Different Levels Of Quantization Research Articles

Anatomy of Deep Learning Image Classification and Object Detection on Commercial Edge Devices: A Case Study on Face Mask Detection

Developing efficient on-the-edge Deep Learning (DL) applications is a challenging and non-trivial task, as first different DL models need to be explored with different trade-offs between accuracy and complexity, second, various optimization options, frameworks and libraries are available that need to be explored, third, a wide range of edge devices are available with different computation and memory constraints. As such, trade-offs arise among inference time, energy consumption, efficiency (throughput/watt) and value (throughput/dollar). To shed some light in this problem, a case study is delivered where seven Image Classification (IC) and six Object Detection (OD) State-of-The-Art (SOTA) DL models were used to detect face masks on the following commercial off-the-shelf edge devices: Raspberry PI 4, Intel Neural Compute Stick 2, Jetson Nano, Jetson Xavier NX, and i.MX 8M Plus. First, a full end-to-end video pipeline face mask wearing detection architecture is developed. Then, the thirteen DL models were optimized, evaluated and compared on the edge devices, in terms of accuracy and inference time. To leverage the computational power of the edge devices, the models have been optimized, first, by using the SOTA optimization frameworks (TensorFlow Lite, OpenVINO, TensorRT, eIQ) and, second, by evaluating/comparing different optimization options, e.g., different levels of quantization. Note that the five edge devices are evaluated and compared too, in terms of inference time, value and efficiency. Last, we obtain insightful observations on which optimization frameworks, libraries and options to use and on how to select the right device depending on the target metric (inference time, efficiency and value). For example, we show that Jetson Xavier NX platform is the best in terms of latency and efficiency (FPS/Watt), while Jetson Nano is the best in terms of value (FPS/$).

Interval patterns are dependent on metrical position in jazz solos

During jazz improvisation, performers employ short recurrent musical motifs called licks. Past research has focused on the pitch, intervallic, and rhythmic characteristics of licks, but less attention has been paid to whether they tend to start in the same place within the measure ( metrical dependence). Licks might be metrically dependent, and where a given lick starts in a measure ( metrical position) may thus be part of the performer’s mental representation of that lick. Here we report the use of a corpus study to investigate whether licks are metrically dependent. We analyzed a subset of solos, all those in 4/4 time ( n = 435), from the Weimar Jazz Database (WJD; Pfleiderer et al., 2017). Using a sliding window technique, we identified melodic sequences ( interval n-grams) between 3 and 10 intervals in length. We counted the number of times each interval n-gram occurred, and noted the metrical position of the initial note of each occurrence, using different levels of quantization (8th and 16th note). We compared the entropy of the distribution of metrical positions for each n-gram—with lower values indicating a stronger metrical dependence—against simulated counterparts that assumed no relationship between an n-gram and its metrical position (no metrical dependence). Overall, we found that shorter n-grams were metrically dependent, with varying results for longer n-grams. We suggest two possible explanations: either mental representations of licks may encode their metrical features or the metrical position may make certain licks more accessible to the performer. On the basis of our findings we discuss future studies that could employ our methods.

Partition function and coherent states for the quantum multiverse

Abstract Motivated by a wider acceptance of the quantum cosmology and the idea of multiverse, we follow the second and third quantization formalism for the conceptual comprehension of quantum multiverse. The other motivating feature to pursue the present study is that at the different levels of quantization the equation of quantum state is similar to that of harmonic oscillator, which encourages us to give a glancing taste of the coherent states in the domain of quantum cosmology. Third quantization enables one to model the universe as a particle in a quantum field theory of the minisuperspace and the wave description prompts one to recognize interactions among the universes of the multiverse. This picture of the system further instigates us to investigate the partition function and the thermodynamic equations of the system. The path integral formalism with proper mathematical extensions are employed to find the partition function followed by the usual techniques of statistical mechanics to examine the thermodynamic properties.

Hierarchical Color Quantization Based on Self-organization

In this paper, a new hierarchical color quantization method based on self-organizing maps that provides different levels of quantization is presented. Color quantization (CQ) is a typical image processing task, which consists of selecting a small number of code vectors from a set of available colors to represent a high color resolution image with minimum perceptual distortion. Several techniques have been proposed for CQ based on splitting algorithms or cluster analysis. Artificial neural networks and, more concretely, self-organizing models have been usually utilized for this purpose. The self-organizing map (SOM) is one of the most useful algorithms for color image quantization. However, it has some difficulties related to its fixed network architecture and the lack of representation of hierarchical relationships among data. The growing hierarchical SOM (GHSOM) tries to face these problems derived from the SOM model. The architecture of the GHSOM is established during the unsupervised learning process according to the input data. Furthermore, the proposed color quantizer allows the evaluation of different color quantization rates under different codebook sizes, according to the number of levels of the generated neural hierarchy. The experimental results show the good performance of this approach compared to other quantizers based on self-organization.

Multiple quantization and the concept of information

The understanding of the meaning of quantization seems to be the main problem in understanding quantum structures. In this paper first the difference between quantized particle vs. radiation fields in the formalism of canonical quantization is discussed. Next von Weizsacker's concept of “multiple quantization” which leads to an understanding of quantization as an iteration of probability theory is explained. Finally a connection between quantization and the idea of a “general theory of information” is considered. This brings together semantic information with the different levels of quantization and expresses the philosophical attitude of this paper concerning the interpretation of quantum theory.

A Class of All Digital Phase Locked Loops: Modeling and Analysis

An all digital phase locked loop which tracks the phase of the incoming signal once per carrier cycle is proposed. The different elements and their functions, and the phase lock operation are explained in detail. The general digital loop operation is governed by a nonlinear difference equation from which a suitable model is developed. The lock range for the general model is derived. The performance of the digital loop for phase step and frequency step inputs for different levels of quantization without loop filter are studied. The analytical results are checked by simulating the actual system on the digital computer.