Inductive biases, pretraining and fine-tuning jointly account for brain responses to speech

Our ability to comprehend speech remains, to date, unrivaled by deep learning models. This feat could result from the brain’s ability to fine-tune generic sound representations for speech-specific processes. To test this hypothesis, we compare i) five types of deep neural networks to ii) human brain responses elicited by spoken sentences and recorded in 102 Dutch subjects using functional Magnetic Resonance Imaging (fMRI). Each network was either trained on an acoustics scene classification, a speech-to-text task (based on Bengali, English, or Dutch), or not trained. The similarity between each model and the brain is assessed by correlating their respective activations after an optimal linear projection. The differences in brain-similarity across networks revealed three main results. First, speech representations in the brain can be accounted for by random deep networks. Second, learning to classify acoustic scenes leads deep nets to increase their brain similarity. Third, learning to process phonetically-related speech inputs (i.e., Dutch vs English) leads deep nets to reach higher levels of brain-similarity than learning to process phonetically-distant speech inputs (i.e. Dutch vs Bengali). Together, these results suggest that the human brain fine-tunes its heavily-trained auditory hierarchy to learn to process speech.

Locally Recoverable Codes correcting many erasures over small fields

We define linear codes which are s-Locally Recoverable Codes (or s-LRC), i.e. codes which are LRC in s ways, the case $$s=1$$ s = 1 roughly corresponding to the classical case of LRC codes. We use them to describe codes which correct many erasures, although they have small minimum distance. Any letter of a received word may be corrected using s different local codes. We use the Segre embedding of s local codes and then a linear projection.

Modelling of methoprene concentration at storm sewer outfall

The objective of this report is to compare the predicted concentration of methoprene, a larvicide used in the City of Toronto to control the widespread of West Nile Virus by suppressing mosquito growth, at storm sewer outfall during a typical year rainfalls (1980 rainfalls) with the Interim Provincial Water Quality Objectives (IPWQO). The methoprene that is under investigation in this report is in form of ingot. Extending from an existing spreadsheet-based model that simulates the methoprene concentration within two monitored catch basins in the Newtonbrook sewershed of North York, methoprene concentration at the sewer outfall during the 1980 rainfalls is predicted by linear projection upon calibration of the model with the methoprene mass at the outfall measured in year 2005. Results show that predicted methoprene concentration at outfall exceeds the IPWQO in six days out of the one-hundred-day period. It is recommended that to better mimic the actual situation, traveling time effect and sensitivity analysis on catch basin sump volume be included in future study.

Modelling of methoprene concentration at storm sewer outfall

The objective of this report is to compare the predicted concentration of methoprene, a larvicide used in the City of Toronto to control the widespread of West Nile Virus by suppressing mosquito growth, at storm sewer outfall during a typical year rainfalls (1980 rainfalls) with the Interim Provincial Water Quality Objectives (IPWQO). The methoprene that is under investigation in this report is in form of ingot. Extending from an existing spreadsheet-based model that simulates the methoprene concentration within two monitored catch basins in the Newtonbrook sewershed of North York, methoprene concentration at the sewer outfall during the 1980 rainfalls is predicted by linear projection upon calibration of the model with the methoprene mass at the outfall measured in year 2005. Results show that predicted methoprene concentration at outfall exceeds the IPWQO in six days out of the one-hundred-day period. It is recommended that to better mimic the actual situation, traveling time effect and sensitivity analysis on catch basin sump volume be included in future study.

Interference Identification for Time-Varying Polyhedra

Identification of when and where moving areas intersect is an important problem in maritime operations and air traffic control. This problem can become particularly complicated when considering large numbers of objects, and when taking into account the curvature of the earth. In this paper, we present an approach to conflict identification as a series of stages where the earlier stages are fast, but may result in a false detection of a conflict. These early stages are used to reduce the number of potential conflict pairs for the later stages, which are slower, but more precise. The stages use R-trees, polygon intersection, linear projection and nonlinear programming. Our approach is generally applicable to objects moving in piece-wise straight lines on a 2D plane, and we present a specific case where the Mercator Projection is used to transform objects moving along rhumb lines on the earth into straight lines to fit in our approach. We present several examples to demonstrate our methods, as well as to quantify the empirical time complexity by using randomly generated areas.

Interference Identification for Time-Varying Polyhedra

Identification of when and where moving areas intersect is an important problem in maritime operations and air traffic control. This problem can become particularly complicated when considering large numbers of objects, and when taking into account the curvature of the earth. In this paper, we present an approach to conflict identification as a series of stages where the earlier stages are fast, but may result in a false detection of a conflict. These early stages are used to reduce the number of potential conflict pairs for the later stages, which are slower, but more precise. The stages use R-trees, polygon intersection, linear projection and nonlinear programming. Our approach is generally applicable to objects moving in piece-wise straight lines on a 2D plane, and we present a specific case where the Mercator Projection is used to transform objects moving along rhumb lines on the earth into straight lines to fit in our approach. We present several examples to demonstrate our methods, as well as to quantify the empirical time complexity by using randomly generated areas.

Geometry on plane and optimal linear prognosis

In the article we consider a new scalar production in the linear subspace of n vectors (by analogy with the geometry of Lobachevsky, all two vectors are orthogonal as projections of the third vector). With help of the production we obtain a new form of linear projection of the additional vector to the subspace. The rapid algorithm of finding of optimum linear estimation of additional vector on the known vectors ensues from the projection.

Tibetan speech synthesis based on an improved neural network

Nowadays, Tibetan speech synthesis based on neural network has become the mainstream synthesis method. Among them, the griffin-lim vocoder is widely used in Tibetan speech synthesis because of its relatively simple synthesis.Aiming at the problem of low fidelity of griffin-lim vocoder, this paper uses WaveNet vocoder instead of griffin-lim for Tibetan speech synthesis.This paper first uses convolution operation and attention mechanism to extract sequence features.And then uses linear projection and feature amplification module to predict mel spectrogram.Finally, use WaveNet vocoder to synthesize speech waveform. Experimental data shows that our model has a better performance in Tibetan speech synthesis.