Contextual bandits are a modern staple tool for active sequential experimentation in the tech industry. They involve online learning algorithms that adaptively (over time) learn policies to map observed contexts $X_t$ to actions $A_t$ in an attempt to maximize stochastic rewards $R_t$. This adaptivity raises interesting but hard statistical inference questions, especially counterfactual ones: for example, it is often of interest to estimate the properties of a hypothetical policy that is different from the logging policy that was used to collect the data -- a problem known as "off-policy evaluation" (OPE). Using modern martingale techniques, we present a comprehensive framework for OPE inference that relax many unnecessary assumptions made in past work, significantly improving on them theoretically and empirically. Our methods remain valid in very general settings, and can be employed while the original experiment is still running (that is, not necessarily post-hoc), when the logging policy may be itself changing (due to learning), and even if the context distributions are drifting over time. More concretely, we derive confidence sequences for various functionals of interest in OPE. These include doubly robust ones for time-varying off-policy mean reward values, but also confidence bands for the entire CDF of the off-policy reward distribution. All of our methods (a) are valid at arbitrary stopping times (b) only make nonparametric assumptions, and (c) do not require known bounds on the maximal importance weights, and (d) adapt to the empirical variance of the reward and weight distributions. In summary, our methods enable anytime-valid off-policy inference using adaptively collected contextual bandit data.
A limiting factor for readout fidelity for superconducting qubits is the relaxation of the qubit to the ground state before the time needed for the resonator to reach its final target state. A technique known as excited state promoted (ESP) readout was proposed to reduce this effect and further improve the readout contrast on superconducting hardware. In this work, we use readout data from five-qubit IBMQ devices to measure the effectiveness of using deep neural networks, like feedforward neural networks, and various classification algorithms, like k-nearest neighbors, decision trees, and Gaussian naive Bayes, for single-qubit and multi-qubit discrimination. These methods were compared to standardly used linear and quadratic discriminant analysis algorithms based on their qubit-state-assignment fidelity performance, robustness to readout crosstalk, and training time.
Visual-based target tracking is easily influenced by multiple factors, such as background clutter, targets fast-moving, illumination variation, object shape change, occlusion, etc. These factors influence the tracking accuracy of a target tracking task. To address this issue, an efficient real-time target tracking method based on a low-dimension adaptive feature fusion is proposed to allow us the simultaneous implementation of the high-accuracy and real-time target tracking. First, the adaptive fusion of a histogram of oriented gradient (HOG) feature and color feature is utilized to improve the tracking accuracy. Second, a convolution dimension reduction method applies to the fusion between the HOG feature and color feature to reduce the over-fitting caused by their high-dimension fusions. Third, an average correlation energy estimation method is used to extract the relative confidence adaptive coefficients to ensure tracking accuracy. We experimentally confirm the proposed method on an OTB100 data set. Compared with nine popular target tracking algorithms, the proposed algorithm gains the highest tracking accuracy and success tracking rate. Compared with the traditional Sum of Template and Pixel-wise LEarners (STAPLE) algorithm, the proposed algorithm can obtain a higher success rate and accuracy, improving by 0.023 and 0.019, respectively. The experimental results also demonstrate that the proposed algorithm can reach the real-time target tracking with 50 fps. The proposed method paves a more promising way for real-time target tracking tasks under a complex environment, such as appearance deformation, illumination change, motion blur, background, similarity, scale change, and occlusion.
A backdoor attack places triggers in victims' deep learning models to enable a targeted misclassification at testing time. In general, triggers are fixed artifacts attached to samples, making backdoor attacks easy to spot. Only recently, a new trigger generation harder to detect has been proposed: the stylistic triggers that apply stylistic transformations to the input samples (e.g., a specific writing style). Currently, stylistic backdoor literature lacks a proper formalization of the attack, which is established in this paper. Moreover, most studies of stylistic triggers focus on text and images, while there is no understanding of whether they can work in sound. This work fills this gap. We propose JingleBack, the first stylistic backdoor attack based on audio transformations such as chorus and gain. Using 444 models in a speech classification task, we confirm the feasibility of stylistic triggers in audio, achieving 96% attack success.
In this paper, we present a novel approach to adapt a sequence-to-sequence Transformer-Transducer ASR system to the keyword spotting (KWS) task. We achieve this by replacing the keyword in the text transcription with a special token <kw> and training the system to detect the <kw> token in an audio stream. At inference time, we create a decision function inspired by conventional KWS approaches, to make our approach more suitable for the KWS task. Furthermore, we introduce a specific keyword spotting loss by adapting the sequence-discriminative Minimum Bayes-Risk training technique. We find that our approach significantly outperforms ASR based KWS systems. When compared with a conventional keyword spotting system, our proposal has similar performance while bringing the advantages and flexibility of sequence-to-sequence training. Additionally, when combined with the conventional KWS system, our approach can improve the performance at any operation point.
We present HOReeNet, which tackles the novel task of manipulating images involving hands, objects, and their interactions. Especially, we are interested in transferring objects of source images to target images and manipulating 3D hand postures to tightly grasp the transferred objects. Furthermore, the manipulation needs to be reflected in the 2D image space. In our reenactment scenario involving hand-object interactions, 3D reconstruction becomes essential as 3D contact reasoning between hands and objects is required to achieve a tight grasp. At the same time, to obtain high-quality 2D images from 3D space, well-designed 3D-to-2D projection and image refinement are required. Our HOReeNet is the first fully differentiable framework proposed for such a task. On hand-object interaction datasets, we compared our HOReeNet to the conventional image translation algorithms and reenactment algorithm. We demonstrated that our approach could achieved the state-of-the-art on the proposed task.
Hierarchical decomposition of control is unavoidable in large dynamical systems. In reinforcement learning (RL), it is usually solved with subgoals defined at higher policy levels and achieved at lower policy levels. Reaching these goals can take a substantial amount of time, during which it is not verified whether they are still worth pursuing. However, due to the randomness of the environment, these goals may become obsolete. In this paper, we address this gap in the state-of-the-art approaches and propose a method in which the validity of higher-level actions (thus lower-level goals) is constantly verified at the higher level. If the actions, i.e. lower level goals, become inadequate, they are replaced by more appropriate ones. This way we combine the advantages of hierarchical RL, which is fast training, and flat RL, which is immediate reactivity. We study our approach experimentally on seven benchmark environments.
Actively monitoring machine learning models during production operations helps ensure prediction quality and detection and remediation of unexpected or undesired conditions. Monitoring models already deployed in big data environments brings the additional challenges of adding monitoring in parallel to the existing modelling workflow and controlling resource requirements. In this paper, we describe (1) a framework for monitoring machine learning models; and, (2) its implementation for a big data supply chain application. We use our implementation to study drift in model features, predictions, and performance on three real data sets. We compare hypothesis test and information theoretic approaches to drift detection in features and predictions using the Kolmogorov-Smirnov distance and Bhattacharyya coefficient. Results showed that model performance was stable over the evaluation period. Features and predictions showed statistically significant drifts; however, these drifts were not linked to changes in model performance during the time of our study.
Reinforcement learning has recently been used to approach well-known NP-hard combinatorial problems in graph theory. Among these problems, Hamiltonian cycle problems are exceptionally difficult to analyze, even when restricted to individual instances of structurally complex graphs. In this paper, we use Monte Carlo Tree Search (MCTS), the search algorithm behind many state-of-the-art reinforcement learning algorithms such as AlphaZero, to create autonomous agents that learn to play the game of Snake, a game centered on properties of Hamiltonian cycles on grid graphs. The game of Snake can be formulated as a single-player discounted Markov Decision Process (MDP) where the agent must behave optimally in a stochastic environment. Determining the optimal policy for Snake, defined as the policy that maximizes the probability of winning - or win rate - with higher priority and minimizes the expected number of time steps to win with lower priority, is conjectured to be NP-hard. Performance-wise, compared to prior work in the Snake game, our algorithm is the first to achieve a win rate over $0.5$ (a uniform random policy achieves a win rate $< 2.57 \times 10^{-15}$), demonstrating the versatility of AlphaZero in approaching NP-hard environments.
Large-scale transformer models have become the de-facto architectures for various machine learning applications, e.g., CV and NLP. However, those large models also introduce prohibitive training costs. To mitigate this issue, we propose a novel random and layerwise token dropping method (random-LTD), which skips the computation of a subset of the input tokens at all middle layers. Particularly, random-LTD achieves considerable speedups and comparable accuracy as the standard training baseline. Compared to other token dropping methods, random-LTD does not require (1) any importance score-based metrics, (2) any special token treatment (e.g., [CLS]), and (3) many layers in full sequence length training except the first and the last layers. Besides, a new LayerToken learning rate schedule is proposed for pretraining problems that resolve the heavy tuning requirement for our proposed training mechanism. Finally, we demonstrate that random-LTD can be applied to broader applications, including GPT and BERT pretraining as well as ViT and GPT finetuning tasks. Our results show that random-LTD can save about 33.3% theoretical compute cost and 25.6% wall-clock training time while achieving similar zero-shot evaluations on GPT-31.3B as compared to baseline.