Search Results for author:
Found 96 papers, 60 papers with code
Learning to Progressively Plan
For problem solving, making reactive decisions based on problem description is fast but inaccurate, while search-based planning using heuristics gives better solutions but could be exponentially slow.
Student Specialization in Deep Rectified Networks With Finite Width and Input Dimension
Under mild conditions on dataset and teacher network, we prove that when the gradient is small at every data sample, each teacher node is \emph{specialized} by at least one student node \emph{at the lowest layer}.
RLCD: Reinforcement Learning from Contrast Distillation for Language Model Alignment
We propose Reinforcement Learning from Contrast Distillation (RLCD), a method for aligning language models to follow natural language principles without using human feedback.
Landscape Surrogate: Learning Decision Losses for Mathematical Optimization Under Partial Information
The implicit approach may not require optimal solutions as labels and is capable of handling problem uncertainty; however, it is slow to train and deploy due to frequent calls to optimizer $\mathbf{g}$ during both training and testing.
Extending Context Window of Large Language Models via Positional Interpolation
We present Position Interpolation (PI) that extends the context window sizes of RoPE-based pretrained LLMs such as LLaMA models to up to 32768 with minimal fine-tuning (within 1000 steps), while demonstrating strong empirical results on various tasks that require long context, including passkey retrieval, language modeling, and long document summarization from LLaMA 7B to 65B.
H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models
Based on these insights, we propose Heavy Hitter Oracle (H$_2$O), a KV cache eviction policy that dynamically retains a balance of recent and H$_2$ tokens.
Scan and Snap: Understanding Training Dynamics and Token Composition in 1-layer Transformer
More specifically, with the assumption (a) no positional encoding, (b) long input sequence, and (c) the decoder layer learns faster than the self-attention layer, we prove that self-attention acts as a \emph{discriminative scanning algorithm}: starting from uniform attention, it gradually attends more to distinct key tokens for a specific next token to be predicted, and pays less attention to common key tokens that occur across different next tokens.
Pre-train and Search: Efficient Embedding Table Sharding with Pre-trained Neural Cost Models
In this work, we explore a "pre-train, and search" paradigm for efficient sharding.
A Cookbook of Self-Supervised Learning
Self-supervised learning, dubbed the dark matter of intelligence, is a promising path to advance machine learning.
Searching Large Neighborhoods for Integer Linear Programs with Contrastive Learning
Integer Linear Programs (ILPs) are powerful tools for modeling and solving a large number of combinatorial optimization problems.
Learning Compiler Pass Orders using Coreset and Normalized Value Prediction
Finding the optimal pass sequence of compilation can lead to a significant reduction in program size and/or improvement in program efficiency.
Sample-efficient Surrogate Model for Frequency Response of Linear PDEs using Self-Attentive Complex Polynomials
Linear Partial Differential Equations (PDEs) govern the spatial-temporal dynamics of physical systems that are essential to building modern technology.
DOC: Improving Long Story Coherence With Detailed Outline Control
In human evaluations of automatically generated stories, DOC substantially outperforms a strong Re3 baseline (Yang et al., 2022) on plot coherence (22. 5% absolute gain), outline relevance (28. 2%), and interestingness (20. 7%).
Local Branching Relaxation Heuristics for Integer Linear Programs
LNS relies on heuristics to select neighborhoods to search in.
EurNet: Efficient Multi-Range Relational Modeling of Spatial Multi-Relational Data
We study EurNets in two important domains for image and protein structure modeling.
SurCo: Learning Linear Surrogates For Combinatorial Nonlinear Optimization Problems
Optimization problems with nonlinear cost functions and combinatorial constraints appear in many real-world applications but remain challenging to solve efficiently compared to their linear counterparts.
Re3: Generating Longer Stories With Recursive Reprompting and Revision
We consider the problem of automatically generating longer stories of over two thousand words.
DreamShard: Generalizable Embedding Table Placement for Recommender Systems
Although prior work has explored learning-based approaches for the device placement of computational graphs, embedding table placement remains to be a challenging problem because of 1) the operation fusion of embedding tables, and 2) the generalizability requirement on unseen placement tasks with different numbers of tables and/or devices.
Efficient Planning in a Compact Latent Action Space
Planning-based reinforcement learning has shown strong performance in tasks in discrete and low-dimensional continuous action spaces.
AutoShard: Automated Embedding Table Sharding for Recommender Systems
This is a significant design challenge of distributed systems named embedding table sharding, i. e., how we should partition the embedding tables to balance the costs across devices, which is a non-trivial task because 1) it is hard to efficiently and precisely measure the cost, and 2) the partition problem is known to be NP-hard.
Denoised MDPs: Learning World Models Better Than the World Itself
The ability to separate signal from noise, and reason with clean abstractions, is critical to intelligence.
Understanding the Role of Nonlinearity in Training Dynamics of Contrastive Learning
First, the presence of nonlinearity can lead to many local optima even in 1-layer setting, each corresponding to certain patterns from the data distribution, while with linear activation, only one major pattern can be learned.
On the Importance of Asymmetry for Siamese Representation Learning
Many recent self-supervised frameworks for visual representation learning are based on certain forms of Siamese networks.
Understanding Curriculum Learning in Policy Optimization for Solving Combinatorial Optimization Problems
For a canonical online CO problem, Secretary Problem, we formally prove that distribution shift is reduced exponentially with curriculum learning even if the curriculum is randomly generated.
Understanding Deep Contrastive Learning via Coordinate-wise Optimization
We show that Contrastive Learning (CL) under a broad family of loss functions (including InfoNCE) has a unified formulation of coordinate-wise optimization on the network parameter $\boldsymbol{\theta}$ and pairwise importance $\alpha$, where the \emph{max player} $\boldsymbol{\theta}$ learns representation for contrastiveness, and the \emph{min player} $\alpha$ puts more weights on pairs of distinct samples that share similar representations.
Learning Bounded Context-Free-Grammar via LSTM and the Transformer:Difference and Explanations
With the forced decomposition, we show that the performance upper bounds of LSTM and Transformer in learning CFL are close: both of them can simulate a stack and perform stack operation along with state transitions.
NovelD: A Simple yet Effective Exploration Criterion
We analyze NovelD thoroughly in MiniGrid and found that empirically it helps the agent explore the environment more uniformly with a focus on exploring beyond the boundary.
DyFormer: A Scalable Dynamic Graph Transformer with Provable Benefits on Generalization Ability
To achieve efficient and scalable training, we propose temporal-union graph structure and its associated subgraph-based node sampling strategy.
Understanding Dimensional Collapse in Contrastive Self-supervised Learning
It has been shown that non-contrastive methods suffer from a lesser collapse problem of a different nature: dimensional collapse, whereby the embedding vectors end up spanning a lower-dimensional subspace instead of the entire available embedding space.
Towards Demystifying Representation Learning with Non-contrastive Self-supervision
Non-contrastive methods of self-supervised learning (such as BYOL and SimSiam) learn representations by minimizing the distance between two views of the same image.
Multi-objective Optimization by Learning Space Partitions
In this paper, we propose LaMOO, a novel multi-objective optimizer that learns a model from observed samples to partition the search space and then focus on promising regions that are likely to contain a subset of the Pareto frontier.
Multi-objective Optimization by Learning Space Partition
In this paper, we propose LaMOO, a novel multi-objective optimizer that learns a model from observed samples to partition the search space and then focus on promising regions that are likely to contain a subset of the Pareto frontier.
NASViT: Neural Architecture Search for Efficient Vision Transformers with Gradient Conflict aware Supernet Training
In this work, we observe that the poor performance is due to a gradient conflict issue: the gradients of different sub-networks conflict with that of the supernet more severely in ViTs than CNNs, which leads to early saturation in training and inferior convergence.
Ranked #7 on
Neural Architecture Search
on ImageNet
CompilerGym: Robust, Performant Compiler Optimization Environments for AI Research
What is needed is an easy, reusable experimental infrastructure for real world compiler optimization tasks that can serve as a common benchmark for comparing techniques, and as a platform to accelerate progress in the field.
Latent Execution for Neural Program Synthesis
While recent works demonstrated limited success on domain-specific languages (DSL), it remains highly challenging to apply them to real-world programming languages, such as C. Due to complicated syntax and token variation, there are three major challenges: (1) unlike many DSLs, programs in languages like C need to compile first and are not executed via interpreters; (2) the program search space grows exponentially when the syntax and semantics of the programming language become more complex; and (3) collecting a large-scale dataset of real-world programs is non-trivial.
Learning Space Partitions for Path Planning
Path planning, the problem of efficiently discovering high-reward trajectories, often requires optimizing a high-dimensional and multimodal reward function.
MADE: Exploration via Maximizing Deviation from Explored Regions
As a proof of concept, we evaluate the new intrinsic reward on tabular examples across a variety of model-based and model-free algorithms, showing improvements over count-only exploration strategies.
Latent Execution for Neural Program Synthesis Beyond Domain-Specific Languages
Program synthesis from input-output (IO) examples has been a long-standing challenge.
Understanding Robustness in Teacher-Student Setting: A New Perspective
Extensive experiments show that student specialization correlates strongly with model robustness in different scenarios, including student trained via standard training, adversarial training, confidence-calibrated adversarial training, and training with robust feature dataset.
Provably Efficient Policy Optimization for Two-Player Zero-Sum Markov Games
Policy-based methods with function approximation are widely used for solving two-player zero-sum games with large state and/or action spaces.
Understanding self-supervised Learning Dynamics without Contrastive Pairs
While contrastive approaches of self-supervised learning (SSL) learn representations by minimizing the distance between two augmented views of the same data point (positive pairs) and maximizing views from different data points (negative pairs), recent \emph{non-contrastive} SSL (e. g., BYOL and SimSiam) show remarkable performance {\it without} negative pairs, with an extra learnable predictor and a stop-gradient operation.
N-Bref : A High-fidelity Decompiler Exploiting Programming Structures
In software development, decompilation aims to reverse engineer binary executables.
BeBold: Exploration Beyond the Boundary of Explored Regions
In this paper, we analyze the pros and cons of each method and propose the regulated difference of inverse visitation counts as a simple but effective criterion for IR.
FP-NAS: Fast Probabilistic Neural Architecture Search
Furthermore, to search fast in the multi-variate space, we propose a coarse-to-fine strategy by using a factorized distribution at the beginning which can reduce the number of architecture parameters by over an order of magnitude.
Multi-Agent Collaboration via Reward Attribution Decomposition
In this work, we propose Collaborative Q-learning (CollaQ) that achieves state-of-the-art performance in the StarCraft multi-agent challenge and supports ad hoc team play.
Understanding Self-supervised Learning with Dual Deep Networks
We propose a novel theoretical framework to understand contrastive self-supervised learning (SSL) methods that employ dual pairs of deep ReLU networks (e. g., SimCLR).
Real-world Video Adaptation with Reinforcement Learning
Client-side video players employ adaptive bitrate (ABR) algorithms to optimize user quality of experience (QoE).
Joint Policy Search for Multi-agent Collaboration with Imperfect Information
Based on this, we propose Joint Policy Search(JPS) that iteratively improves joint policies of collaborative agents in imperfect information games, without re-evaluating the entire game.
Learning Search Space Partition for Black-box Optimization using Monte Carlo Tree Search
If the nonlinear partition function and the local model fits well with ground-truth black-box function, then good partitions and candidates can be reached with much fewer samples.
Towards Automated Neural Interaction Discovery for Click-Through Rate Prediction
Click-Through Rate (CTR) prediction is one of the most important machine learning tasks in recommender systems, driving personalized experience for billions of consumers.
Few-shot Neural Architecture Search
supernet, to approximate the performance of every architecture in the search space via weight-sharing.
FBNetV3: Joint Architecture-Recipe Search using Predictor Pretraining
To address this, we present Neural Architecture-Recipe Search (NARS) to search both (a) architectures and (b) their corresponding training recipes, simultaneously.
Ranked #5 on
Neural Architecture Search
on ImageNet
Deep Symbolic Superoptimization Without Human Knowledge
Deep symbolic superoptimization refers to the task of applying deep learning methods to simplify symbolic expressions.
FBNetV2: Differentiable Neural Architecture Search for Spatial and Channel Dimensions
We propose a masking mechanism for feature map reuse, so that memory and computational costs stay nearly constant as the search space expands.
Ranked #68 on
Neural Architecture Search
on ImageNet
Coda: An End-to-End Neural Program Decompiler
Furthermore, Coda outperforms the sequence-to-sequence model with attention by a margin of 70% program accuracy.
Student Specialization in Deep ReLU Networks With Finite Width and Input Dimension
Under mild conditions on dataset and teacher network, we prove that when the gradient is small at every data sample, each teacher node is \emph{specialized} by at least one student node \emph{at the lowest layer}.
A⋆MCTS: SEARCH WITH THEORETICAL GUARANTEE USING POLICY AND VALUE FUNCTIONS
We apply our theoretical framework to different models for the noise distribution of the policy and value network as well as the distribution of rewards, and show that for these general models, the sample complexity is polynomial in D, where D is the depth of the search tree.
Toward Understanding Generalization of Over-parameterized Deep ReLU network trained with SGD in Student-teacher Setting
Our analysis shows that over-parameterization plays two roles: (1) it is a necessary condition for alignment to happen at the critical points, and (2) in training dynamics, it helps student nodes cover more teacher nodes with fewer iterations.
Mean Field Models for Neural Networks in Teacher-student Setting
For the two-layer networks, we derive the necessary condition of the stationary distributions of the mean field equation and explain an empirical phenomenon concerning training speed differences using the Wasserstein flow description.
Neural Architecture Search by Learning Action Space for Monte Carlo Tree Search
As a result, using manually designed action space to perform NAS often leads to sample-inefficient explorations of architectures and thus can be sub-optimal.
Simple is Better: Training an End-to-end Contract Bridge Bidding Agent without Human Knowledge
While playing is relatively easy for modern software, bidding is challenging and requires agents to learn a communication protocol to reach the optimal contract jointly, with their own private information.
All Simulations Are Not Equal: Simulation Reweighing for Imperfect Information Games
We use simulation reweighing in the playing phase of the game contract bridge, and show that it outperforms previous state-of-the-art Monte Carlo simulation based methods, and achieves better play per decision.
Bayesian Relational Memory for Semantic Visual Navigation
We introduce a new memory architecture, Bayesian Relational Memory (BRM), to improve the generalization ability for semantic visual navigation agents in unseen environments, where an agent is given a semantic target to navigate towards.
A Neural-based Program Decompiler
Reverse engineering of binary executables is a critical problem in the computer security domain.
Sample-Efficient Neural Architecture Search by Learning Action Space
To improve the sample efficiency, this paper proposes Latent Action Neural Architecture Search (LaNAS), which learns actions to recursively partition the search space into good or bad regions that contain networks with similar performance metrics.
Playing the lottery with rewards and multiple languages: lottery tickets in RL and NLP
The lottery ticket hypothesis proposes that over-parameterization of deep neural networks (DNNs) aids training by increasing the probability of a "lucky" sub-network initialization being present rather than by helping the optimization process (Frankle & Carbin, 2019).
One ticket to win them all: generalizing lottery ticket initializations across datasets and optimizers
Perhaps surprisingly, we found that, within the natural images domain, winning ticket initializations generalized across a variety of datasets, including Fashion MNIST, SVHN, CIFAR-10/100, ImageNet, and Places365, often achieving performance close to that of winning tickets generated on the same dataset.
Hierarchical Decision Making by Generating and Following Natural Language Instructions
We explore using latent natural language instructions as an expressive and compositional representation of complex actions for hierarchical decision making.
Luck Matters: Understanding Training Dynamics of Deep ReLU Networks
We analyze the dynamics of training deep ReLU networks and their implications on generalization capability.
A theoretical framework for deep and locally connected ReLU network
In this paper, we propose a novel theoretical framework for such networks with ReLU nonlinearity.
M^3RL: Mind-aware Multi-agent Management Reinforcement Learning
Most of the prior work on multi-agent reinforcement learning (MARL) achieves optimal collaboration by directly controlling the agents to maximize a common reward.
AlphaX: eXploring Neural Architectures with Deep Neural Networks and Monte Carlo Tree Search
Neural Architecture Search (NAS) has shown great success in automating the design of neural networks, but the prohibitive amount of computations behind current NAS methods requires further investigations in improving the sample efficiency and the network evaluation cost to get better results in a shorter time.
Ranked #7 on
Neural Architecture Search
on CIFAR-10 Image Classification
(Params metric)
ELF OpenGo: An Analysis and Open Reimplementation of AlphaZero
The AlphaGo, AlphaGo Zero, and AlphaZero series of algorithms are remarkable demonstrations of deep reinforcement learning's capabilities, achieving superhuman performance in the complex game of Go with progressively increasing autonomy.
Sample-Efficient Neural Architecture Search by Learning Action Space for Monte Carlo Tree Search
To improve the sample efficiency, this paper proposes Latent Action Neural Architecture Search (LaNAS), which learns actions to recursively partition the search space into good or bad regions that contain networks with similar performance metrics.
Ranked #19 on
Image Classification
on CIFAR-10
FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search
Due to this, previous neural architecture search (NAS) methods are computationally expensive.
Ranked #846 on
Image Classification
on ImageNet
Mixed Precision Quantization of ConvNets via Differentiable Neural Architecture Search
Recent work in network quantization has substantially reduced the time and space complexity of neural network inference, enabling their deployment on embedded and mobile devices with limited computational and memory resources.
Learning to Perform Local Rewriting for Combinatorial Optimization
Search-based methods for hard combinatorial optimization are often guided by heuristics.
M$^3$RL: Mind-aware Multi-agent Management Reinforcement Learning
Most of the prior work on multi-agent reinforcement learning (MARL) achieves optimal collaboration by directly controlling the agents to maximize a common reward.
Learning and Planning with a Semantic Model
Building deep reinforcement learning agents that can generalize and adapt to unseen environments remains a fundamental challenge for AI.
A theoretical framework for deep locally connected ReLU network
Understanding theoretical properties of deep and locally connected nonlinear network, such as deep convolutional neural network (DCNN), is still a hard problem despite its empirical success.
Algorithmic Framework for Model-based Deep Reinforcement Learning with Theoretical Guarantees
Model-based reinforcement learning (RL) is considered to be a promising approach to reduce the sample complexity that hinders model-free RL.
Neural Architecture Search using Deep Neural Networks and Monte Carlo Tree Search
Neural Architecture Search (NAS) has shown great success in automating the design of neural networks, but the prohibitive amount of computations behind current NAS methods requires further investigations in improving the sample efficiency and the network evaluation cost to get better results in a shorter time.
3D Interpreter Networks for Viewer-Centered Wireframe Modeling
3D-INN is trained on real images to estimate 2D keypoint heatmaps from an input image; it then predicts 3D object structure from heatmaps using knowledge learned from synthetic 3D shapes.
Building Generalizable Agents with a Realistic and Rich 3D Environment
To generalize to unseen environments, an agent needs to be robust to low-level variations (e. g. color, texture, object changes), and also high-level variations (e. g. layout changes of the environment).
Latent forward model for Real-time Strategy game planning with incomplete information
Model-free deep reinforcement learning approaches have shown superhuman performance in simulated environments (e. g., Atari games, Go, etc).
CoDraw: Collaborative Drawing as a Testbed for Grounded Goal-driven Communication
The game involves two players: a Teller and a Drawer.
Gradient Descent Learns One-hidden-layer CNN: Don't be Afraid of Spurious Local Minima
We consider the problem of learning a one-hidden-layer neural network with non-overlapping convolutional layer and ReLU activation, i. e., $f(\mathbf{Z}, \mathbf{w}, \mathbf{a}) = \sum_j a_j\sigma(\mathbf{w}^T\mathbf{Z}_j)$, in which both the convolutional weights $\mathbf{w}$ and the output weights $\mathbf{a}$ are parameters to be learned.
When is a Convolutional Filter Easy To Learn?
We show that (stochastic) gradient descent with random initialization can learn the convolutional filter in polynomial time and the convergence rate depends on the smoothness of the input distribution and the closeness of patches.
ELF: An Extensive, Lightweight and Flexible Research Platform for Real-time Strategy Games
In addition, our platform is flexible in terms of environment-agent communication topologies, choices of RL methods, changes in game parameters, and can host existing C/C++-based game environments like Arcade Learning Environment.
Channel-Recurrent Autoencoding for Image Modeling
Despite recent successes in synthesizing faces and bedrooms, existing generative models struggle to capture more complex image types, potentially due to the oversimplification of their latent space constructions.
An Analytical Formula of Population Gradient for two-layered ReLU network and its Applications in Convergence and Critical Point Analysis
We train our network with gradient descent on $\mathbf{w}$ to mimic the output of a teacher network with the same architecture and fixed parameters $\mathbf{w}^*$.
Single Image 3D Interpreter Network
In this work, we propose 3D INterpreter Network (3D-INN), an end-to-end framework which sequentially estimates 2D keypoint heatmaps and 3D object structure, trained on both real 2D-annotated images and synthetic 3D data.
Simple Baseline for Visual Question Answering
We describe a very simple bag-of-words baseline for visual question answering.
Better Computer Go Player with Neural Network and Long-term Prediction
Against human players, the newest versions, darkfores2, achieve a stable 3d level on KGS Go Server as a ranked bot, a substantial improvement upon the estimated 4k-5k ranks for DCNN reported in Clark & Storkey (2015) based on games against other machine players.
Semantic Amodal Segmentation
Specifically, we create an amodal segmentation of each image: the full extent of each region is marked, not just the visible pixels.
Convolutional networks and learning invariant to homogeneous multiplicative scalings
The conventional classification schemes -- notably multinomial logistic regression -- used in conjunction with convolutional networks (convnets) are classical in statistics, designed without consideration for the usual coupling with convnets, stochastic gradient descent, and backpropagation.
