-
Performance Prediction for Large Systems via Text-to-Text Regression
Authors:
Yash Akhauri,
Bryan Lewandowski,
Cheng-Hsi Lin,
Adrian N. Reyes,
Grant C. Forbes,
Arissa Wongpanich,
Bangding Yang,
Mohamed S. Abdelfattah,
Sagi Perel,
Xingyou Song
Abstract:
In many industries, predicting metric outcomes of large systems is a fundamental problem, driven largely by traditional tabular regression. However, such methods struggle on complex systems data in the wild such as configuration files or system logs, where feature engineering is often infeasible. We propose text-to-text regression as a general, scalable alternative. For predicting resource efficie…
▽ More
In many industries, predicting metric outcomes of large systems is a fundamental problem, driven largely by traditional tabular regression. However, such methods struggle on complex systems data in the wild such as configuration files or system logs, where feature engineering is often infeasible. We propose text-to-text regression as a general, scalable alternative. For predicting resource efficiency on Borg, Google's massive compute cluster scheduling system, a 60M parameter encoder-decoder, trained from random initialization, achieves up to a near perfect 0.99 (0.9 average) rank correlation across the entire fleet, and 100x lower MSE than tabular approaches. The model also easily adapts to new tasks in only 500 few-shot examples and captures the densities of complex outcome distributions. Ablation studies highlight the importance of using encoders, increasing sequence length, and the model's inherent uncertainty quantification. These findings pave the way for universal simulators of real-world outcomes.
△ Less
Submitted 26 June, 2025;
originally announced June 2025.
-
An incremental algorithm based on multichannel non-negative matrix partial co-factorization for ambient denoising in auscultation
Authors:
Juan De La Torre Cruz,
Francisco Jesus Canadas Quesada,
Damian Martinez-Munoz,
Nicolas Ruiz Reyes,
Sebastian Garcia Galan,
Julio Jose Carabias Orti
Abstract:
The aim of this study is to implement a method to remove ambient noise in biomedical sounds captured in auscultation. We propose an incremental approach based on multichannel non-negative matrix partial co-factorization (NMPCF) for ambient denoising focusing on high noisy environment with a Signal-to-Noise Ratio (SNR) <= -5 dB. The first contribution applies NMPCF assuming that ambient noise can b…
▽ More
The aim of this study is to implement a method to remove ambient noise in biomedical sounds captured in auscultation. We propose an incremental approach based on multichannel non-negative matrix partial co-factorization (NMPCF) for ambient denoising focusing on high noisy environment with a Signal-to-Noise Ratio (SNR) <= -5 dB. The first contribution applies NMPCF assuming that ambient noise can be modelled as repetitive sound events simultaneously found in two single-channel inputs captured by means of different recording devices. The second contribution proposes an incremental algorithm, based on the previous multichannel NMPCF, that refines the estimated biomedical spectrogram throughout a set of incremental stages by eliminating most of the ambient noise that was not removed in the previous stage at the expense of preserving most of the biomedical spectral content. The ambient denoising performance of the proposed method, compared to some of the most relevant state-of-the-art methods, has been evaluated using a set of recordings composed of biomedical sounds mixed with ambient noise that typically surrounds a medical consultation room to simulate high noisy environments with a SNR from -20 dB to -5 dB. Experimental results report that: (i) the performance drop suffered by the proposed method is lower compared to MSS and NLMS; (ii) unlike what happens with MSS and NLMS, the proposed method shows a stable trend of the average SDR and SIR results regardless of the type of ambient noise and the SNR level evaluated; and (iii) a remarkable advantage is the high robustness of the estimated biomedical sounds when the two single-channel inputs suffer from a delay between them.
△ Less
Submitted 1 November, 2024;
originally announced November 2024.
-
Digital compensation of the side-band-rejection ratio in a fully analog 2SB sub-millimeter receiver
Authors:
R. Rodriguez,
R. Finger,
F. P. Mena,
A. Alvear,
R. Fuentes,
A. Khudchenko,
R. Hesper,
A. M. Baryshev,
N. Reyes,
L. Bronfman
Abstract:
In observational radio astronomy, sideband-separating receivers are preferred, particularly under high atmospheric noise, which is usually the case in the sub-millimeter range. However, obtaining a good rejection ratio between the two sidebands is difficult since, unavoidably, imbalances in the different analog components appear. We describe a method to correct these imbalances without making any…
▽ More
In observational radio astronomy, sideband-separating receivers are preferred, particularly under high atmospheric noise, which is usually the case in the sub-millimeter range. However, obtaining a good rejection ratio between the two sidebands is difficult since, unavoidably, imbalances in the different analog components appear. We describe a method to correct these imbalances without making any change in the analog part of the sideband-separating receiver, specifically, keeping the intermediate-frequency hybrid in place. This opens the possibility of implementing the method in any existing receiver. We have built hardware to demonstrate the validity of the method and tested it on a fully analog receiver operating between 600 and 720GHz. We have tested the stability of calibration and performance vs time and after full resets of the receiver. We have performed an error analysis to compare the digital compensation in two configurations of analog receivers, with and without intermediate frequency (IF) hybrid. An average compensated sideband rejection ratio of 46dB is obtained. Degradation of the compensated sideband rejection ratio on time and after several resets of the receiver is minimal. A receiver with an IF hybrid is more robust to systematic errors. Moreover, we have shown that the intrinsic random errors in calibration have the same impact for configuration without IF hybrid and for a configuration with IF hybrid with analog rejection ratio better than 10dB. Compensated rejection ratios above 40dB are obtained even in the presence of high analog rejection. The method is robust allowing its use under normal operational conditions at any telescope. We also demonstrate that a full analog receiver is more robust against systematic errors. Finally, the error bars associated to the compensated rejection ratio are almost independent of whether IF hybrid is present or not.
△ Less
Submitted 11 June, 2018;
originally announced June 2018.
