Methods for learning to search for structured prediction typically imitate a reference policy, with existing theoretical guarantees demonstrating low regret compared to that reference. This is unsatisfactory in many applications where the reference policy is suboptimal and the goal of learning is to improve upon it. Ca…
On-device research index
arXiv research
A locally-built, LLM-digested index of recent arXiv papers in quant finance, geometry/topology, and statistical ML — keyword search served straight from SQLite on this machine.
168,742 papers · 148 categories
Trend · papers per month
19 results for “learning-to-search”
CrossBeam learns to search more efficiently in program synthesis.
problem Efficiently searching through vast program spaces.
method Trains a neural model to guide program synthesis, combining previously explored programs.
result CrossBeam explores much smaller portions of the program space compared to state-of-the-art methods.
This paper improves autoregressive model training by focusing on test metrics, not just likelihood.
problem Training autoregressive models to perform better on specific metrics like METEOR score.
method Follows the learning-to-search approach, constructing a reference policy and choosing test metric-related costs.
result The standard KL loss only learns high-probability tokens and can be improved with ranking objectives.
GLSearch uses GNN to learn efficient search strategies for finding large common subgraphs.
problem Finding the Maximum Common Subgraph (MCS) between two graphs is NP-hard and hard to solve efficiently.
method GLSearch combines GNN and DQN to learn optimal node pairs for expansion in a branch and bound algorithm.
result GLSearch finds significantly larger common subgraphs than heuristic search methods given the same computation budget.
We present a novel view that unifies two frameworks that aim to solve sequential prediction problems: learning to search (L2S) and recurrent neural networks (RNN). We point out equivalences between elements of the two frameworks. By complementing what is missing from one framework comparing to the other, we introduce a…
Transformer learns to search through reinforcement learning, mimicking DFS.
problem Understanding how transformers learn search capabilities in RL.
method Two-head transformer, depth-wise curriculum, discounted returns.
result Transformer policy generalizes depth and prioritizes high-probability branches.
Machine learning finds Z/2 eigenfunctions on a sphere.
problem Finding Z/2 eigenfunctions on the sphere.
method Created a multivalued neural network and used JAX to implement it. Fixed branch points at tetrahedron and cube vertices, and allowed AI to move them in the third case.
result Found Z/2 eigenfunctions for three cases.
We propose a general framework for sequential and dynamic acquisition of useful information in order to solve a particular task. While our goal could in principle be tackled by general reinforcement learning, our particular setting is constrained enough to allow more efficient algorithms. In this paper, we work under t…
Predicting structured outputs can be computationally onerous due to the combinatorially large output spaces. In this paper, we focus on reducing the prediction time of a trained black-box structured classifier without losing accuracy. To do so, we train a speedup classifier that learns to mimic a black-box classifier u…
We study the problem of learning a good search policy for combinatorial search spaces. We propose retrospective imitation learning, which, after initial training by an expert, improves itself by learning from \textit{retrospective inspections} of its own roll-outs. That is, when the policy eventually reaches a feasible…
Efficiently find near-optimal medical treatments with less trial and error.
problem Finding effective medical treatments through trial and error.
method Formalizes the problem, uses a causal inference framework, and proposes model-based dynamic programming and greedy algorithms.
result Our methods compare favorably to model-free reinforcement learning, offering a more transparent trade-off between search time and treatment efficacy.
We propose SEARNN, a novel training algorithm for recurrent neural networks (RNNs) inspired by the "learning to search" (L2S) approach to structured prediction. RNNs have been widely successful in structured prediction applications such as machine translation or parsing, and are commonly trained using maximum likelihoo…
The neural architecture search (NAS) algorithm with reinforcement learning can be a powerful and novel framework for the automatic discovering process of neural architectures. However, its application is restricted by noncontinuous and high-dimensional search spaces, which result in difficulty in optimization. To resol…
Planning problems are among the most important and well-studied problems in artificial intelligence. They are most typically solved by tree search algorithms that simulate ahead into the future, evaluate future states, and back-up those evaluations to the root of a search tree. Among these algorithms, Monte-Carlo tree …
Improves deep learning for Airbnb search ranking.
problem Challenges in ranking inventory and handling new listings.
method New ranking neural network architecture, positional bias handling, and cold start solutions.
result Significant improvements in inventory ranking and new listing treatment.
Statistical model checking for PCTL on MDPs using reinforcement learning.
problem Model checking PCTL specifications on MDPs with statistical methods.
method Reinforcement learning for policy search, statistical model checking with UCB-based Q-learning.
result Provably guaranteed statistical model checking method for PCTL specifications on MDPs.
New method uses reinforcement learning to calibrate financial models.
problem Finding continuous-time diffusion models that fit market option prices.
method Multi-Agent Reinforcement Learning (MARL) to search stochastic process space.
result Algorithm learns local volatility and path-dependence for Bermudan options.
ROOTS learns to represent and render 3D scenes with object-centric models.
problem Learning to represent and render 3D scenes with object-centric compositionality.
method Probabilistic generative model for learning object representations and scene rendering from partial observations.
result The model can infer 3D object representations and render scenes from arbitrary viewpoints.
Unsupervised space partitioning improves ANNS performance without pre-processing.
problem Efficient nearest neighbor search in high-dimensional spaces.
method Custom unsupervised learning framework for space partitioning and learning-to-search.
result Our method outperforms state-of-the-art approaches on ANNS benchmarks.