arXiv research

This study examines how learning algorithms affect collective action in machine learning.

Study shows small groups can influence machine learning algorithms.

Algorithm designs neural group actions for symmetric transformations.

The paper classifies circle actions on 6D manifolds with isolated fixed points.

In autonomous vehicle (AV) control, allowing mistakes can be quite dangerous and costly in the real world. For this reason we investigate methods of training an AV without allowing the agent to explore and instead having a human explorer collect the data. Supervised learning has been explored for AV control, but it enc…

Recourse explanations can become invalid if collective actions change statistical data.

Look-ahead reasoning helps predict strategic user behavior on learning platforms.

In this article we collect a series of observations that constrain actions of many groups on compact manifolds. In particular, we show that "generic" finitely generated groups have no smooth volume preserving actions on compact manifolds while also producing many finitely presented, torsion free groups with the same pr…

Action chunking and data exploration improve behavior cloning in robotics.

Researchers identify graph components for unicellular collections.

Let $Ω=(ω_{j})_{j\in I}$ be a collection of pairwise non-isotopic simple closed curves on the closed, orientable, genus $g$ surface $S_{g}$, such that $ω_{i}$ and $ω_{j}$ intersect exactly once for $i\neq j$. It was recently demonstrated by Malestein, Rivin, and Theran that the cardinality of such a collection is no mo…

We calculate the cohomology rings of a collection of seven dimensional manifolds supporting an S^3 x S^3-action with one dimensional orbit space. These manifolds are of interest to differential geometers studying non-negative and positive sectional curvature. From this collection, we identify several families of manifo…

Survey on finite group actions on manifolds.

Collectives can manipulate learning platforms by coordinated data submission, requiring strategic assessments and algorithms.

In his work on the Farrell-Jones Conjecture, Arthur Bartels introduced the concept of a "finitely $\mathcal{F}$-amenable" group action, where $\mathcal{F}$ is a family of subgroups. We show how a finitely $\mathcal{F}$-amenable action of a countable group $G$ on a compact metric space, where the asymptotic dimensions o…

Indices of vector fields and 1-forms studied for singular varieties and actions.

Unified stopping rules ensure accurate policies in contextual learning.

We derive a class of macroscopic differential equations that describe collective adaptation, starting from a discrete-time stochastic microscopic model. The behavior of each agent is a dynamic balance between adaptation that locally achieves the best action and memory loss that leads to randomized behavior. We show tha…

In this paper we propose a hybrid architecture of actor-critic algorithms for reinforcement learning in parameterized action space, which consists of multiple parallel sub-actor networks to decompose the structured action space into simpler action spaces along with a critic network to guide the training of all sub-acto…

Let Phi : M --> g^* be a proper moment map associated to an action of a compact connected Lie group, G, on a connected symplectic manifold, (M,ω). A collective function is a pullback via Φof a smooth function on g^*. In this paper we present four new results about the relationship between the collective functions and t…

Simplifies large action space bandits by selecting representative actions.

Paper proposes efficient sample collection strategy for RL.

Let $S_g$ denote the closed orientable surface of genus $g$. We construct exponentially many mapping class group orbits of collections of $2g+1$ simple closed curves on $S_g$ which pairwise intersect exactly once, extending a result of the first author and further answering a question of Malestein-Rivin-Theran. To dist…

We prove that for any infinite-type orientable surface S there exists a collection of essential curves Γ in S such that any homeomorphism that preserves the isotopy classes of the elements of Γ is isotopic to the identity. The collection Γ is countable and has infinite complement in C(S), the curve complex of S. As a c…

We document a mechanism operating in complex adaptive systems leading to dynamical pockets of predictability (``prediction days''), in which agents collectively take predetermined courses of action, transiently decoupled from past history. We demonstrate and test it out-of-sample on synthetic minority and majority game…

Study of laminations for pseudo-Anosov flows on three-manifolds.

A novel approach learns goal-conditioned policies for locomotion using batch RL.

Church-Ellenberg-Farb used the language of FI-modules to prove that the cohomology of certain sequences of hyperplane arrangements with S_n-actions satisfies representation stability. Here we lift their results to the level of the arrangements themselves, and define when a collection of arrangements is "finitely genera…

A new algorithm for cryo-EM data collection that balances reward and latency.

New method reduces bias and variance in OPE for large action spaces.

Twitter, a popular social network, presents great opportunities for on-line machine learning research. However, previous research has focused almost entirely on learning from passively collected data. We study the problem of learning to acquire followers through normative user behavior, as opposed to the mass following…

Poisson and symplectic structures discussed in lecture notes.

In most real-world settings such as recommender systems, finance, and healthcare, collecting useful information is costly and requires an active choice on the part of the decision maker. The decision-maker needs to learn simultaneously what observations to make and what actions to take. This paper incorporates the info…

We describe a collection of graded rings which surject onto Webster rings for sl(2) and which should be related to certain categories of singular Soergel bimodules. In the first non-trivial case, we construct a categorical braid group action which categorifies the Burau representation.

New algorithm learns policies without uniform overlap assumption.

We propose a statistical model to understand people's perception of their carbon footprint. Driven by the observation that few people think of CO2 impact in absolute terms, we design a system to probe people's perception from simple pairwise comparisons of the relative carbon footprint of their actions. The formulation…

We classify symplectic actions of 2-tori on compact, connected symplectic 4-manifolds, up to equivariant symplectomorphisms. This extends results of Atiyah, Guillemin-Sternberg, Delzant and Benoist. The classification is in terms of a collection of invariants, which are invariants of the topology of the manifold, of th…

We tackle the Multi-task Batch Reinforcement Learning problem. Given multiple datasets collected from different tasks, we train a multi-task policy to perform well in unseen tasks sampled from the same distribution. The task identities of the unseen tasks are not provided. To perform well, the policy must infer the tas…

New algorithm for reward-free RL with linear function approximation, reducing sample complexity.

StakeBench evaluates language understanding by linking comments to market commitments, improving model alignment with real-world outcomes.

Criterion for polystability in Lie group actions on manifolds.

A reinforcement learning agent tries to maximize its cumulative payoff by interacting in an unknown environment. It is important for the agent to explore suboptimal actions as well as to pick actions with highest known rewards. Yet, in sensitive domains, collecting more data with exploration is not always possible, but…

Sharp conditions link separators to R-trees for space transformations.

Study circle actions on unitary manifolds with discrete fixed points.

Unified method for balancing simulation and data collection.

Study designs logging policies to minimize off-policy evaluation error.

The paper defines infinite Schottky groups and their applications to infinite type surfaces.

Investigates offline RL in factorisable action spaces, overcoming overestimation bias.