The most common use of data visualization is to minimize the complexity for proper understanding. A graph is one of the most commonly used representations for understanding relational data. It produces a simplified representation of data that is challenging to comprehend if kept in a textual format. In this study, we propose a methodology to utilize the relational properties of source code in the form of a graph to identify Just-in-Time (JIT) bug prediction in software systems during different revisions of software evolution and maintenance. We presented a method to convert the source codes of commit patches to equivalent graph representations and named it Source Code Graph (SCG). To understand and compare multiple source code graphs, we extracted several structural properties of these graphs, such as the density, number of cycles, nodes, edges, etc. We then utilized the attribute values of those SCGs to visualize and detect buggy software commits. We process more than 246K software commits from 12 subject systems in this investigation. Our investigation on these 12 open-source software projects written in C++ and Java programming languages shows that if we combine the features from SCG with conventional features used in similar studies, we will get the increased performance of Machine Learning (ML) based buggy commit detection models. We also find the increase of F1~Scores in predicting buggy and non-buggy commits statistically significant using the Wilcoxon Signed Rank Test. Since SCG-based feature values represent the style or structural properties of source code updates or changes in the software system, it suggests the importance of careful maintenance of source code style or structure for keeping a software system bug-free.
Prior literature on adversarial attack methods has mainly focused on attacking with and defending against a single threat model, e.g., perturbations bounded in Lp ball. However, multiple threat models can be combined into composite perturbations. One such approach, composite adversarial attack (CAA), not only expands the perturbable space of the image, but also may be overlooked by current modes of robustness evaluation. This paper demonstrates how CAA's attack order affects the resulting image, and provides real-time inferences of different models, which will facilitate users' configuration of the parameters of the attack level and their rapid evaluation of model prediction. A leaderboard to benchmark adversarial robustness against CAA is also introduced.
Unconstrained lip-to-speech synthesis aims to generate corresponding speeches from silent videos of talking faces with no restriction on head poses or vocabulary. Current works mainly use sequence-to-sequence models to solve this problem, either in an autoregressive architecture or a flow-based non-autoregressive architecture. However, these models suffer from several drawbacks: 1) Instead of directly generating audios, they use a two-stage pipeline that first generates mel-spectrograms and then reconstructs audios from the spectrograms. This causes cumbersome deployment and degradation of speech quality due to error propagation; 2) The audio reconstruction algorithm used by these models limits the inference speed and audio quality, while neural vocoders are not available for these models since their output spectrograms are not accurate enough; 3) The autoregressive model suffers from high inference latency, while the flow-based model has high memory occupancy: neither of them is efficient enough in both time and memory usage. To tackle these problems, we propose FastLTS, a non-autoregressive end-to-end model which can directly synthesize high-quality speech audios from unconstrained talking videos with low latency, and has a relatively small model size. Besides, different from the widely used 3D-CNN visual frontend for lip movement encoding, we for the first time propose a transformer-based visual frontend for this task. Experiments show that our model achieves $19.76\times$ speedup for audio waveform generation compared with the current autoregressive model on input sequences of 3 seconds, and obtains superior audio quality.
Along with advances in optical sensors is the common practice of building an imaging system with heterogeneous cameras. While high-resolution (HR) videos acquisition and analysis are benefited from hybrid sensors, the intrinsic characteristics of multiple cameras lead to an interesting motion transfer problem. Unfortunately, most of the existing methods provide no theoretical analysis and require intensive training data. In this paper, we propose an algorithm using time series analysis for motion transfer among multiple cameras. Specifically, we firstly identify seasonality in motion data and then build an addictive time series model to extract patterns that could be transferred across cameras. Our approach has a complete and clear mathematical formulation, thus being efficient and interpretable. Through quantitative evaluations on real-world data, we demonstrate the effectiveness of our method. Furthermore, our motion transfer algorithm could combine with and facilitate downstream tasks, e.g., enhancing pose estimation on LR videos with inherent patterns extracted from HR ones. Code is available at https://github.com/IndigoPurple/TSAMT.
Any human activity can be represented as a temporal sequence of actions performed to achieve a certain goal. Unlike machine-made time series, these action sequences are highly disparate as the time taken to finish a similar action might vary between different persons. Therefore, understanding the dynamics of these sequences is essential for many downstream tasks such as activity length prediction, goal prediction, etc. Existing neural approaches that model an activity sequence are either limited to visual data or are task specific, i.e., limited to next action or goal prediction. In this paper, we present ProActive, a neural marked temporal point process (MTPP) framework for modeling the continuous-time distribution of actions in an activity sequence while simultaneously addressing three high-impact problems -- next action prediction, sequence-goal prediction, and end-to-end sequence generation. Specifically, we utilize a self-attention module with temporal normalizing flows to model the influence and the inter-arrival times between actions in a sequence. Moreover, for time-sensitive prediction, we perform an early detection of sequence goal via a constrained margin-based optimization procedure. This in-turn allows ProActive to predict the sequence goal using a limited number of actions. Extensive experiments on sequences derived from three activity recognition datasets show the significant accuracy boost of ProActive over the state-of-the-art in terms of action and goal prediction, and the first-ever application of end-to-end action sequence generation.
Two outstanding challenges for machine learning (ML) accelerated chemical discovery are the synthesizability of candidate molecules or materials and the fidelity of the data used in ML model training. To address the first challenge, we construct a hypothetical design space of 32.5M transition metal complexes (TMCs), in which all of the constituent fragments (i.e., metals and ligands) and ligand symmetries are synthetically accessible. To address the second challenge, we search for consensus in predictions among 23 density functional approximations across multiple rungs of Jacob's ladder. To accelerate the screening of these 32.5M TMCs, we use efficient global optimization to sample candidate low-spin chromophores that simultaneously have low absorption energies and low static correlation. Despite the scarcity (i.e., $<$ 0.01\%) of potential chromophores in this large chemical space, we identify transition metal chromophores with high likelihood (i.e., $>$ 10\%) as the ML models improve during active learning. This represents a 1,000 fold acceleration in discovery corresponding to discoveries in days instead of years. Analyses of candidate chromophores reveal a preference for Co(III) and large, strong-field ligands with more bond saturation. We compute the absorption spectra of promising chromophores on the Pareto front by time-dependent density functional theory calculations and verify that two thirds of them have desired excited state properties. Although these complexes have never been experimentally explored, their constituent ligands demonstrated interesting optical properties in literature, exemplifying the effectiveness of our construction of realistic TMC design space and active learning approach.
We present a framework to generate periodic trajectory references for a 3D under-actuated bipedal robot, using a linear inverted pendulum (LIP) based controller with adaptive neural regulation. We use the LIP template model to estimate the robot's center of mass (CoM) position and velocity at the end of the current step, and formulate a discrete controller that determines the next footstep location to achieve a desired walking profile. This controller is equipped on the frontal plane with a Neural-Network-based adaptive term that reduces the model mismatch between the template and physical robot that particularly affects the lateral motion. Then, the foot placement location computed for the LIP model is used to generate task space trajectories (CoM and swing foot trajectories) for the actual robot to realize stable walking. We use a fast, real-time QP-based inverse kinematics algorithm that produces joint references from the task space trajectories, which makes the formulation independent of the knowledge of the robot dynamics. Finally, we implemented and evaluated the proposed approach in simulation and hardware experiments with a Digit robot obtaining stable periodic locomotion for both cases.
In many sequential tasks, a model needs to remember relevant events from the distant past to make correct predictions. Unfortunately, a straightforward application of gradient based training requires intermediate computations to be stored for every element of a sequence. This requires prohibitively large computing memory if a sequence consists of thousands or even millions elements, and as a result, makes learning of very long-term dependencies infeasible. However, the majority of sequence elements can usually be predicted by taking into account only temporally local information. On the other hand, predictions affected by long-term dependencies are sparse and characterized by high uncertainty given only local information. We propose MemUP, a new training method that allows to learn long-term dependencies without backpropagating gradients through the whole sequence at a time. This method can be potentially applied to any gradient based sequence learning. MemUP implementation for recurrent architectures shows performances better or comparable to baselines while requiring significantly less computing memory.
Comparing different AutoML frameworks is notoriously challenging and often done incorrectly. We introduce an open and extensible benchmark that follows best practices and avoids common mistakes when comparing AutoML frameworks. We conduct a thorough comparison of 9 well-known AutoML frameworks across 71 classification and 33 regression tasks. The differences between the AutoML frameworks are explored with a multi-faceted analysis, evaluating model accuracy, its trade-offs with inference time, and framework failures. We also use Bradley-Terry trees to discover subsets of tasks where the relative AutoML framework rankings differ. The benchmark comes with an open-source tool that integrates with many AutoML frameworks and automates the empirical evaluation process end-to-end: from framework installation and resource allocation to in-depth evaluation. The benchmark uses public data sets, can be easily extended with other AutoML frameworks and tasks, and has a website with up-to-date results.
3D hand-mesh reconstruction from RGB images facilitates many applications, including augmented reality (AR). However, this requires not only real-time speed and accurate hand pose and shape but also plausible mesh-image alignment. While existing works already achieve promising results, meeting all three requirements is very challenging. This paper presents a novel pipeline by decoupling the hand-mesh reconstruction task into three stages: a joint stage to predict hand joints and segmentation; a mesh stage to predict a rough hand mesh; and a refine stage to fine-tune it with an offset mesh for mesh-image alignment. With careful design in the network structure and in the loss functions, we can promote high-quality finger-level mesh-image alignment and drive the models together to deliver real-time predictions. Extensive quantitative and qualitative results on benchmark datasets demonstrate that the quality of our results outperforms the state-of-the-art methods on hand-mesh/pose precision and hand-image alignment. In the end, we also showcase several real-time AR scenarios.