Revisiting Additive Quantization
نویسندگان
چکیده
We revisit Additive Quantization (AQ), an approach to vector quantization that uses multiple, full-dimensional, and non-orthogonal codebooks. Despite its elegant and simple formulation, AQ has failed to achieve state-of-the-art performance on standard retrieval benchmarks, because the encoding problem, which amounts to MAP inference in multiple fully-connected Markov Random Fields (MRFs), has proven to be hard to solve. We demonstrate that the performance of AQ can be improved to surpass the state of the art by leveraging iterated local search, a stochastic local search approach known to work well for a range of NP-hard combinatorial problems. We further show a direct application of our approach to a recent formulation of vector quantization that enforces sparsity of the codebooks. Unlike previous work, which required specialized optimization techniques, our formulation can be plugged directly into state-of-the-art lasso optimizers. This results in a conceptually simple, easily implemented method that outperforms the previous state of the art in solving sparse vector quantization. Our implementation is publicly available.
منابع مشابه
Revisiting the Tangent Groupoid
The setup of a quantization in a Tangent Groupoid structure is reviewed from the point of view of the action of Butcher Group. In this sense, quantization is implied by renormalization.
متن کاملVector Quantization of Speech with Noise Cancellation
This thesis is an investigation of robust vector quantization, with the purpose of providing a system for the application of data compression and speech enhancement. Vector quantization is widely used in data compression systems. However, the performance of these systems will degrade in a noisy environment. The proposed robust vector quantization system solves the problem of optimal quantizatio...
متن کاملStatistical Theory of Quantization - Instrumentation and Measurement, IEEE Transactions on
The effect of uniform quantization can often be modeled by an additive noise that is uniformly distributed, uncorrelated with the input signal, and has a white spectrum. This paper surveys the theory behind this model, and discusses the conditions of its validity. The applicationi of the model to floating-point quantization is demonstrated.
متن کاملStatistical Theory of Quantization
1 Statistical Theory of Quantization Bernard Widrow, Life Fellow, IEEE , Istv an Koll ar, Senoir Member, IEEE , and Ming-Chang Liu Abstract |The e ect of uniform quantization can often be modeled by an additive noise that is uniformly distributed, uncorrelated with the input signal, and has a white spectrum. This paper surveys the theory behind this model, and discusses the conditions of its va...
متن کاملWatermark Detection after Quantization Attacks
The embedding of additive noise sequences is often used to hide information in digital audio, image or video documents. However, the embedded information might be impaired by involuntary or malicious “attacks.” This paper shows that quantization attacks cannot be described appropriately by an additive white Gaussian noise (AWGN) channel. The robustness of additive watermarks against quantizatio...
متن کامل