More Like this

1. Role of Regularization in the Prediction of Valence from Speech

   https://doi.org/10.21437/Interspeech.2018-2508  

   Sridhar, Kusha; Parthasarathy, Srinivas; Busso, Carlos (September 2018, Interspeech 2018)

   Regularization plays a key role in improving the prediction of emotions using attributes such as arousal, valence and dominance. Regularization is particularly important with deep neural networks (DNNs), which have millions of parameters. While previous studies have reported competitive performance for arousal and dominance, the prediction results for valence using acoustic features are significantly lower. We hypothesize that higher regularization can lead to better results for valence. This study focuses on exploring the role of dropout as a form of regularization for valence, suggesting the need for higher regularization. We analyze the performance of regression models for valence, arousal and dominance as a function of the dropout probability. We observe that the optimum dropout rates are consistent for arousal and dominance. However, the optimum dropout rate for valence is higher. To understand the need for higher regularization for valence, we perform an empirical analysis to explore the nature of emotional cues conveyed in speech. We compare regression models with speakerdependent and speaker-independent partitions for training and testing. The experimental evaluation suggests stronger speaker dependent traits for valence. We conclude that higher regularization is needed for valence to force the network to learn global patterns that generalize across speakers. 

   more » « less
2. Effects of ferromagnetic branch structures on coercivity of alnico permanent magnets

   https://doi.org/10.1063/9.0000298  

   Won, Hoyun; Hong, Yang-Ki; Choi, Minyeong; Mankey, Gary J.; Lee, Jongkook; Lee, Taegyu; Lim, Tae Won (March 2022, AIP Advances)

   The effects of various branches geometry and dimensions such as length, thickness, and width for H-, U-, O-, YH- and YU-shaped alnico structures on coercivity ( H ci ) using micromagnetic simulation for coherent rotation and curling modes are investigated. The simulation results suggest that the H-shaped structure needs long and short branch length for the coherent rotation and curling, respectively, regardless of branch thickness and width to realize high H ci . Short branch length with thin thickness and short width are recommended for both rotations for the U- and O-shaped structures. Lastly, both Y-shaped structures need branch with long length, thin thickness, and mid-long width for the coherent rotation, but short width for the YH-shaped and mid-long width for the YU-shaped regardless of length and thickness for curling are desired. Furthermore, among the five studied structures, H- or YH-shaped structure for coherent rotation and O-shaped structure for curling are highly recommended for fabrication to realize a high H ci . 

   more » « less
3. Faster Linear Algebra for Distance Matrices

   Indyk, Piotr; Silwal, Sandeep (January 2022, Conference on Neural Information Processing Systems)

   The distance matrix of a dataset X of n points with respect to a distance function f represents all pairwise distances between points in X induced by f. Due to their wide applicability, distance matrices and related families of matrices have been the focus of many recent algorithmic works. We continue this line of research and take a broad view of algorithm design for distance matrices with the goal of designing fast algorithms, which are specifically tailored for distance matrices, for fundamental linear algebraic primitives. Our results include efficient algorithms for computing matrix-vector products for a wide class of distance matrices, such as the l1 metric for which we get a linear runtime, as well as a quadratic lower bound for any algorithm which computes a matrix-vector product for the l_infty case. Our upper bound results have many further downstream applications, including the fastest algorithm for computing a relative error low-rank approximation for the distance matrix induced by l1 and l2 functions and the fastest algorithm for computing an additive error lowrank approximation for the l2 metric, in addition to applications for fast matrix multiplication among others. We also give algorithms for constructing distance matrices and show that one can construct an approximate l2 distance matrix in time faster than the bound implied by the Johnson-Lindenstrauss lemma. 

   more » « less
4. Excess mortality in Belarus during the COVID-19 pandemic as the case study of a country with limited non-pharmaceutical interventions and limited reporting

   https://doi.org/10.1038/s41598-022-09345-z  

   Kirpich, Alexander; Shishkin, Aleksandr; Weppelmann, Thomas A.; Tchernov, Alexander Perez; Skums, Pavel; Gankin, Yuriy (December 2022, Scientific Reports)

   Abstract Public health intervention to contain the ongoing COVID-19 pandemic significantly differed by country since the SARS-CoV-2 spread varied regionally in time and in scale. Since vaccinations were not available until the end of 2020 non-pharmaceutical interventions (NPIs) remained the only strategies to mitigate the pandemic spread at that time. Belarus in Europe is one of a few countries with a high Human Development Index where no lockdowns have ever been implemented and only limited NPIs have taken place for a period of time. Therefore, the Belarusian case was evaluated and compared in terms of the mortality burden. Since the COVID-19 mortality was low, the excess overall mortality was studied for Belarus. Since no overall mortality data have been reported past June 2020 the analysis was complemented by the study of Google Trends funeral-related search queries up until August 2021. Depending on the model, the Belarusian mortality for June of 2020 was 29 to 39% higher than otherwise expected with the corresponding estimated excess death was from 2953 to 3690 while the reported COVID-19 mortality for June 2020 was only 157 cases. The Belarusian excess mortality for June 2020 was higher than for all neighboring countries with an excess of 5% for Poland, 5% for Ukraine, 8% for Russia, 11% for Lithuania and 11% for Latvia. The relationship between Google Trends and mortality time series was studied using Granger’s test and the results were statistically significant. The results for Google Trends searches did vary by key phrase with the largest excess of 138% for April 2020 and 148% for September 2020 was observed for a key phrase “coffin”, while the largest excess of 218% for January 2021 was observed for “funeral services”. In summary, there are indications of the excess overall mortality in Belarus, which is larger than the reported COVID-19-related mortality. 

   more » « less
5. Estimating Shapley Values of Training Utterances for Automatic Speech Recognition Models

   https://doi.org/10.1109/ICASSP49357.2023.10097237  

   Raza Syed, Ali; Mandel, Michael I. (June 2023, IEEE International Conference on Acoustics Speech and Signal Processing)

   Data Valuation in machine learning is concerned with quantifying the relative contribution of a training example to a model’s performance. Quantifying the importance of training examples is useful for identifying high and low quality data to curate training datasets and for address data quality issues. Shapley values have gained traction in machine learning for curating training data and identifying data quality issues. While computing the Shapley values of training examples is computationally prohibitive, approximation methods have been used successfully for classification models in computer vision tasks. We investigate data valuation for Automatic Speech Recognition models which perform a structured prediction task and propose a method for estimating Shapley values for these models. We show that a proxy model can be learned for the acoustic model component of an end-to-end ASR and used to estimate Shapley values for acoustic frames. We present a method for using the proxy acoustic model to estimate Shapley values for variable length utterances and demonstrate that the Shapley values provide a signal of example quality. 

   more » « less