Search

Your search keyword '"Huang, Ping-Chen"' showing total 23 results
Start Over Author "Huang, Ping-Chen"
23 results on '"Huang, Ping-Chen"'

1. Efficient Decoding of Compositional Structure in Holistic Representations

Author

Kleyko, Denis, Bybee, Connor, Huang, Ping-Chen, Kymn, Christopher J., Olshausen, Bruno A., Frady, E. Paxon, and Sommer, Friedrich T.

Subjects
Computer Science - Neural and Evolutionary Computing, Computer Science - Information Retrieval, Computer Science - Information Theory
Abstract

We investigate the task of retrieving information from compositional distributed representations formed by Hyperdimensional Computing/Vector Symbolic Architectures and present novel techniques which achieve new information rate bounds. First, we provide an overview of the decoding techniques that can be used to approach the retrieval task. The techniques are categorized into four groups. We then evaluate the considered techniques in several settings that involve, e.g., inclusion of external noise and storage elements with reduced precision. In particular, we find that the decoding techniques from the sparse coding and compressed sensing literature (rarely used for Hyperdimensional Computing/Vector Symbolic Architectures) are also well-suited for decoding information from the compositional distributed representations. Combining these decoding techniques with interference cancellation ideas from communications improves previously reported bounds (Hersche et al., 2021) of the information rate of the distributed representations from 1.20 to 1.40 bits per dimension for smaller codebooks and from 0.60 to 1.26 bits per dimension for larger codebooks., Comment: 28 pages, 5 figures

Published
2023
Full Text
View/download PDF

2. Computing with Hypervectors for Efficient Speaker Identification

Author

Huang, Ping-Chen, Kleyko, Denis, Rabaey, Jan M., Olshausen, Bruno A., and Kanerva, Pentti

Subjects
Computer Science - Sound, Computer Science - Machine Learning, Electrical Engineering and Systems Science - Audio and Speech Processing
Abstract

We introduce a method to identify speakers by computing with high-dimensional random vectors. Its strengths are simplicity and speed. With only 1.02k active parameters and a 128-minute pass through the training data we achieve Top-1 and Top-5 scores of 31% and 52% on the VoxCeleb1 dataset of 1,251 speakers. This is in contrast to CNN models requiring several million parameters and orders of magnitude higher computational complexity for only a 2$\times$ gain in discriminative power as measured in mutual information. An additional 92 seconds of training with Generalized Learning Vector Quantization (GLVQ) raises the scores to 48% and 67%. A trained classifier classifies 1 second of speech in 5.7 ms. All processing was done on standard CPU-based machines.

Published
2022

3. Analysis and Evaluation of Fruit Quality and Volatile Composition of F1 Hybrids between ‘Yuluxiang’ and ‘Shinseiki’ Pear

Author

YUE Yu, HUANG Ping, CHEN Hu, YU Chunliang, PENG Jianying

Subjects
pear, hybrid offspring, fruit quality, gas chromatography-mass spectrometry, volatile aroma, Food processing and manufacture, TP368-456
Abstract

In order to explore the genetic variation between hybrid fruits and their parents and conduct comprehensive evaluation of hybrid fruits, seven promising F1 hybrids (7-93, 8-44, 8-176, 9-188, 9-193, 10-101 and 10-173) between ‘Yuluxiang’ and ‘Shinseiki’ pears, the female parent ‘Yuluxiang’ and the popular cultivar ‘Akizuki’ were evaluated for quality indicators such as single fruit mass, fruit shape index, hardness, soluble solid content, soluble sugar content, titratable acid content, sugar/acid ratio, vitamin C content, and aroma composition. Meanwhile, the differences in aroma among different cultivars were evaluated using orthogonal partial least squares-discriminant analysis (OPLS-DA). The results showed that there was a significant difference in fruit quality among the seven F1 hybrids, with 8-44 having a fruit shape index equal to 1 and the highest soluble solids and sugar content. Altogether, 83 volatile components were detected from the nine cultivars, of which 10-173 and 8-44 had the highest total amount of volatile components, and 10-173 and 10-101 contained the highest number of volatile compounds. The major aroma components in pears were alcohols, esters, and aldehydes. The OPLS-DA results showed that the aroma components of 10-101 and 10-173 were significantly different from those of the other strains. Fruit quality traits of 10-101 had a significantly positive correlation with ethyl acetate and hexyl acetate; fruit quality traits of 10-173 had a significantly positive correlation with ethyl octanoate and α-farnesene and trans-2-hexenal, and their contents were significantly higher than those in the other strains, making the flavor more prominent. According to variable importance in the projection (VIP) results, we predicted 42 aroma markers that differed among varieties, and through odor activity value (OAV) analysis, we identified key volatile compounds that contributed to the characteristic odor of each strain. Through correlation analysis, we identified the association between pear quality traits and characteristic aromas. Principal component analysis (PCA) showed that the comprehensive scores of 8-44, 10-173 and 10-101 had the highest ranking, indicating that they had better fruit quality than the other strains. In summary, 8-44, 10-101, and 10-173 can be regarded as potential pear strains. This study can provide a reference for the breeding and promotion of new pear varieties.

Published
2023
Full Text
View/download PDF

7. Bio-Inspired Artificial Olfactory System

Author

Huang, Ping-Chen

Subjects
Electrical engineering, Bio-Inspired, Nanotechnologies, Neural Computation, Olfactory System
Abstract

The escalating statistical variation of device behaviors in the nano-scale era has led to a search for alternative computational paradigms where randomness can be treated as opportunities rather than nuisance. This will allow computational systems to be built with circuits designed at the nominal case and have better access to the advantages of scaled technologies. Such computational paradigms may emerge as we explore the information processing in the biological sensory systems, which achieve unparalleled performance and energy efficiency with mediocre and unreliable components. Implementations of signal processing tasks such as feature extraction, learning, or recognition can especially benefit from bio-inspired computational models. In this thesis, we present a bio-inspired artificial olfactory system, which consists of an analog feature extraction front end that operates efficiently in the low-precision regime and a spike pattern classifier that exploits the randomness in circuits.The analog front end implements a novel trainable feature extraction algorithm for metal-oxide gas sensor arrays. The algorithm extracts one composite feature of all analytes and transforms the sensor responses into concentration-invariant spike patterns. This composite feature is extracted by performing the gradient decent algorithm during training. This 6-channel analog frond end consumes 519 nW/channel in the training mode, and 463 nW/channel in the recognition mode.The spike pattern classifier consists of a transformation of the input spikes into high-dimensional sparse vectors and a cortical memory model. The transformation is based on a random sampling scheme that can be efficiently performed with circuits exhibiting large parametric variations. Moreover, sparse representations allow fast and robust pattern storage and retrieval with associative memories such as the correlation matrix memory. It’s realized today that hyper-dimensional computing architectures like this may be a perfect match to the emerging nano-scale devices. We show how this classifier can be densely and efficiently implemented in a 3-D CNFET-RRAM technology.

Published
2015

14. A novel coupled fiber ring laser and its application in CATV signals transmission

Author

Huang-Ping Chen, Huang-Cuang Lin, and Shyh-Lin Tsao

Subjects
Optical fiber, Materials science, business.industry, Polarization-maintaining optical fiber, law.invention, Optics, Fiber Bragg grating, Fiber optic sensor, law, Fiber laser, Dispersion-shifted fiber, business, Plastic optical fiber, Photonic-crystal fiber
Abstract

We demonstrate a fiber laser with three coupled fiber rings and a 2/spl times/2 electro-optical modulator for transmitting CATV signals in this paper. The system performance including CNR, CSO, and CTB are also reported.

Published
2004
Full Text
View/download PDF

22. A novel structural analysis of glycerophosphocholines as TFA/K+adducts by electrospray ionization ion trap tandem mass spectrometry

Author

Ho, Yen‐Peng and Huang, Ping‐Chen

Abstract

When collisionally activated dissociation (CAD) of glycerophosphocholine (GPC) species is examined using quadrupole ion trap mass spectrometry (QITMS), the spectral patterns differ from those obtained using sector or quadrupole mass spectrometry. Methods employed in the structural analysis of GPCs using a sector or quadrupole mass spectrometers are not necessarily useful for an ion trap mass spectrometer. A novel method is presented for structurally analyzing GPCs that involves the CAD of trifluoroacetic acid (TFA) adducts of kaliated GPCs. Solutions of GPCs in 0.1% TFA/methanol were electrosprayed to produce precursor ions by attaching a trifluoroacetic acid (TFA) molecule to a kaliated GPC molecule. The CAD‐MS/MS spectra obtained by QITMS revealed a dramatic increase in the abundance of fragment ions, corresponding to the losses of sn‐1 and sn‐2 fatty acyl substituents. A preferential loss of the sn‐1 fatty acyl group over the loss of the sn‐2 fatty acyl group was observed among the GPC standards examined. A GPC extract from egg yolk was directly analyzed by this method without prior separation. The identities and positions of fatty acyl substituents of over 20 GPC species were identified. Some isomers present in very low relative abundance, which could not be analyzed by QITMS/MS using other ions as precursors, were identified by the TFA attachment method. Copyright © 2002 John Wiley & Sons, Ltd.

Published
2002
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results