С. В. Колесников

Denis Tarasov, Alexander Nikulin, Dmitry Akimov et al.

CORL is an open-source library that provides thoroughly benchmarked single-file implementations of both deep offline and offline-to-online reinforcement learning algorithms. It emphasizes a simple developing experience with a straightforward codebase and a modern analysis tracking tool. In CORL, we isolate methods implementation into separate single files, making performance-relevant details easier to recognize. Additionally, an experiment tracking feature is available to help log metrics, hyperparameters, dependencies, and more to the cloud. Finally, we have ensured the reliability of the implementations by benchmarking commonly employed D4RL datasets providing a transparent source of results that can be reused for robust evaluation tools such as performance profiles, probability of improvement, or expected online performance.

Vladislav Kurenkov, Alexander Nikulin, Denis Tarasov et al.

NetHack is known as the frontier of reinforcement learning research where learning-based methods still need to catch up to rule-based solutions. One of the promising directions for a breakthrough is using pre-collected datasets similar to recent developments in robotics, recommender systems, and more under the umbrella of offline reinforcement learning (ORL). Recently, a large-scale NetHack dataset was released; while it was a necessary step forward, it has yet to gain wide adoption in the ORL community. In this work, we argue that there are three major obstacles for adoption: resource-wise, implementation-wise, and benchmark-wise. To address them, we develop an open-source library that provides workflow fundamentals familiar to the ORL community: pre-defined D4RL-style tasks, uncluttered baseline implementations, and reliable evaluation tools with accompanying configs and logs synced to the cloud.

Dmitriy Akimov, Vladislav Kurenkov, Alexander Nikulin et al.

Offline reinforcement learning aims to train a policy on a pre-recorded and fixed dataset without any additional environment interactions. There are two major challenges in this setting: (1) extrapolation error caused by approximating the value of state-action pairs not well-covered by the training data and (2) distributional shift between behavior and inference policies. One way to tackle these problems is to induce conservatism - i.e., keeping the learned policies closer to the behavioral ones. To achieve this, we build upon recent works on learning policies in latent action spaces and use a special form of Normalizing Flows for constructing a generative model, which we use as a conservative action encoder. This Normalizing Flows action encoder is pre-trained in a supervised manner on the offline dataset, and then an additional policy model - controller in the latent space - is trained via reinforcement learning. This approach avoids querying actions outside of the training dataset and therefore does not require additional regularization for out-of-dataset actions. We evaluate our method on various locomotion and navigation tasks, demonstrating that our approach outperforms recently proposed algorithms with generative action models on a large portion of datasets.

Mikhail Andronov, Sergey Kolesnikov

The evaluation of recommender systems from a practical perspective is a topic of ongoing discourse within the research community. While many current evaluation methods reduce performance to a single value metric as an easy way to compare models, it relies on the assumption that the methods' performance remains constant over time. In this study, we examine this assumption and propose the Cross-Validation Thought Time (CVTT) technique as a more comprehensive evaluation method, focusing on model performance over time. By utilizing the proposed technique, we conduct an in-depth analysis of the performance of popular RecSys algorithms. Our findings indicate that (1) the performance of the recommenders varies over time for all reviewed datasets, (2) using simple evaluation approaches can lead to a substantial decrease in performance in real-world evaluation scenarios, and (3) excessive data usage can lead to suboptimal results.

Viacheslav Sinii, Alexander Nikulin, Vladislav Kurenkov et al.

Recently, it has been shown that transformers pre-trained on diverse datasets with multi-episode contexts can generalize to new reinforcement learning tasks in-context. A key limitation of previously proposed models is their reliance on a predefined action space size and structure. The introduction of a new action space often requires data re-collection and model re-training, which can be costly for some applications. In our work, we show that it is possible to mitigate this issue by proposing the Headless-AD model that, despite being trained only once, is capable of generalizing to discrete action spaces of variable size, semantic content and order. By experimenting with Bernoulli and contextual bandits, as well as a gridworld environment, we show that Headless-AD exhibits significant capability to generalize to action spaces it has never encountered, even outperforming specialized models trained for a specific set of actions on several environment configurations. Implementation is available at: https://github.com/corl-team/headless-ad.

Stanislav Dereka, Ivan Karpukhin, Sergey Kolesnikov

Large-scale datasets are essential for the success of deep learning in image retrieval. However, manual assessment errors and semi-supervised annotation techniques can lead to label noise even in popular datasets. As previous works primarily studied annotation quality in image classification tasks, it is still unclear how label noise affects deep learning approaches to image retrieval. In this work, we show that image retrieval methods are less robust to label noise than image classification ones. Furthermore, we, for the first time, investigate different types of label noise specific to image retrieval tasks and study their effect on model performance.

Ivan Karpukhin, Stanislav Dereka, Sergey Kolesnikov

Classification tasks are usually evaluated in terms of accuracy. However, accuracy is discontinuous and cannot be directly optimized using gradient ascent. Popular methods minimize cross-entropy, hinge loss, or other surrogate losses, which can lead to suboptimal results. In this paper, we propose a new optimization framework by introducing stochasticity to a model's output and optimizing expected accuracy, i.e. accuracy of the stochastic model. Extensive experiments on linear models and deep image classification show that the proposed optimization method is a powerful alternative to widely used classification losses.

Ilya Zisman, Vladislav Kurenkov, Alexander Nikulin et al.

Recently, extensive studies in Reinforcement Learning have been carried out on the ability of transformers to adapt in-context to various environments and tasks. Current in-context RL methods are limited by their strict requirements for data, which needs to be generated by RL agents or labeled with actions from an optimal policy. In order to address this prevalent problem, we propose AD$^\varepsilon$, a new data acquisition approach that enables in-context Reinforcement Learning from noise-induced curriculum. We show that it is viable to construct a synthetic noise injection curriculum which helps to obtain learning histories. Moreover, we experimentally demonstrate that it is possible to alleviate the need for generation using optimal policies, with in-context RL still able to outperform the best suboptimal policy in a learning dataset by a 2x margin.

Maksim Zhdanov, Stanislav Dereka, Sergey Kolesnikov

Uncertainty estimation is crucial in safety-critical applications, where robust out-of-distribution (OOD) detection is essential. Traditional Bayesian methods, though effective, are often hindered by high computational demands. As an alternative, Laplace approximation offers a more practical and efficient approach to uncertainty estimation. In this paper, we introduce the Identity Curvature Laplace Approximation (ICLA), a novel method that challenges the conventional posterior covariance formulation by using identity curvature and optimizing prior precision. This innovative design significantly enhances OOD detection performance on well-known datasets such as CIFAR-10, CIFAR-100, and ImageNet, while maintaining calibration scores. We attribute this improvement to the alignment issues between typical feature embeddings and curvature as measured by the Fisher information matrix. Our findings are further supported by demonstrating that incorporating Fisher penalty or sharpness-aware minimization techniques can greatly enhance the uncertainty estimation capabilities of standard Laplace approximation.

Ivan Karpukhin, Stanislav Dereka, Sergey Kolesnikov

In recent years, deep metric learning and its probabilistic extensions claimed state-of-the-art results in the face verification task. Despite improvements in face verification, probabilistic methods received little attention in the research community and practical applications. In this paper, we, for the first time, perform an in-depth analysis of known probabilistic methods in verification and retrieval tasks. We study different design choices and propose a simple extension, achieving new state-of-the-art results among probabilistic methods. Finally, we study confidence prediction and show that it correlates with data quality, but contains little information about prediction error probability. We thus provide a new confidence evaluation benchmark and establish a baseline for future confidence prediction research. PyTorch implementation is publicly released.

Sergey Kolesnikov, Oleg Lashinin, Michail Pechatov et al.

Over the past decade, tremendous progress has been made in Recommender Systems (RecSys) for well-known tasks such as next-item and next-basket prediction. On the other hand, the recently proposed next-period recommendation (NPR) task is not covered as much. Current works about NPR are mostly based around distinct problem formulations, methods, and proprietary datasets, making solutions difficult to reproduce. In this article, we aim to fill the gap in RecSys methods evaluation on the NPR task using publicly available datasets and (1) introduce the TTRS, a large-scale financial transactions dataset suitable for RecSys methods evaluation; (2) benchmark popular RecSys approaches on several datasets for the NPR task. When performing our analysis, we found a strong repetitive consumption pattern in several real-world datasets. With this setup, our results suggest that the repetitive nature of data is still hard to generalize for the evaluated RecSys methods, and novel item prediction performance is still questionable.

Vladislav Kurenkov, Sergey Kolesnikov

In this work, we argue for the importance of an online evaluation budget for a reliable comparison of deep offline RL algorithms. First, we delineate that the online evaluation budget is problem-dependent, where some problems allow for less but others for more. And second, we demonstrate that the preference between algorithms is budget-dependent across a diverse range of decision-making domains such as Robotics, Finance, and Energy Management. Following the points above, we suggest reporting the performance of deep offline RL algorithms under varying online evaluation budgets. To facilitate this, we propose to use a reporting tool from the NLP field, Expected Validation Performance. This technique makes it possible to reliably estimate expected maximum performance under different budgets while not requiring any additional computation beyond hyperparameter search. By employing this tool, we also show that Behavioral Cloning is often more favorable to offline RL algorithms when working within a limited budget.

Łukasz Kidziński, Carmichael Ong, Sharada Prasanna Mohanty et al.

In the NeurIPS 2018 Artificial Intelligence for Prosthetics challenge, participants were tasked with building a controller for a musculoskeletal model with a goal of matching a given time-varying velocity vector. Top participants were invited to describe their algorithms. In this work, we describe the challenge and present thirteen solutions that used deep reinforcement learning approaches. Many solutions use similar relaxations and heuristics, such as reward shaping, frame skipping, discretization of the action space, symmetry, and policy blending. However, each team implemented different modifications of the known algorithms by, for example, dividing the task into subtasks, learning low-level control, or by incorporating expert knowledge and using imitation learning.

Łukasz Kidziński, Sharada Prasanna Mohanty, Carmichael Ong et al.

In the NIPS 2017 Learning to Run challenge, participants were tasked with building a controller for a musculoskeletal model to make it run as fast as possible through an obstacle course. Top participants were invited to describe their algorithms. In this work, we present eight solutions that used deep reinforcement learning approaches, based on algorithms such as Deep Deterministic Policy Gradient, Proximal Policy Optimization, and Trust Region Policy Optimization. Many solutions use similar relaxations and heuristics, such as reward shaping, frame skipping, discretization of the action space, symmetry, and policy blending. However, each of the eight teams implemented different modifications of the known algorithms.