arXiv research

Proposes a deep neural network for early disk drive failure prediction.

Paper proposes a method to predict disk failures using multi-layer domain adaptive learning.

RODMAN improves ML-based disk failure prediction accuracy in cloud environments.

This paper addresses the problem of evaluating learning systems in safety critical domains such as autonomous driving, where failures can have catastrophic consequences. We focus on two problems: searching for scenarios when learned agents fail and assessing their probability of failure. The standard method for agent e…

New approach uses dynamic programming to efficiently discover failures in autonomous vehicle simulations.

Researchers develop PAIN to improve self-driving safety through adversarial training.

While machine learning systems show high success rate in many complex tasks, research shows they can also fail in very unexpected situations. Rise of machine learning products in safety-critical industries cause an increase in attention in evaluating model robustness and estimating failure probability in machine learni…

This work improves safety validation of autonomous vehicles by finding interpretable failures.

Bayesian inference models failure distributions in autonomous systems.

With the development of cloud computing and big data, the reliability of data storage systems becomes increasingly important. Previous researchers have shown that machine learning algorithms based on SMART attributes are effective methods to predict hard drive failures. In this paper, we use SMART attributes to predict…

GE finds failures in autonomous systems without domain heuristics.

While autonomous vehicle (AV) technology has shown substantial progress, we still lack tools for rigorous and scalable testing. Real-world testing, the $\textit{de-facto}$ evaluation method, is dangerous to the public. Moreover, due to the rare nature of failures, billions of miles of driving are needed to statisticall…

CARL safely adapts RL agents for safety-critical tasks.

In this paper, we present a framework to control a self-driving car by fusing raw information from RGB images and depth maps. A deep neural network architecture is used for mapping the vision and depth information, respectively, to steering commands. This fusion of information from two sensor sources allows to provide …

To improve efficiency and reduce failures in autonomous vehicles, research has focused on developing robust and safe learning methods that take into account disturbances in the environment. Existing literature in robust reinforcement learning poses the learning problem as a two player game between the autonomous system…

Develops methods to improve reliability of deep learning for autonomous driving.

With emerging smart communities, improving overall system availability is becoming a major concern. In order to improve the reliability of the components in a system we propose an inference model to predict Remaining Useful Life (RUL) of those components. In this paper we work with components of backend data servers su…

Convolutional neural networks are commonly used to control the steering angle for autonomous cars. Most of the time, multiple long range cameras are used to generate lateral failure cases. In this paper we present a novel model to generate this data and label augmentation using only one short range fisheye camera. We p…

Sensor fusion is a key technology that integrates various sensory inputs to allow for robust decision making in many applications such as autonomous driving and robot control. Deep neural networks have been adopted for sensor fusion in a body of recent studies. Among these, the so-called netgated architecture was propo…

We show how to measure the failure of the Whitney trick in dimension 4 by constructing higher- order intersection invariants of Whitney towers built from iterated Whitney disks on immersed surfaces in 4-manifolds. For Whitney towers on immersed disks in the 4-ball, we identify some of these new invariants with previous…

During the development of autonomous systems such as driverless cars, it is important to characterize the scenarios that are most likely to result in failure. Adaptive Stress Testing (AST) provides a way to search for the most-likely failure scenario as a Markov decision process (MDP). Our previous work used a deep rei…

Feedback alignment methods need to be evaluated for accuracy and gradient cosine similarity.

Many open problems and important theorems in low-dimensional topology have been formulated as statements about certain 2--complexes called gropes. This paper describes a precise correspondence between embedded gropes in 4--manifolds and the failure of the Whitney move in terms of iterated `towers' of Whitney disks. The…

The conformal Laplacian's algebraic structure is explored in 2D, revealing a central charge.

Failing to distinguish between a sheepdog and a skyscraper should be worse and penalized more than failing to distinguish between a sheepdog and a poodle; after all, sheepdogs and poodles are both breeds of dogs. However, existing metrics of failure (so-called "loss" or "win") used in textual or visual classification/r…

New insights explain why $β$-VAEs fail at disentanglement.

D3M debiases models by selectively removing problematic examples.

VerifAI toolkit improves neural network-based aircraft taxiing system safety.

New framework assesses LLM security risks in BFSI.

The minimum error entropy (MEE) criterion has been verified as a powerful approach for non-Gaussian signal processing and robust machine learning. However, the implementation of MEE on robust classification is rather a vacancy in the literature. The original MEE only focuses on minimizing the Renyi's quadratic entropy …

Given a genus-$g$ Heegaard splitting of the $3$-sphere with $g \ge 3$, we show that the primitive disk complex for the splitting is not weakly closed under disk surgery operation. That is, there exist two primitive disks in one of the handlebodies of the splitting such that any disk surgery on one along the other one y…

Note on connectedness of primitive disk complex.

Let $K$ be an unknot in $8$-bridge position in the $3$-sphere. We give an example of a pair of weak reducing disks $D_1$ and $D_2$ for $K$ such that both disks obtained from $D_i$ ($i = 1, 2$) by a surgery along any outermost disk in $D_{3-i}$, cut off by an outermost arc of $D_i \cap D_{3-i}$ in $D_{3-i}$, are not wea…

Machine and reinforcement learning (RL) are increasingly being applied to plan and control the behavior of autonomous systems interacting with the physical world. Examples include self-driving vehicles, distributed sensor networks, and agile robots. However, when machine learning is to be applied in these new settings,…

New knots bound multiple non-isotopic ribbon disks.

We study cobordisms and cobordisms rel boundary of PL locally-flat disk knots $D^{n-2}\into D^n$. Cobordisms of disk knots that do not fix the boundary sphere knots are easily classified by the cobordism properties of these boundaries, and any two even-dimensional disk knots with isotopic boundary knots are cobordant r…

Khovanov homology fails to differentiate certain slice disks.

For a genus two Heegaard splitting of a lens space, the primitive disk complex is defined to be the full subcomplex of the disk complex for one of the handlebodies of the splitting spanned by all vertices of primitive disks. In this work, we describe the complete combinatorial structure of the primitive disk complex fo…

We study the classification of slice disks of knots up to isotopy and diffeomorphism using an invariant in knot Floer homology. We compute the invariant of a slice disk obtained by deform-spinning, and show that it can be effectively used to distinguish non-isotopic slice disks with diffeomorphic complements. Given a s…

In recent years, deep learning methods have outperformed other methods in image recognition. This has fostered imagination of potential application of deep learning technology including safety relevant applications like the interpretation of medical images or autonomous driving. The passage from assistance of a human d…

A criterion for Whitney disks connects intersections in 3-manifold homology.

New disks found with similar outer shapes.

We construct an infinite family of slice disks with the same exterior, which gives an affirmative answer to an old question asked by Hitt and Sumners in 1981. Furthermore, we prove that these slice disks are ribbon disks.

Study constructs disks with curved boundaries in a 3D ball.

Economics tool predicts failure times in reliability systems.

New spanning 3-disks found for unlink in 4-sphere.

Even-dimensional simply connected manifolds that are rational homology spheres and double disk bundles are homeomorphic to spheres.

We study the relationship between fibered ribbon 1-knots and fibered ribbon 2-knots by studying fibered slice disks with handlebody fibers. We give a characterization of fibered homotopy-ribbon disks and give analogues of the Stallings twist for fibered disks and 2-knots. As an application, we produce infinite families…