arXiv research

Edge language models show bias over time, especially on resource-constrained devices.

xYOLO speeds up object detection on low-end hardware for humanoid soccer robots.

We present a machine learning based method for noise classification using a low-power and inexpensive IoT unit. We use Mel-frequency cepstral coefficients for audio feature extraction and supervised classification algorithms (that is, support vector machine and k-nearest neighbors) for noise classification. We evaluate…

Anomalies in the ambient magnetic field can be used as features in indoor positioning and navigation. By using Maxwell's equations, we derive and present a Bayesian non-parametric probabilistic modeling approach for interpolation and extrapolation of the magnetic field. We model the magnetic field components jointly by…

Human computer interaction facilitates intelligent communication between humans and computers, in which gesture recognition plays a prominent role. This paper proposes a machine learning system to identify dynamic gestures using tri-axial acceleration data acquired from two public datasets. These datasets, uWave and So…

Paper proposes scalable multi-label classification for edge devices using CNN.

Federated learning (FL) allows model training from local data collected by edge/mobile devices while preserving data privacy, which has wide applicability to image and vision applications. A challenge is that client devices in FL usually have much more limited computation and communication resources compared to servers…

CodeReef enables sharing ML models across platforms efficiently.

We implement a differentiable Neural Architecture Search (NAS) method inspired by FBNet for discovering neural networks that are heavily optimized for a particular target device. The FBNet NAS method discovers a neural network from a given search space by optimizing over a loss function which accounts for accuracy and …

Bayesian active learning improves stress and affect detection on wearable devices.

Decision-tree-based ensemble classification methods (DTEMs) are a prevalent tool for supervised anomaly detection. However, due to the continued growth of datasets, DTEMs result in increasing drawbacks such as growing memory footprints, longer training times, and slower classification latencies at lower throughput. In …

Paper proves PI consensus algorithm converges exponentially under restricted secant inequality.

We show that the triangle with angles Pi/12, Pi/3 and 7*Pi/12 has the lattice property and compute this triangle's Veech group.

This paper introduces a novel low-cost device prototype for the automatic diagnosis of diseases, utilizing inputted symptoms and personal background. The engineering goal is to solve the problem of limited healthcare access with a single device. Diagnosing diseases automatically is an immense challenge, owing to their …

By seeking the narrowest prediction intervals (PIs) that satisfy the specified coverage probability requirements, the recently proposed quality-based PI learning principle can extract high-quality PIs that better summarize the predictive certainty in regression tasks, and has been widely applied to solve many practical…

SBI provides more accurate pole positions than chi-squared minimization in model misspecification.

New research suggests privileged information doesn't improve model performance.

Every torus knot can be represented as a Fourier-(1,1,2) knot which is the simplest possible Fourier representation for such a knot. This answers a question of Kauffman and confirms the conjecture made by Boocher, Daigle, Hoste and Zheng. In particular, the torus knot T(p,q) can be parameterized as x(t)=cos(pt), y(t)=c…

This study evaluates methods for constructing prediction intervals with neural networks.

This paper improves SVM prediction uncertainty quantification methods.

Proposes a new cost function for neural networks to improve prediction interval quality.

CryptoNAS improves PI accuracy by 3.4% with 2.4x less latency.

Let K be a knot of genus g. If K is fibered, then it is well known that the knot group pi(K) splits only over a free group of rank 2g. We show that if K is not fibered, then pi(K) splits over non-free groups of arbitrarily large rank. Furthermore, if K is not fibered, then pi(K) splits over every free group of rank at …

Lower bounds for PI on multi-action MDPs are established, showing complexity grows with action count.

A K(pi,1)-foliation is one for which the universal covers of all leaves are contractible (thus all leaves are K(pi,1)'s for some pi). In the first part of the paper we show that the tangential Lusternik--Schnirelmann category cat F of a K(pi,1)-foliation F on a manifold M is bounded from below by t-codim F for any t wi…

Activity recognition is the ability to identify and recognize the action or goals of the agent. The agent can be any object or entity that performs action that has end goals. The agents can be a single agent performing the action or group of agents performing the actions or having some interaction. Human activity recog…

New Ricci curvature means derived from plane curvatures.

This paper presents a method to automatically generate high-quality prediction intervals for neural networks.

Gromov and Lawson conjectured that a closed spin manifold M of dimension n with fundamental group pi admits a metric with positive scalar curvature if and only if an associated element in KO_n(B pi) vanishes. In this note we present counter examples to the `if' part of this conjecture for groups pi which are torsion fr…

Sigma-Pi-Sigma neural networks (SPSNNs) as a kind of high-order neural networks can provide more powerful mapping capability than the traditional feedforward neural networks (Sigma-Sigma neural networks). In the existing literature, in order to reduce the number of the Pi nodes in the Pi layer, a special multinomial P_…

Proposes a method to generate prediction intervals using weighted asymmetric loss functions.

PI-SAC agents learn predictive information to improve RL efficiency.

In this article we construct a minimal symplectic 4-manifold R that has small Euler characteristic (e(R)=8) and two essential Lagrangian tori with nice properties. These properties make R particularly suitable for constructing interesting examples of symplectic manifolds with small Euler characteristic. In particular, …

pi-VAE models neural activity with interpretable latent variables.

Given a closed manifold N and a self-indexing Morse function f: N --> R with up to four distinct Morse indices, we construct a symplectic Lefschetz fibration pi: E --> C which models the complexification of f on the disk cotangent bundle, f_C : D(T*N) --> C, when f is real analytic. By construction, pi: E --> C comes w…

Paper proposes a method to create smaller, more efficient deep generative audio models.

Adaptive PI by reweighting nonconformity scores improves model uncertainty reflection.

Let pi be a free group of rank 2. Its outer automorphism group Out(pi) acts on the space of equivalence classes of representations in Hom(pi, SL(2,C)). Let SLm(2,R) denote ths subset of GL(2,R) consisting of matrices of determinant -1 and let ISL(2,R) denote the subgroup (SL(2,R) union i SLm(2,R)) of SL(2,C). The repre…

Proposes PI-VAE for solving SDEs with limited measurements.

New algorithms explain Naive Bayes classifiers in polynomial time and delay.

A framework to explain decoder-only sequence classification models using intermediate predictions.

Infinite rank groups found in 3-manifolds with infinite fundamental groups.

This paper considers the generation of prediction intervals (PIs) by neural networks for quantifying uncertainty in regression tasks. It is axiomatic that high-quality PIs should be as narrow as possible, whilst capturing a specified portion of data. We derive a loss function directly from this axiom that requires no d…

Given k>=2, we construct a (2k-2)-parameter family of properly embedded minimal surfaces in H^2 x R invariant by a vertical translation T, called Saddle Towers, which have total intrinsic curvature 4 pi(1-k), genus zero and 2k vertical Scherk-type ends in the quotient by T. As limits of those Saddle Towers, we obtain J…

We show that many 3-manifold groups have no nonabelian surface subgroups. For example, any link of an isolated complex surface singularity has this property. In fact, we determine the exact class of closed graph-manifolds which have no immersed pi_1-injective surface of negative Euler characteristic. We also determine …

Let S be a closed, oriented surface of genus at least 2, and consider the extension 1 -> pi_1 S -> MCG(S,p) -> MCG(S) -> 1, where MCG(S) is the mapping class group of S, and MCG(S,p) is the mapping class group of S punctured at p. We prove that any quasi-isometry of MCG(S,p) which coarsely respects the cosets of the no…

Let Pi: M -> B be an onto maximal rank map or a Riemannian submersion between Riemannian manifolds M and B. Initially, we prove necessary and sufficient conditions for any fiber F to be roughly isometric to M. Then, we prove necessary and sufficient conditions for Pi to be a rough isometry. As a corollary M is roughly …

PIF detects anomalies in structured patterns using preference embedding.