This modification helps reduce online computation time and allows for GPU acceleration. We show using repre- sentative examples that the proposed QV-Tree Search is able to actively localize the robot in order to reach the goal location with high probability. The results of the proposed method is also compared with the A* and MDP algorithms, neither of which handles state uncertainty directly. The comparison shows that QV-Tree Search is able to drive the robot to the goal with higher success rate and fewer steps.
In this paper we describe the Open Vision Computer (OVC) which was designed to support high speed, vision guided autonomous drone flight. In particular our aim was to develop a system that would be suitable for relatively small-scale flying platforms where size, weight, power consumption and computational performance were all important considerations. This manuscript describes the primary features of our OVC system and explains how they are used to support fully autonomous indoor and outdoor exploration and navigation operations on our Falcon 250 quadrotor platform.
In this paper, we are interested in building lightweight and efficient convolutional neural networks. Inspired by the success of two design patterns, composition of structured sparse kernels, e.g., interleaved group convolutions (IGC), and composition of low-rank kernels, e.g., bottle-neck modules, we study the combination of such two design patterns, using the composition of structured sparse low-rank kernels, to form a convolutional kernel. Rather than introducing a complementary condition over channels, we introduce a loose complementary condition, which is formulated by imposing the complementary condition over super-channels, to guide the design for generating a dense convolutional kernel. The resulting network is called IGCV3. We empirically demonstrate that the combination of low-rank and sparse kernels boosts the performance and the superiority of our proposed approach to the state-of-the-arts, IGCV2 and MobileNetV2 over image classification on CIFAR and ImageNet and object detection on COCO.
Lexical analysis is believed to be a crucial step towards natural language understanding and has been widely studied. Recent years, end-to-end lexical analysis models with recurrent neural networks have gained increasing attention. In this report, we introduce a deep Bi-GRU-CRF network that jointly models word segmentation, part-of-speech tagging and named entity recognition tasks. We trained the model using several massive corpus pre-tagged by our best Chinese lexical analysis tool, together with a small, yet high-quality human annotated corpus. We conducted balanced sampling between different corpora to guarantee the influence of human annotations, and fine-tune the CRF decoding layer regularly during the training progress. As evaluated by linguistic experts, the model achieved a 95.5% accuracy on the test set, roughly 13% relative error reduction over our (previously) best Chinese lexical analysis tool. The model is computationally efficient, achieving the speed of 2.3K characters per second with one thread.
The total variation distance is a core statistical distance between probability measures that satisfies the metric axioms, with value always falling in $[0,1]$. This distance plays a fundamental role in machine learning and signal processing: It is a member of the broader class of $f$-divergences, and it is related to the probability of error in Bayesian hypothesis testing. Since the total variation distance does not admit closed-form expressions for statistical mixtures (like Gaussian mixture models), one often has to rely in practice on costly numerical integrations or on fast Monte Carlo approximations that however do not guarantee deterministic lower and upper bounds. In this work, we consider two methods for bounding the total variation of univariate mixture models: The first method is based on the information monotonicity property of the total variation to design guaranteed nested deterministic lower bounds. The second method relies on computing the geometric lower and upper envelopes of weighted mixture components to derive deterministic bounds based on density ratio. We demonstrate the tightness of our bounds in a series of experiments on Gaussian, Gamma and Rayleigh mixture models.
We investigate the problem of $L_p$-norm constrained coding, i.e. converting signal into code that lies inside an $L_p$-ball and most faithfully reconstructs the signal. While previous works known as sparse coding have addressed the cases of $L_0$ and $L_1$ norms, more general cases with other $p$ values, especially with unknown $p$, remain a difficulty. We propose the Frank-Wolfe Network (F-W Net), whose architecture is inspired by unrolling and truncating the Frank-Wolfe algorithm for solving an $L_p$-norm constrained problem. We show that the Frank-Wolfe solver for the $L_p$-norm constraint leads to a novel closed-form nonlinear unit, which is parameterized by $p$ and termed $pool_p$. The $pool_p$ unit links the conventional pooling, activation, and normalization operations, making F-W Net distinct from existing deep networks either heuristically designed or converted from projected gradient descent algorithms. We further show that the hyper-parameter $p$ can be made learnable instead of pre-chosen in F-W Net, which gracefully solves the $L_p$-norm constrained coding problem with unknown $p$. We evaluate the performance of F-W Net on an extensive range of simulations as well as the task of handwritten digit recognition, where F-W Net exhibits strong learning capability. We then propose a convolutional version of F-W Net, and apply the convolutional F-W Net into image denoising and super-resolution tasks, where F-W Net all demonstrates impressive effectiveness, flexibility, and robustness.
High speed navigation through unknown environments is a challenging problem in robotics. It requires fast computation and tight integration of all the subsystems on the robot such that the latency in the perception-action loop is as small as possible. Aerial robots add a limitation of payload capacity, which restricts the amount of computation that can be carried onboard. This requires efficient algorithms for each component in the navigation system. In this paper, we describe our quadrotor system which is able to smoothly navigate through mixed indoor and outdoor environments and is able to fly at speeds of more than 18 m/s. We provide an overview of our system and details about the specific component technologies that enable the high speed navigation capability of our platform. We demonstrate the robustness of our system through high speed autonomous flights and navigation through a variety of obstacle rich environments.
We propose a new generic type of stochastic neurons, called $q$-neurons, that considers activation functions based on Jackson's $q$-derivatives with stochastic parameters $q$. Our generalization of neural network architectures with $q$-neurons is shown to be both scalable and very easy to implement. We demonstrate experimentally consistently improved performances over state-of-the-art standard activation functions, both on training and testing loss functions.
Constructing an occupancy representation of the environment is a fundamental problem for robot autonomy. Many accurate and efficient methods exist that address this problem but most assume that the occupancy states of different elements in the map representation are statistically independent. The focus of this paper is to provide a model that captures correlation of the occupancy of map elements. Correlation is important not only for improved accuracy but also for quantifying uncertainty in the map and for planning autonomous mapping trajectories based on the correlation among known and unknown areas. Recent work proposes Gaussian Process (GP) regression to capture covariance information and enable resolution-free occupancy estimation. The drawback of techniques based on GP regression (or classification) is that the computation complexity scales cubically with the length of the measurement history. Our main contribution is a new approach for occupancy mapping that models the binary nature of occupancy measurements precisely, via a Bernoulli distribution, and provides an efficient approximation of GP classification with complexity that does not scale with time. We prove that the error between the estimates provided by our method and those provided by GP classification is negligible. The proposed method is evaluated using both simulated data and real data collected using a Velodyne Puck 3-D range sensor.