Automatic differentiation and the optimization of differential equation models in biology

A computational revolution unleashed the power of artificial neural networks. At the heart of that revolution is automatic differentiation, which calculates the derivative of a performance measure relative to a large number of parameters. Differentiation enhances the discovery of improved performance in large models, an achievement that was previously difficult or impossible. Recently, a second computational advance optimizes the temporal trajectories traced by differential equations. Optimization requires differentiating a measure of performance over a trajectory, such as the closeness of tracking the environment, with respect to the parameters of the differential equations. Because model trajectories are usually calculated numerically by multistep algorithms, such as Runge-Kutta, the automatic differentiation must be passed through the numerical algorithm. This article explains how such automatic differentiation of trajectories is achieved. It also discusses why such computational breakthroughs are likely to advance theoretical and statistical studies of biological problems, in which one can consider variables as dynamic paths over time and space. Many common problems arise between improving success in computational learning models over performance landscapes, improving evolutionary fitness over adaptive landscapes, and improving statistical fits to data over information landscapes.

Towards a Rigorous Evaluation of Time-series Anomaly Detection

In recent years, proposed studies on time-series anomaly detection (TAD) report high F1 scores on benchmark TAD datasets, giving the impression of clear improvements. However, most studies apply a peculiar evaluation protocol called point adjustment (PA) before scoring. In this paper, we theoretically and experimentally reveal that the PA protocol has a great possibility of overestimating the detection performance; that is, even a random anomaly score can easily turn into a state-of-the-art TAD method. Therefore, the comparison of TAD methods with F1 scores after the PA protocol can lead to misguided rankings. Furthermore, we question the potential of existing TAD methods by showing that an untrained model obtains comparable detection performance to the existing methods even without PA. Based on our findings, we propose a new baseline and an evaluation protocol. We expect that our study will help a rigorous evaluation of TAD and lead to further improvement in future researches.

A State Transition Model for Mobile Notifications via Survival Analysis

Mobile notifications have become a major communication channel for social networking services to keep users informed and engaged. As more mobile applications push notifications to users, they constantly face decisions on what to send, when and how. A lack of research and methodology commonly leads to heuristic decision making. Many notifications arrive at an inappropriate moment or introduce too many interruptions, failing to provide value to users and spurring users' complaints. In this paper we explore unique features of interactions between mobile notifications and user engagement. We propose a state transition framework to quantitatively evaluate the effectiveness of notifications. Within this framework, we develop a survival model for badging notifications assuming a log-linear structure and a Weibull distribution. Our results show that this model achieves more flexibility for applications and superior prediction accuracy than a logistic regression model. In particular, we provide an online use case on notification delivery time optimization to show how we make better decisions, drive more user engagement, and provide more value to users.

Robust Constrained Multi-objective Evolutionary Algorithm based on Polynomial Chaos Expansion for Trajectory Optimization

An integrated optimization method based on the constrained multi-objective evolutionary algorithm (MOEA) and non-intrusive polynomial chaos expansion (PCE) is proposed, which solves robust multi-objective optimization problems under time-series dynamics. The constraints in such problems are difficult to handle, not only because the number of the dynamic constraints is multiplied by the discretized time steps but also because each of them is probabilistic. The proposed method rewrites a robust formulation into a deterministic problem via the PCE, and then sequentially processes the generated constraints in population generation, trajectory generation, and evaluation by the MOEA. As a case study, the landing trajectory design of supersonic transport (SST) with wind uncertainty is optimized. Results demonstrate the quantitative influence of the constraint values over the optimized solution sets and corresponding trajectories, proposing robust flight controls.

Landscape Learning for Neural Network Inversion

Many machine learning methods operate by inverting a neural network at inference time, which has become a popular technique for solving inverse problems in computer vision, robotics, and graphics. However, these methods often involve gradient descent through a highly non-convex loss landscape, causing the optimization process to be unstable and slow. We introduce a method that learns a loss landscape where gradient descent is efficient, bringing massive improvement and acceleration to the inversion process. We demonstrate this advantage on a number of methods for both generative and discriminative tasks, including GAN inversion, adversarial defense, and 3D human pose reconstruction.

Cut and Continuous Paste towards Real-time Deep Fall Detection

Deep learning based fall detection is one of the crucial tasks for intelligent video surveillance systems, which aims to detect unintentional falls of humans and alarm dangerous situations. In this work, we propose a simple and efficient framework to detect falls through a single and small-sized convolutional neural network. To this end, we first introduce a new image synthesis method that represents human motion in a single frame. This simplifies the fall detection task as an image classification task. Besides, the proposed synthetic data generation method enables to generate a sufficient amount of training dataset, resulting in satisfactory performance even with the small model. At the inference step, we also represent real human motion in a single image by estimating mean of input frames. In the experiment, we conduct both qualitative and quantitative evaluations on URFD and AIHub airport datasets to show the effectiveness of our method.

Conditionally Elicitable Dynamic Risk Measures for Deep Reinforcement Learning

We propose a novel framework to solve risk-sensitive reinforcement learning (RL) problems where the agent optimises time-consistent dynamic spectral risk measures. Based on the notion of conditional elicitability, our methodology constructs (strictly consistent) scoring functions that are used as penalizers in the estimation procedure. Our contribution is threefold: we (i) devise an efficient approach to estimate a class of dynamic spectral risk measures with deep neural networks, (ii) prove that these dynamic spectral risk measures may be approximated to any arbitrary accuracy using deep neural networks, and (iii) develop a risk-sensitive actor-critic algorithm that uses full episodes and does not require any additional nested transitions. We compare our conceptually improved reinforcement learning algorithm with the nested simulation approach and illustrate its performance in two settings: statistical arbitrage and portfolio allocation on both simulated and real data.

Cross-Scale Vector Quantization for Scalable Neural Speech Coding

Bitrate scalability is a desirable feature for audio coding in real-time communications. Existing neural audio codecs usually enforce a specific bitrate during training, so different models need to be trained for each target bitrate, which increases the memory footprint at the sender and the receiver side and transcoding is often needed to support multiple receivers. In this paper, we introduce a cross-scale scalable vector quantization scheme (CSVQ), in which multi-scale features are encoded progressively with stepwise feature fusion and refinement. In this way, a coarse-level signal is reconstructed if only a portion of the bitstream is received, and progressively improves the quality as more bits are available. The proposed CSVQ scheme can be flexibly applied to any neural audio coding network with a mirrored auto-encoder structure to achieve bitrate scalability. Subjective results show that the proposed scheme outperforms the classical residual VQ (RVQ) with scalability. Moreover, the proposed CSVQ at 3 kbps outperforms Opus at 9 kbps and Lyra at 3kbps and it could provide a graceful quality boost with bitrate increase.

High Resolution Time-Frequency Generation with Generative Adversarial Networks

Signal representation in Time-Frequency (TF) domain is valuable in many applications including radar imaging and inverse synthetic aparture radar. TF representation allows us to identify signal components or features in a mixed time and frequency plane. There are several well-known tools, such as Wigner-Ville Distribution (WVD), Short-Time Fourier Transform (STFT) and various other variants for such a purpose. The main requirement for a TF representation tool is to give a high-resolution view of the signal such that the signal components or features are identifiable. A commonly used method is the reassignment process which reduces the cross-terms by artificially moving smoothed WVD values from their actual location to the center of the gravity for that region. In this article, we propose a novel reassignment method using the Conditional Generative Adversarial Network (CGAN). We train a CGAN to perform the reassignment process. Through examples, it is shown that the method generates high-resolution TF representations which are better than the current reassignment methods.

Actor-Critic Scheduling for Path-Aware Air-to-Ground Multipath Multimedia Delivery

Reinforcement Learning (RL) has recently found wide applications in network traffic management and control because some of its variants do not require prior knowledge of network models. In this paper, we present a novel scheduler for real-time multimedia delivery in multipath systems based on an Actor-Critic (AC) RL algorithm. We focus on a challenging scenario of real-time video streaming from an Unmanned Aerial Vehicle (UAV) using multiple wireless paths. The scheduler acting as an RL agent learns in real-time the optimal policy for path selection, path rate allocation and redundancy estimation for flow protection. The scheduler, implemented as a module of the GStreamer framework, can be used in real or simulated settings. The simulation results show that our scheduler can target a very low loss rate at the receiver by dynamically adapting in real-time the scheduling policy to the path conditions without performing training or relying on prior knowledge of network channel models.