Hyperspherical Normalization for Scalable Deep Reinforcement Learning

Scaling up the model size and computation has brought consistent performance improvements in supervised learning. However, this lesson often fails to apply to reinforcement learning (RL) because training the model on non-stationary data easily leads to overfitting and unstable optimization. In response, we introduce SimbaV2, a novel RL architecture designed to stabilize optimization by (i) constraining the growth of weight and feature norm by hyperspherical normalization; and (ii) using a distributional value estimation with reward scaling to maintain stable gradients under varying reward magnitudes. Using the soft actor-critic as a base algorithm, SimbaV2 scales up effectively with larger models and greater compute, achieving state-of-the-art performance on 57 continuous control tasks across 4 domains. The code is available at https://dojeon-ai.github.io/SimbaV2.

SimBa: Simplicity Bias for Scaling Up Parameters in Deep Reinforcement Learning

Recent advances in CV and NLP have been largely driven by scaling up the number of network parameters, despite traditional theories suggesting that larger networks are prone to overfitting. These large networks avoid overfitting by integrating components that induce a simplicity bias, guiding models toward simple and generalizable solutions. However, in deep RL, designing and scaling up networks have been less explored. Motivated by this opportunity, we present SimBa, an architecture designed to scale up parameters in deep RL by injecting a simplicity bias. SimBa consists of three components: (i) an observation normalization layer that standardizes inputs with running statistics, (ii) a residual feedforward block to provide a linear pathway from the input to output, and (iii) a layer normalization to control feature magnitudes. By scaling up parameters with SimBa, the sample efficiency of various deep RL algorithms-including off-policy, on-policy, and unsupervised methods-is consistently improved. Moreover, solely by integrating SimBa architecture into SAC, it matches or surpasses state-of-the-art deep RL methods with high computational efficiency across DMC, MyoSuite, and HumanoidBench. These results demonstrate SimBa's broad applicability and effectiveness across diverse RL algorithms and environments.

Adapting Pretrained ViTs with Convolution Injector for Visuo-Motor Control

Vision Transformers (ViT), when paired with large-scale pretraining, have shown remarkable performance across various computer vision tasks, primarily due to their weak inductive bias. However, while such weak inductive bias aids in pretraining scalability, this may hinder the effective adaptation of ViTs for visuo-motor control tasks as a result of the absence of control-centric inductive biases. Such absent inductive biases include spatial locality and translation equivariance bias which convolutions naturally offer. To this end, we introduce Convolution Injector (CoIn), an add-on module that injects convolutions which are rich in locality and equivariance biases into a pretrained ViT for effective adaptation in visuo-motor control. We evaluate CoIn with three distinct types of pretrained ViTs (CLIP, MVP, VC-1) across 12 varied control tasks within three separate domains (Adroit, MetaWorld, DMC), and demonstrate that CoIn consistently enhances control task performance across all experimented environments and models, validating the effectiveness of providing pretrained ViTs with control-centric biases.

Investigating Pre-Training Objectives for Generalization in Vision-Based Reinforcement Learning

Recently, various pre-training methods have been introduced in vision-based Reinforcement Learning (RL). However, their generalization ability remains unclear due to evaluations being limited to in-distribution environments and non-unified experimental setups. To address this, we introduce the Atari Pre-training Benchmark (Atari-PB), which pre-trains a ResNet-50 model on 10 million transitions from 50 Atari games and evaluates it across diverse environment distributions. Our experiments show that pre-training objectives focused on learning task-agnostic features (e.g., identifying objects and understanding temporal dynamics) enhance generalization across different environments. In contrast, objectives focused on learning task-specific knowledge (e.g., identifying agents and fitting reward functions) improve performance in environments similar to the pre-training dataset but not in varied ones. We publicize our codes, datasets, and model checkpoints at https://github.com/dojeon-ai/Atari-PB.

Slow and Steady Wins the Race: Maintaining Plasticity with Hare and Tortoise Networks

This study investigates the loss of generalization ability in neural networks, revisiting warm-starting experiments from Ash & Adams. Our empirical analysis reveals that common methods designed to enhance plasticity by maintaining trainability provide limited benefits to generalization. While reinitializing the network can be effective, it also risks losing valuable prior knowledge. To this end, we introduce the Hare & Tortoise, inspired by the brain's complementary learning system. Hare & Tortoise consists of two components: the Hare network, which rapidly adapts to new information analogously to the hippocampus, and the Tortoise network, which gradually integrates knowledge akin to the neocortex. By periodically reinitializing the Hare network to the Tortoise's weights, our method preserves plasticity while retaining general knowledge. Hare & Tortoise can effectively maintain the network's ability to generalize, which improves advanced reinforcement learning algorithms on the Atari-100k benchmark. The code is available at https://github.com/dojeon-ai/hare-tortoise.

Do's and Don'ts: Learning Desirable Skills with Instruction Videos

Unsupervised skill discovery is a learning paradigm that aims to acquire diverse behaviors without explicit rewards. However, it faces challenges in learning complex behaviors and often leads to learning unsafe or undesirable behaviors. For instance, in various continuous control tasks, current unsupervised skill discovery methods succeed in learning basic locomotions like standing but struggle with learning more complex movements such as walking and running. Moreover, they may acquire unsafe behaviors like tripping and rolling or navigate to undesirable locations such as pitfalls or hazardous areas. In response, we present DoDont (Do's and Don'ts), an instruction-based skill discovery algorithm composed of two stages. First, in an instruction learning stage, DoDont leverages action-free instruction videos to train an instruction network to distinguish desirable transitions from undesirable ones. Then, in the skill learning stage, the instruction network adjusts the reward function of the skill discovery algorithm to weight the desired behaviors. Specifically, we integrate the instruction network into a distance-maximizing skill discovery algorithm, where the instruction network serves as the distance function. Empirically, with less than 8 instruction videos, DoDont effectively learns desirable behaviors and avoids undesirable ones across complex continuous control tasks. Code and videos are available at https://mynsng.github.io/dodont/

Learning to Discover Skills through Guidance

In the field of unsupervised skill discovery (USD), a major challenge is limited exploration, primarily due to substantial penalties when skills deviate from their initial trajectories. To enhance exploration, recent methodologies employ auxiliary rewards to maximize the epistemic uncertainty or entropy of states. However, we have identified that the effectiveness of these rewards declines as the environmental complexity rises. Therefore, we present a novel USD algorithm, skill discovery with guidance (DISCO-DANCE), which (1) selects the guide skill that possesses the highest potential to reach unexplored states, (2) guides other skills to follow guide skill, then (3) the guided skills are dispersed to maximize their discriminability in unexplored states. Empirical evaluation demonstrates that DISCO-DANCE outperforms other USD baselines in challenging environments, including two navigation benchmarks and a continuous control benchmark. Qualitative visualizations and code of DISCO-DANCE are available at https://mynsng.github.io/discodance.

Towards Validating Long-Term User Feedbacks in Interactive Recommendation Systems

Interactive Recommender Systems (IRSs) have attracted a lot of attention, due to their ability to model interactive processes between users and recommender systems. Numerous approaches have adopted Reinforcement Learning (RL) algorithms, as these can directly maximize users' cumulative rewards. In IRS, researchers commonly utilize publicly available review datasets to compare and evaluate algorithms. However, user feedback provided in public datasets merely includes instant responses (e.g., a rating), with no inclusion of delayed responses (e.g., the dwell time and the lifetime value). Thus, the question remains whether these review datasets are an appropriate choice to evaluate the long-term effects of the IRS. In this work, we revisited experiments on IRS with review datasets and compared RL-based models with a simple reward model that greedily recommends the item with the highest one-step reward. Following extensive analysis, we can reveal three main findings: First, a simple greedy reward model consistently outperforms RL-based models in maximizing cumulative rewards. Second, applying higher weighting to long-term rewards leads to a degradation of recommendation performance. Third, user feedbacks have mere long-term effects on the benchmark datasets. Based on our findings, we conclude that a dataset has to be carefully verified and that a simple greedy baseline should be included for a proper evaluation of RL-based IRS approaches.

ST-RAP: A Spatio-Temporal Framework for Real Estate Appraisal

In this paper, we introduce ST-RAP, a novel Spatio-Temporal framework for Real estate APpraisal. ST-RAP employs a hierarchical architecture with a heterogeneous graph neural network to encapsulate temporal dynamics and spatial relationships simultaneously. Through comprehensive experiments on a large-scale real estate dataset, ST-RAP outperforms previous methods, demonstrating the significant benefits of integrating spatial and temporal aspects in real estate appraisal. Our code and dataset are available at https://github.com/dojeon-ai/STRAP.

Enhancing Generalization and Plasticity for Sample Efficient Reinforcement Learning

In Reinforcement Learning (RL), enhancing sample efficiency is crucial, particularly in scenarios when data acquisition is costly and risky. In principle, off-policy RL algorithms can improve sample efficiency by allowing multiple updates per environment interaction. However, these multiple updates often lead to overfitting, which decreases the network's ability to adapt to new data. We conduct an empirical analysis of this challenge and find that generalizability and plasticity constitute different roles in improving the model's adaptability. In response, we propose a combined usage of Sharpness-Aware Minimization (SAM) and a reset mechanism. SAM seeks wide, smooth minima, improving generalization, while the reset mechanism, through periodic reinitialization of the last few layers, consistently injects plasticity into the model. Through extensive empirical studies, we demonstrate that this combined usage improves sample efficiency and computational cost on the Atari-100k and DeepMind Control Suite benchmarks.