SearcharxivSearch

arXiv subjects

Andrei Petrovskii

Publications and source records attributed to Andrei Petrovskii.

5 recordsLinked to original sources

CNN+LSTM Architecture for Speech Emotion Recognition with Data Augmentation

In this work we design a neural network for recognizing emotions in speech, using the IEMOCAP dataset. Following the latest advances in audio analysis, we use an architecture involving both convolutional layers, for extracting high-level features from raw spectrograms, and recurrent ones for aggregating long-term dependencies. We examine the techniques of data augmentation with vocal track length perturbation, layer-wise optimizer adjustment, batch normalization of recurrent layers and obtain highly competitive results of 64.5% for weighted accuracy and 61.7% for unweighted accuracy on four emotions.

cs.SD

Diagonal form factors and hexagon form factors

We study the heavy-heavy-light (HHL) three-point functions in the planar N = 4 super-Yang- Mills theory using the recently proposed hexagon bootstrap program [arXiv:1505.06745]. We prove the conjecture of Bajnok, Janik and Wereszczynski [arXiv:1404.4556] on the polynomial L-dependence of HHL structure constant up to the leading finite-size corrections, where L is the length of the heavy operators. The proof is presented for a specific set-up but the method can be applied to more general situations.

hep-th

Diagonal Form Factors and Heavy-Heavy-Light Three-Point Functions at Weak Coupling

In this paper we consider a special kind of three-point functions of HHL type at weak coupling in N=4 SYM theory and analyze its volume dependence. At strong coupling this kind of three-point functions were studied recently by Bajnok, Janik and Wereszczynski [1]. The authors considered some cases of HHL correlator in the su(2) sector and, relying on their explicit results, formulated a conjecture about the form of the volume dependence of the symmetric HHL structure constant to be valid at any coupling up to wrapping corrections. In order to test this hypothesis we considered the HHL correlator in su(2) sector at weak coupling and directly showed that, up to one loop, the finite volume dependence has exactly the form proposed in [1]. Another side of the conjecture suggests that computation of the symmetric structure constant is equivalent to computing the corresponding set of infinite volume form factors, which can be extracted as the coefficients of finite volume expansion. In this sense, extracting appropriate coefficients from our result gives a prediction for the corresponding infinite volume form factors.

hep-th

From Spin Vertex to String Vertex

In the recent publication arXiv:1410.8860 the spin vertex was introduced as a new approach for computing three-point functions in N = 4 SYM. In this note we consider the BMN limit of the spin vertex for scalar excitations and show that it reproduces the string vertex in the light-cone string field theory which describes the string interactions in the pp-wave background at the leading order of BMN expansion. This is achieved by introducing a polynomial representation for the spin vertex. We derive the Neumann coefficients from the spin vertex at weak coupling and show they match with the Neumann coefficients from the string field theory.

hep-th

String Bits and the Spin Vertex

We initiate a novel formalism for computing correlation functions of trace operators in the planar N=4 SYM theory. The central object in our formalism is the spin vertex, which is the weak coupling analogy of the string vertex in string field theory. We construct the spin vertex explicitly for all sectors at the leading order using a set of bosonic and fermionic oscillators. We prove that the vertex has trivial monodromy, or put in other words, it is a Yangian invariant. Since the monodromy of the vertex is the product of the monodromies of the three states, the Yangian invariance of the vertex implies an infinite exact symmetry for the three-point function. We conjecture that this infinite symmetry can be lifted to any loop order.

hep-th