Learning a Code: Machine Learning for Approximate Non-Linear Coded Computation

Machine learning algorithms are typically run on large scale, distributed compute infrastructure that routinely face a number of unavailabilities such as failures and temporary slowdowns. Adding redundant computations using coding-theoretic tools called ‘codes’ is an emerging technique to alleviate the adverse effects of such unavailabilities. A code consists of an encoding function that proactively introduces redundant computation and a decoding function that reconstructs unavailable outputs using the available ones. Past work focuses on using codes to provide resilience for linear computations and specific iterative optimization algorithms. However, computations performed for a variety of applications including inference on state-of-the-art machine learning algorithms, such as neural networks, typically fall outside this realm. In this paper, we propose taking a learning-based approach to designing codes that can handle non-linear computations. We present carefully designed neural network architectures and a training methodology for learning encoding and decoding functions that produce approximate reconstructions of unavailable computation results. We present extensive experimental results demonstrating the effectiveness of the proposed approach: we show that the our learned codes can accurately reconstruct 64 - 98\% of the unavailable predictions from neural-network based image classifiers on the MNIST, Fashion-MNIST, and CIFAR-10 datasets. To the best of our knowledge, this work proposes the first learning-based approach for designing codes, and also presents the first coding-theoretic solution that can provide resilience for any non-linear (differentiable) computation. Our results show that learning can be an effective technique for designing codes, and that learned codes are a highly promising approach for bringing the benefits of coding to non-linear computations.


Agreement-based Learning

Model selection is a problem that has occupied machine learning researchers for a long time. Recently, its importance has become evident through applications in deep learning. We propose an agreement-based learning framework that prevents many of the pitfalls associated with model selection. It relies on coupling the training of multiple models by encouraging them to agree on their predictions while training. In contrast with other model selection and combination approaches used in machine learning, the proposed framework is inspired by human learning. We also propose a learning algorithm defined within this framework which manages to significantly outperform alternatives in practice, and whose performance improves further with the availability of unlabeled data. Finally, we describe a number of potential directions for developing more flexible agreement-based learning algorithms.


Dynamically Hierarchy Revolution: DirNet for Compressing Recurrent Neural Network on Mobile Devices

Recurrent neural networks (RNNs) achieve cutting-edge performance on a variety of problems. However, due to their high computational and memory demands, deploying RNNs on resource constrained mobile devices is a challenging task. To guarantee minimum accuracy loss with higher compression rate and driven by the mobile resource requirement, we introduce a novel model compression approach DirNet based on an optimized fast dictionary learning algorithm, which 1) dynamically mines the dictionary atoms of the projection dictionary matrix within layer to adjust the compression rate 2) adaptively changes the sparsity of sparse codes cross the hierarchical layers. Experimental results on language model and an ASR model trained with a 1000h speech dataset demonstrate that our method significantly outperforms prior approaches. Evaluated on off-the-shelf mobile devices, we are able to reduce the size of original model by eight times with real-time model inference and negligible accuracy loss.


ML-Leaks: Model and Data Independent Membership Inference Attacks and Defenses on Machine Learning Models

Machine learning (ML) has become a core component of many real-world applications and training data is a key factor that drives current progress. This huge success has led Internet companies to deploy machine learning as a service (MLaaS). Recently, the first membership inference attack has shown that extraction of information on the training set is possible in such MLaaS settings, which has severe security and privacy implications. However, the early demonstrations of the feasibility of such attacks have many assumptions on the adversary such as using multiple so-called shadow models, knowledge of the target model structure and having a dataset from the same distribution as the target model’s training data. We relax all 3 key assumptions, thereby showing that such attacks are very broadly applicable at low cost and thereby pose a more severe risk than previously thought. We present the most comprehensive study so far on this emerging and developing threat using eight diverse datasets which show the viability of the proposed attacks across domains. In addition, we propose the first effective defense mechanisms against such broader class of membership inference attacks that maintain a high level of utility of the ML model.


Secure and Efficient Skyline Queries on Encrypted Data

Outsourcing data and computation to cloud server provides a cost-effective way to support large scale data storage and query processing. However, due to security and privacy concerns, sensitive data (e.g., medical records) need to be protected from the cloud server and other unauthorized users. One approach is to outsource encrypted data to the cloud server and have the cloud server perform query processing on the encrypted data only. It remains a challenging task to support various queries over encrypted data in a secure and efficient way such that the cloud server does not gain any knowledge about the data, query, and query result. In this paper, we study the problem of secure skyline queries over encrypted data. The skyline query is particularly important for multi-criteria decision making but also presents significant challenges due to its complex computations. We propose a fully secure skyline query protocol on data encrypted using semantically-secure encryption. As a key subroutine, we present a new secure dominance protocol, which can be also used as a building block for other queries. Furthermore, we demonstrate two optimizations, data partitioning and lazy merging, to further reduce the computation load. Finally, we provide both serial and parallelized implementations and empirically study the protocols in terms of efficiency and scalability under different parameter settings, verifying the feasibility of our proposed solutions.


Efficient and Scalable Batch Bayesian Optimization Using K-Means

We present K-Means Batch Bayesian Optimization (KMBBO), a novel batch sampling algorithm for Bayesian Optimization (BO). KMBBO uses unsupervised learning to efficiently estimate peaks of the model acquisition function. We show in empirical experiments that our method outperforms the current state-of-the-art batch allocation algorithms on a variety of test problems including tuning of algorithm hyper-parameters and a challenging drug discovery problem. In order to accommodate the real-world problem of high dimensional data, we propose a modification to KMBBO by combining it with compressed sensing to project the optimization into a lower dimensional subspace. We demonstrate empirically that this 2-step method is competitive with algorithms where no dimensionality reduction has taken place.


Auto-Correlation and Coherence Time of Interference in Poisson Networks

The dynamics of interference over space and time influences the performance of wireless communication systems, yet its features are still not fully understood. This article analyzes the temporal dynamics of the interference in Poisson networks accounting for three key correlation sources: the location of nodes, the wireless channel, and the network traffic. We derive expressions for the auto-correlation function of interference. These are presented as a framework that enables us to arbitrarily combine the three correlation sources to match a wide range of interference scenarios. We then introduce the interference coherence time – analogously to the well-known channel coherence time – and analyze its features for each correlation source. We find that the coherence time behaves very different for the different interference scenarios considered and depends on the network parameters. Having accurate knowledge of the coherence time can thus be an important design input for protocols, e.g., retransmission and medium access control.


iFair: Learning Individually Fair Data Representations for Algorithmic Decision Making

People are rated and ranked, towards algorithmic decision making in an increasing number of applications, typically based on machine learning. Research on how to incorporate fairness into such tasks has prevalently pursued the paradigm of group fairness: ensuring that each ethnic or social group receives its fair share in the outcome of classifiers and rankings. In contrast, the alternative paradigm of individual fairness has received relatively little attention. This paper introduces a method for probabilistically clustering user records into a low-rank representation that captures individual fairness yet also achieves high accuracy in classification and regression models. Our notion of individual fairness requires that users who are similar in all task-relevant attributes such as job qualification, and disregarding all potentially discriminating attributes such as gender, should have similar outcomes. Since the case for fairness is ubiquitous across many tasks, we aim to learn general representations that can be applied to arbitrary downstream use-cases. We demonstrate the versatility of our method by applying it to classification and learning-to-rank tasks on two real-world datasets. Our experiments show substantial improvements over the best prior work for this setting.


Holographic Neural Architectures

Representation learning is at the heart of what makes deep learning effective. In this work, we introduce a new framework for representation learning that we call ‘Holographic Neural Architectures’ (HNAs). In the same way that an observer can experience the 3D structure of a holographed object by looking at its hologram from several angles, HNAs derive Holographic Representations from the training set. These representations can then be explored by moving along a continuous bounded single dimension. We show that HNAs can be used to make generative networks, state-of-the-art regression models and that they are inherently highly resistant to noise. Finally, we argue that because of their denoising abilities and their capacity to generalize well from very few examples, models based upon HNAs are particularly well suited for biological applications where training examples are rare or noisy.


Algorithmic Regularization in Learning Deep Homogeneous Models: Layers are Automatically Balanced

We study the implicit regularization imposed by gradient descent for learning multi-layer homogeneous functions including feed-forward fully connected and convolutional deep neural networks with linear, ReLU or Leaky ReLU activation. We rigorously prove that gradient flow (i.e. gradient descent with infinitesimal step size) effectively enforces the differences between squared norms across different layers to remain invariant without any explicit regularization. This result implies that if the weights are initially small, gradient flow automatically balances the magnitudes of all layers. Using a discretization argument, we analyze gradient descent with positive step size for the non-convex low-rank asymmetric matrix factorization problem without any regularization. Inspired by our findings for gradient flow, we prove that gradient descent with step sizes \eta_t = O\left(t^{-\left( \frac12+\delta\right)} \right) (0<\delta\le\frac12) automatically balances two low-rank factors and converges to a bounded global optimum. Furthermore, for rank-1 asymmetric matrix factorization we give a finer analysis showing gradient descent with constant step size converges to the global minimum at a globally linear rate. We believe that the idea of examining the invariance imposed by first order algorithms in learning homogeneous models could serve as a fundamental building block for studying optimization for learning deep models.


Learning Graphs from Data: A Signal Representation Perspective

The construction of a meaningful graph topology plays a crucial role in the effective representation, processing, analysis and visualization of structured data. When a natural choice of the graph is not readily available from the datasets, it is thus desirable to infer or learn a graph topology from the data. In this tutorial overview, we survey solutions to the problem of graph learning, including classical viewpoints from statistics and physics, and more recent approaches that adopt a graph signal processing (GSP) perspective. We further emphasize the conceptual similarities and differences between classical and GSP graph inference methods and highlight the potential advantage of the latter in a number of theoretical and practical scenarios. We conclude with several open issues and challenges that are keys to the design of future signal processing and machine learning algorithms for learning graphs from data.


Relational inductive biases, deep learning, and graph networks
Digging Into Self-Supervised Monocular Depth Estimation
Do the laws of physics prohibit counterfactual communication
The Capacity of Private Information Retrieval with Private Side Information Under Storage Constraints
Graph networks as learnable physics engines for inference and control
Diffeomorphic Learning
Smooth Cyclically Monotone Interpolation and Empirical Center-Outward Distribution Functions
Deep Graphs
Vexillary signed permutations revisited
Limit Theory for Moderate Deviation from Integrated GARCH Processes
Image reconstruction through metamorphosis
Challenges in High-dimensional Reinforcement Learning with Evolution Strategies
Quasi-experimental Shift-share Research Designs
Efficient Genomic Interval Queries Using Augmented Range Trees
Implementing Mediators with Asynchronous Cheap Talk
A basic model of mutations
Stability Analysis for Fast Settling Switched DPLL
Relational inductive bias for physical construction in humans and machines
Face Synthesis for Eyeglass-Robust Face Recognition
Simple probability distributions on a Fock-space lattice
Measuring and avoiding side effects using relative reachability
History Playground: A Tool for Discovering Temporal Trends in Massive Textual Corpora
Deep Continuous Conditional Random Fields with Asymmetric Inter-object Constraints for Online Multi-object Tracking
Online Reciprocal Recommendation with Theoretical Performance Guarantees
Revisiting Singing Voice Detection: a Quantitative Review and the Future Outlook
TD or not TD: Analyzing the Role of Temporal Differencing in Deep Reinforcement Learning
Two novel immunization strategies for epidemic control in directed scale-free networks with nonlinear infectivity
Strict Complementarity in MaxCut SDP
Stability results for martingale representations: the general case
Efficient Online Scalar Annotation with Bounded Support
Risk measures with markets volatility
VADU 2018 Open Problem Session
Bilateral identities of the Rogers-Ramanujan type
Shallow decision-making analysis in General Video Game Playing
Text to brain: predicting the spatial distribution of neuroimaging observations from text reports
Learning from Exemplars and Prototypes in Machine Learning and Psychology
Ring Migration Topology Helps Bypassing Local Optima
Confidence Interval Estimators for MOS Values
Global martingale solutions for a stochastic population cross-diffusion system
Covering with Clubs: Complexity and Approximability
A fairer penalty shootout design in soccer
Modeling Realistic Degradations in Non-blind Deconvolution
groupICA: Independent component analysis for grouped data
Cross-intersecting non-empty uniform subfamilies of hereditary families
Optimal Balancing of Time-Dependent Confounders for Marginal Structural Models
On an extension of the promotion time cure model
Efficient Database Generation for Data-driven Security Assessment of Power Systems
A rational decentralized generalized Nash equilibrium seeking for energy markets
Deep Multi-Structural Shape Analysis: Application to Neuroanatomy
Equitable Coloring and Equitable Choosability of Planar Graphs without chordal 4- and 6-Cycles
Approximation Hierarchies for Copositive Tensor Cone
RedNet: Residual Encoder-Decoder Network for indoor RGB-D Semantic Segmentation
Neural Network-Based Equations for Predicting PGA and PGV in Texas, Oklahoma, and Kansas
On the computational complexity of blind detection of binary linear codes
Normative Modeling of Neuroimaging Data using Scalable Multi-Task Gaussian Processes
Topic Modelling of Empirical Text Corpora: Validity, Reliability, and Reproducibility in Comparison to Semantic Maps
A Desirability-Based Axiomatisation for Coherent Choice Functions
Player preferences in $N$-player games
pammtools: Piece-wise exponential Additive Mixed Modeling tools
Delay Performance of Multi-Antenna Multicasting in Wireless Networks
Differential Diagnosis for Pancreatic Cysts in CT Scans Using Densely-Connected Convolutional Networks
A Cascade of 2.5D CNN and LSTM Network for Mitotic Cell Detection in 4D Microscopy Image
Hierarchical Bi-level Multi-Objective Evolution of Single- and Multi-layer Echo State Network Autoencoders for Data Representations
Dynamically borrowing strength from another study
Synthetic data generation for end-to-end thermal infrared tracking
Meta Learner with Linear Nulling
On the total variation regularized estimator over the branched path graph
State-Space Adaptive Nonlinear Self-Interference Cancellation for Full-Duplex Communication
Distributed Learning from Interactions in Social Networks
A Feynman-Kac type formula for a fixed delay CIR model
Integrality of Subgradients and Biconjugates of Integrally Convex Functions
Time-Fractional User Association in Millimeter Wave MIMO Networks
Efficiency of adaptive importance sampling
User Association in Millimeter Wave MIMO Networks
DNN-HMM based Speaker Adaptive Emotion Recognition using Proposed Epoch and MFCC Features
Robust Optimal Control Using Conditional RiskMappings in Infinite Horizon
Automatic Clustering of a Network Protocol with Weakly-Supervised Clustering
Similarity encoding for learning with dirty categorical variables
Non-Darcian flow or fractional derivative
A geometric attractor mechanism for self-organization of entorhinal grid modules
ALMN: Deep Embedding Learning with Geometrical Virtual Point Generating
Sequential Test for the Lowest Mean: From Thompson to Murphy Sampling
Neural Adversarial Training for Semi-supervised Japanese Predicate-argument Structure Analysis
Observability of the heat equation, geometric constants in control theory, and a conjecture of Luc Miller
Counting walks with large steps in an orthant
Segmentation, Incentives and Privacy
Simultaneous compressive image recovery and deep denoiser learning from undersampled measurements
Mechanism Design without Money for Common Goods
PID2018 Benchmark Challenge:Multi-Objective Stochastic Optimization Algorithm
A Game-Theoretic Approach to Recommendation Systems with Strategic Content Providers
MacroPCA: An all-in-one PCA method allowing for missing values as well as cellwise and rowwise outliers
Asymptotic Refinements of a Misspecification-Robust Bootstrap for Generalized Empirical Likelihood Estimators
Stochastic Gradient/Mirror Descent: Minimax Optimality and Implicit Regularization
Private PAC learning implies finite Littlestone dimension
Frieze Vectors and Unitary Friezes
Lagrange Coded Computing: Optimal Design for Resiliency, Security and Privacy
Program Synthesis from Visual Specification
State-Dependent Interference Channel with Correlated States
Provenance-based Intrusion Detection: Opportunities and Challenges
A Causal Exposure Response Function with Local Adjustment for Confounding
NRTR: A No-Recurrence Sequence-to-Sequence Model For Scene Text Recognition
Automatic catheter detection in pediatric X-ray images using a scale-recurrent network and synthetic data
DRCD: a Chinese Machine Reading Comprehension Dataset
Adversarial confidence and smoothness regularizations for scalable unsupervised discriminative learning
Network Reliability Estimation in Theory and Practice
How Much Are You Willing to Share A ‘Poker-Styled’ Selective Privacy Preserving Framework for Recommender Systems
Self-Normalization Properties of Language Modeling
Magnetoenhancement of superconductivity in composite D-wave superconductors
Bayesian Semantic Instance Segmentation in Open Set World
An unsupervised and customizable misspelling generator for mining noisy health-related text sources
Accurate Building Detection in VHR Remote Sensing Images using Geometric Saliency
A Structure-aware Approach for Efficient Graph Processing
Solving Systems of Quadratic Equations via Exponential-type Gradient Descent Algorithm
Large-scale Land Cover Classification in GaoFen-2 Satellite Imagery
Recent advances and opportunities in scene classification of aerial images with deep models
Conservative Exploration using Interleaving
Soccer on Your Tabletop
Critical Percolation and the Incipient Infinite Cluster on Galton-Watson Trees
Elasticizing Linux via Joint Disaggregation of Memory and Computation
Structural Learning of Multivariate Regression Chain Graphs via Decomposition
Measuring Influence on Instagram: a Network-oblivious Approach
Disconnected Manifold Learning for Generative Adversarial Networks
Multi-Agent Reinforcement Learning via Double Averaging Primal-Dual Optimization
Deploying Customized Data Representation and Approximate Computing in Machine Learning Applications
Patch-Based Image Hallucination for Super Resolution with Detail Reconstruction from Similar Sample Images
On Construction of Upper and Lower Bounds for the HOMO-LUMO Spectral Gap
A Comprehensive Comparison between Neural Style Transfer and Universal Style Transfer
On the Flip Side: Identifying Counterexamples in Visual Question Answering
On the Importance of Attention in Meta-Learning for Few-Shot Text Classification
An Aggressive Genetic Programming Approach for Searching Neural Network Structure Under Computational Constraints
On estimation for Brownian motion governed by telegraph process with multiple off states
A bivariate generating function for zeta values and related supercongruences
TernausNetV2: Fully Convolutional Network for Instance Segmentation
Latent Tree Learning with Differentiable Parsers: Shift-Reduce Parsing and Chart Parsing

Advertisements