Hi Everyone, In this article, we understand the paper "The Dual Form of Neural Networks Revisited: Connecting Test Time Predictions to Training Patterns via Spotlights of Attention." Neural networks are widely used for various tasks. Still, their exact working of them is not fully understood. Multiple ways are proposed to understand the trained network, such … Continue reading The Dual Form of Neural Networks Revisited: Connecting Test Time Predictions to Training Patterns via Spotlights of Attention
Author: Achuth Rao MV
An interactive tutorial on Normalizing flows
Refer this notebook.
Probabilistic interpretation of Itakura Saito distance
According to Wikipedia, The Itakura–Saito distance (or Itakura–Saito divergence) (IS-distance) is a measure of the difference between an original spectrum and an approximation of that spectrum. It was proposed by Fumitada Itakura and Shuzo Saito in the 1960s while they were with NTT.[1] This is the distance used in several spectral estimation methods. Especially in speech … Continue reading Probabilistic interpretation of Itakura Saito distance
How variational auto encoder can theoretically model any distribution
Similar ideas are applicable in case of Normalizing flows, You can find more details in another tutorial. Variational Auto Encoder (VAE) is one of the famous generative models and applied for various unsupervised, semi-supervised tasks [1][2]. There is a lot of interesting ways to look at how it works [3]. But in this article, we … Continue reading How variational auto encoder can theoretically model any distribution
Why neural networks are called Universal Approximators- Graphical proof
Hey Guys, We are reading lot about AI/Deep learning to do some really smart stuff such as playing G0 game, recognizing images and speech. The core part of the Deep learning is the deep neural network (DNN). Before we understand DNN, we need look at the shallow neural network and its property called the universal … Continue reading Why neural networks are called Universal Approximators- Graphical proof
MEL-GENERALIZED CEPSTRAL ANALYSIS — A UNIFIED APPROACH TO SPEECH SPECTRAL ESTIMATION
Introduction: As per the source-filter model,we can see our speech production system as excitation at the glottis passing through a acoustic filter(vocal tract). As you see the below figure the excitation is approximately a impulse train(voiced speech) and the filter has smooth frequency response depending on the what phoneme is been spoken(vocal tract shape). In … Continue reading MEL-GENERALIZED CEPSTRAL ANALYSIS — A UNIFIED APPROACH TO SPEECH SPECTRAL ESTIMATION
What is the intuition behind the square wave spectrum(Fourier series)
The spectrum of the square wave contains just the nonzero values only at the odd multiple of its frequency(0 ,f0 ,3*f0 ).The intuition behind it is the spectrum value/Fourier series coefficient at given frequency is the measure of correlation(multiply and add) between the signal and sinusoidal signal of given frequency.Consider and square wave(anti-symmetric around zeros) of some … Continue reading What is the intuition behind the square wave spectrum(Fourier series)
What is the geometric meaning behind the positive definite matrix ?
The definition of the positive definite(PD) of the matrix(H) is v’*H*v>0 for all non zero vector. The inner product between the vector is proportional to cosine of the angle between the vectors. If we consider v’*(H*v)>0 means the angle between the vector v and H*v(transformed vector) is acute.So positive definite matrix is not going to transforms … Continue reading What is the geometric meaning behind the positive definite matrix ?
Welcome
Hello Everyone, This blog is share my understanding of the speech technology(like speech to text,text to speech,speech enhancement) and machine learning(Gaussian mixture models,Hidden Markov Models and Neural networks).Please feel free to post the comments if you have any suggestions. -Achuth
Achuth Rao M V 