Objects play a crucial role in our everyday activities. Though multisensory object-centric learning has shown great potential lately, the modeling of objects in prior work is rather unrealistic. ObjectFolder 1.0 is a recent dataset that introduces 100 virtualized objects with visual, acoustic, and tactile sensory data. However, the dataset is small in scale and the multisensory data is of limited quality, hampering generalization to real-world scenarios. We present ObjectFolder 2.0, a large-scale, multisensory dataset of common household objects in the form of implicit neural representations that significantly enhances ObjectFolder 1.0 in three aspects. First, our dataset is 10 times larger in the amount of objects and orders of magnitude faster in rendering time. Second, we significantly improve the multisensory rendering quality for all three modalities. Third, we show that models learned from virtual objects in our dataset successfully transfer to their real-world counterparts in three challenging tasks: object scale estimation, contact localization, and shape reconstruction. ObjectFolder 2.0 offers a new path and testbed for multisensory learning in computer vision and robotics. The dataset is available at https://github.com/rhgao/ObjectFolder.
Accounting for the effects of confounders is one of the central challenges in causal inference. Unstructured multi-modal data (images, time series, text) contains valuable information about diverse types of confounders, yet it is typically left unused by most existing methods. This paper seeks to develop techniques that leverage this unstructured data within causal inference to correct for additional confounders that may otherwise not be accounted for. We formalize this task and we propose algorithms based on deep structural equations that treat multi-modal unstructured data as proxy variables. We empirically demonstrate on tasks in genomics and healthcare that unstructured data can be used to correct for diverse sources of confounding, potentially enabling the use of large amounts of data that were previously not used in causal inference.
Computation on a large volume of data at high speed and low power requires energy-efficient computing architectures. Spiking neural network (SNN) with bio-inspired spike-timing-dependent plasticity learning (STDP) is a promising solution for energy-efficient neuromorphic systems than conventional artificial neural network (ANN). Previous works on SNN with STDP learning primarily uses memristive devices which are difficult to fabricate. Some reported works on SNN makes use of memristor macro models, which are software-based and cannot give complete insight into circuit implementation challenges. This article presents for the first time, a full circuit-level implementation of the SNN system featuring on-chip unsupervised STDP learning in standard CMOS technology. It does not involve the use of FPGAs, CPUs or GPUs for training the neural network. We demonstrated the complete circuit-level design, implementation and simulation of SNN with on-chip training and inference for pattern classification using 180 nm CMOS technology. A comprehensive comparison of the proposed SNN circuit with the previous related work is also presented. To demonstrate the versatility of the CMOS synapse circuit for application scenarios requiring rate-based learning, we have tuned the pair-based STDP circuit to obtain Bienenstock-Cooper-Munro (BCM) characteristics and applied it to heart rate classification.
We study the problem of robust estimation of the mean vector of a sub-Gaussian distribution. We introduce an estimator based on spectral dimension reduction (SDR) and establish a finite sample upper bound on its error that is minimax-optimal up to a logarithmic factor. Furthermore, we prove that the breakdown point of the SDR estimator is equal to $1/2$, the highest possible value of the breakdown point. In addition, the SDR estimator is equivariant by similarity transforms and has low computational complexity. More precisely, in the case of $n$ vectors of dimension $p$ -- at most $\varepsilon n$ out of which are adversarially corrupted -- the SDR estimator has a squared error of order $\big(\frac{r_\Sigma}{n} + \varepsilon^2\log(1/\varepsilon)\big){\log p}$ and a running time of order $p^3 + n p^2$. Here, $r_\Sigma\le p$ is the effective rank of the covariance matrix of the reference distribution. Another advantage of the SDR estimator is that it does not require knowledge of the contamination rate and does not involve sample splitting. We also investigate extensions of the proposed algorithm and of the obtained results in the case of (partially) unknown covariance matrix.
This report provides a description of the methodology I used in the IEEE-CIS 3rd Technical Challenge. For the forecast, I used a quantile regression forest approach using the solar variables provided by the Bureau of Meterology of Australia (BOM) and many of the weather variables from the European Centre for Medium-Range Weather Forecasting (ECMWF). Groups of buildings and all of the solar instances were trained together as they were observed to be closely correlated over time. Other variables used included Fourier values based on hour of day and day of year, and binary variables for combinations of days of the week. The start dates for the time series were carefully tuned based on phase 1 and cleaning and thresholding was used to reduce the observed error rate for each time series. For the optimization, a four-step approach was used using the forecast developed. First, a mixed-integer program (MIP) was solved for the recurring and recurring plus once-off activities, then each of these was extended using a mixed-integer quadratic program (MIQP). The general strategy was chosen from one of two ("array" from the "array" and "tuples" approaches) while the specific step improvement strategy was chosen from one of five ("no forced discharge").
A robotic swarm may encounter traffic congestion when many robots simultaneously attempt to reach the same area. For solving that efficiently, robots must execute decentralised traffic control algorithms. In this work, we propose a measure for evaluating the access efficiency of a common target area as the number of robots in the swarm rises: the common target area throughput. We demonstrate that the throughput of a target region with a limited area as the time tends to infinity -- the asymptotic throughput -- is finite, opposed to the relation arrival time at target per number of robots that tends to infinity. Using this measure, we can analytically compare the effectiveness of different algorithms. In particular, we propose and formally evaluate three different theoretical strategies for getting to a circular target area: (i) forming parallel queues towards the target area, (ii) forming a hexagonal packing through a corridor going to the target, and (iii) making multiple curved trajectories towards the boundary of the target area. We calculate the throughput for a fixed time and the asymptotic throughput for these strategies. Additionally, we corroborate these results by simulations, showing that when an algorithm has higher throughput, its arrival time per number of robots is lower. Thus, we conclude that using throughput is well suited for comparing congestion algorithms for a common target area in robotic swarms even if we do not have their closed asymptotic equation.
Limited by the time complexity of querying k-hop neighbors in a graph database, most graph algorithms cannot be deployed online and execute millisecond-level inference. This problem dramatically limits the potential of applying graph algorithms in certain areas, such as financial fraud detection. Therefore, we propose Asynchronous Propagation Attention Network, an asynchronous continuous time dynamic graph algorithm for real-time temporal graph embedding. Traditional graph models usually execute two serial operations: first graph computation and then model inference. We decouple model inference and graph computation step so that the heavy graph query operations will not damage the speed of model inference. Extensive experiments demonstrate that the proposed method can achieve competitive performance and 8.7 times inference speed improvement in the meantime.
This paper introduces a distributed leaderless swarm formation control framework to address the problem of collectively driving a swarm of robots to track a time-varying formation. The swarm's formation is captured by the trajectory of an abstract shape that circumscribes the convex hull of robots' positions and is independent of the number of robots and their ordering in the swarm. For each robot in the swarm, given global specifications in terms of the trajectory of the abstract shape parameters, the proposed framework synthesizes a control law that steers the swarm to track the desired formation using the information available at the robot's local neighbors. For this purpose, we generate a suitable local reference trajectory that the robot controller tracks by solving the input-output linearization problem. Here, we select the swarm output to be the parameters of the abstract shape. For this purpose, we design a dynamic average consensus estimator to estimate the abstract shape parameters. The abstract shape parameters are used as the swarm state feedback to generate a suitable robot trajectory. We demonstrate the effectiveness and robustness of the proposed control framework by providing the simulation of coordinated collective navigation of a group of car-like robots in the presence of robots and communication link failures.
Precise representations of 3D faces are beneficial to various computer vision and graphics applications. Due to the data discretization and model linearity, however, it remains challenging to capture accurate identity and expression clues in current studies. This paper presents a novel 3D morphable face model, namely ImFace, to learn a nonlinear and continuous space with implicit neural representations. It builds two explicitly disentangled deformation fields to model complex shapes associated with identities and expressions, respectively, and designs an improved learning strategy to extend embeddings of expressions to allow more diverse changes. We further introduce a Neural Blend-Field to learn sophisticated details by adaptively blending a series of local fields. In addition to ImFace, an effective preprocessing pipeline is proposed to address the issue of watertight input requirement in implicit representations, enabling them to work with common facial surfaces for the first time. Extensive experiments are performed to demonstrate the superiority of ImFace.
Direct multi-task twin support vector machine (DMTSVM) explores the shared information between multiple correlated tasks, then it produces better generalization performance. However, it contains matrix inversion operation when solving the dual problems, so it costs much running time. Moreover, kernel trick cannot be directly utilized in the nonlinear case. To effectively avoid above problems, a novel multi-task nonparallel support vector machine (MTNPSVM) including linear and nonlinear cases is proposed in this paper. By introducing epsilon-insensitive loss instead of square loss in DMTSVM, MTNPSVM effectively avoids matrix inversion operation and takes full advantage of the kernel trick. Theoretical implication of the model is further discussed. To further improve the computational efficiency, the alternating direction method of multipliers (ADMM) is employed when solving the dual problem. The computational complexity and convergence of the algorithm are provided. In addition, the property and sensitivity of the parameter in model are further explored. The experimental results on fifteen benchmark datasets and twelve image datasets demonstrate the validity of MTNPSVM in comparison with the state-of-the-art algorithms. Finally, it is applied to real Chinese Wine dataset, and also verifies its effectiveness.