Jun Luo

Banafsheh Rafiee, Jun Jin, Jun Luo et al. · deepmind

Auxiliary tasks have been argued to be useful for representation learning in reinforcement learning. Although many auxiliary tasks have been empirically shown to be effective for accelerating learning on the main task, it is not yet clear what makes useful auxiliary tasks. Some of the most promising results are on the pixel control, reward prediction, and the next state prediction auxiliary tasks; however, the empirical results are mixed, showing substantial improvements in some cases and marginal improvements in others. Careful investigations of how auxiliary tasks help the learning of the main task is necessary. In this paper, we take a step studying the effect of the target policies on the usefulness of the auxiliary tasks formulated as general value functions. General value functions consist of three core elements: 1) policy 2) cumulant 3) continuation function. Our focus on the role of the target policy of the auxiliary tasks is motivated by the fact that the target policy determines the behavior about which the agent wants to make a prediction and the state-action distribution that the agent is trained on, which further affects the main task learning. Our study provides insights about questions such as: Does a greedy policy result in bigger improvement gains compared to other policies? Is it best to set the auxiliary task policy to be the same as the main task policy? Does the choice of the target policy have a substantial effect on the achieved performance gain or simple strategies for setting the policy, such as using a uniformly random policy, work as well? Our empirical results suggest that: 1) Auxiliary tasks with the greedy policy tend to be useful. 2) Most policies, including a uniformly random policy, tend to improve over the baseline. 3) Surprisingly, the main task policy tends to be less useful compared to other policies.

Amirmohammad Karimi, Jun Jin, Jun Luo et al. · eth-zurich

In classic reinforcement learning algorithms, agents make decisions at discrete and fixed time intervals. The duration between decisions becomes a crucial hyperparameter, as setting it too short may increase the problem's difficulty by requiring the agent to make numerous decisions to achieve its goal while setting it too long can result in the agent losing control over the system. However, physical systems do not necessarily require a constant control frequency, and for learning agents, it is often preferable to operate with a low frequency when possible and a high frequency when necessary. We propose a framework called Continuous-Time Continuous-Options (CTCO), where the agent chooses options as sub-policies of variable durations. These options are time-continuous and can interact with the system at any desired frequency providing a smooth change of actions. We demonstrate the effectiveness of CTCO by comparing its performance to classical RL and temporal-abstraction RL methods on simulated continuous control tasks with various action-cycle times. We show that our algorithm's performance is not affected by the choice of environment interaction frequency. Furthermore, we demonstrate the efficacy of CTCO in facilitating exploration in a real-world visual reaching task for a 7 DOF robotic arm with sparse rewards.

Guangyu Sun, Matias Mendieta, Jun Luo et al.

Personalized Federated Learning (PFL) represents a promising solution for decentralized learning in heterogeneous data environments. Partial model personalization has been proposed to improve the efficiency of PFL by selectively updating local model parameters instead of aggregating all of them. However, previous work on partial model personalization has mainly focused on Convolutional Neural Networks (CNNs), leaving a gap in understanding how it can be applied to other popular models such as Vision Transformers (ViTs). In this work, we investigate where and how to partially personalize a ViT model. Specifically, we empirically evaluate the sensitivity to data distribution of each type of layer. Based on the insights that the self-attention layer and the classification head are the most sensitive parts of a ViT, we propose a novel approach called FedPerfix, which leverages plugins to transfer information from the aggregated model to the local client as a personalization. Finally, we evaluate the proposed approach on CIFAR-100, OrganAMNIST, and Office-Home datasets and demonstrate its effectiveness in improving the model's performance compared to several advanced PFL methods.

Qingfeng Lan, Yangchen Pan, Jun Luo et al.

Artificial neural networks are promising for general function approximation but challenging to train on non-independent or non-identically distributed data due to catastrophic forgetting. The experience replay buffer, a standard component in deep reinforcement learning, is often used to reduce forgetting and improve sample efficiency by storing experiences in a large buffer and using them for training later. However, a large replay buffer results in a heavy memory burden, especially for onboard and edge devices with limited memory capacities. We propose memory-efficient reinforcement learning algorithms based on the deep Q-network algorithm to alleviate this problem. Our algorithms reduce forgetting and maintain high sample efficiency by consolidating knowledge from the target Q-network to the current Q-network. Compared to baseline methods, our algorithms achieve comparable or better performance in both feature-based and image-based tasks while easing the burden of large experience replay buffers.

Quinten Bolding, Baohao Liao, Brandon James Denis et al.

Transformer models have demonstrated remarkable performance in neural machine translation (NMT). However, their vulnerability to noisy input poses a significant challenge in practical implementation, where generating clean output from noisy input is crucial. The MTNT dataset is widely used as a benchmark for evaluating the robustness of NMT models against noisy input. Nevertheless, its utility is limited due to the presence of noise in both the source and target sentences. To address this limitation, we focus on cleaning the noise from the target sentences in MTNT, making it more suitable as a benchmark for noise evaluation. Leveraging the capabilities of large language models (LLMs), we observe their impressive abilities in noise removal. For example, they can remove emojis while considering their semantic meaning. Additionally, we show that LLM can effectively rephrase slang, jargon, and profanities. The resulting datasets, called C-MTNT, exhibit significantly less noise in the target sentences while preserving the semantic integrity of the original sentences. Our human and GPT-4 evaluations also lead to a consistent conclusion that LLM performs well on this task. Lastly, experiments on C-MTNT showcased its effectiveness in evaluating the robustness of NMT models, highlighting the potential of advanced language models for data cleaning and emphasizing C-MTNT as a valuable resource.

Yuqi Liang, Jun Luo, Xiaoxi Guo et al.

In product advertising applications, the automated inpainting of backgrounds utilizing AI techniques in product images has emerged as a significant task. However, the techniques still suffer from issues such as inappropriate background and inconsistent product in generated product images, and existing approaches for evaluating the quality of generated product images are mostly inconsistent with human feedback causing the evaluation for this task to depend on manual annotation. To relieve the issues above, this paper proposes Human Feedback and Product Consistency (HFPC), which can automatically assess the generated product images based on two modules. Firstly, to solve inappropriate backgrounds, human feedback on 44,000 automated inpainting product images is collected to train a reward model based on multi-modal features extracted from BLIP and comparative learning. Secondly, to filter generated product images containing inconsistent products, a fine-tuned segmentation model is employed to segment the product of the original and generated product images and then compare the differences between the above two. Extensive experiments have demonstrated that HFPC can effectively evaluate the quality of generated product images and significantly reduce the expense of manual annotation. Moreover, HFPC achieves state-of-the-art(96.4% in precision) in comparison to other open-source visual-quality-assessment models. Dataset and code are available at: https://github.com/created-Bi/background_inpainting_products_dataset

Jun Luo, Dooman Arefan, Margarita Zuley et al.

Mammography is used as a standard screening procedure for the potential patients of breast cancer. Over the past decade, it has been shown that deep learning techniques have succeeded in reaching near-human performance in a number of tasks, and its application in mammography is one of the topics that medical researchers most concentrate on. In this work, we propose an end-to-end Curriculum Learning (CL) strategy in task space for classifying the three categories of Full-Field Digital Mammography (FFDM), namely Malignant, Negative, and False recall. Specifically, our method treats this three-class classification as a "harder" task in terms of CL, and create an "easier" sub-task of classifying False recall against the combined group of Negative and Malignant. We introduce a loss scheduler to dynamically weight the contribution of the losses from the two tasks throughout the entire training process. We conduct experiments on an FFDM datasets of 1,709 images using 5-fold cross validation. The results show that our curriculum learning strategy can boost the performance for classifying the three categories of FFDM compared to the baseline strategies for model training.

Ehsan Nezhadarya, Ehsan Taghavi, Ryan Razani et al.

While several convolution-like operators have recently been proposed for extracting features out of point clouds, down-sampling an unordered point cloud in a deep neural network has not been rigorously studied. Existing methods down-sample the points regardless of their importance for the output. As a result, some important points in the point cloud may be removed, while less valuable points may be passed to the next layers. In contrast, adaptive down-sampling methods sample the points by taking into account the importance of each point, which varies based on the application, task and training data. In this paper, we propose a permutation-invariant learning-based adaptive down-sampling layer, called Critical Points Layer (CPL), which reduces the number of points in an unordered point cloud while retaining the important points. Unlike most graph-based point cloud down-sampling methods that use $k$-NN search algorithm to find the neighbouring points, CPL is a global down-sampling method, rendering it computationally very efficient. The proposed layer can be used along with any graph-based point cloud convolution layer to form a convolutional neural network, dubbed CP-Net in this paper. We introduce a CP-Net for $3$D object classification that achieves the best accuracy for the ModelNet$40$ dataset among point cloud-based methods, which validates the effectiveness of the CPL.