-
Hitting "Probe"rty with Non-Linearity, and More
Authors:
Avik Pal,
Madhura Pawar
Abstract:
Structural probes learn a linear transformation to find how dependency trees are embedded in the hidden states of language models. This simple design may not allow for full exploitation of the structure of the encoded information. Hence, to investigate the structure of the encoded information to its full extent, we incorporate non-linear structural probes. We reformulate the design of non-linear s…
▽ More
Structural probes learn a linear transformation to find how dependency trees are embedded in the hidden states of language models. This simple design may not allow for full exploitation of the structure of the encoded information. Hence, to investigate the structure of the encoded information to its full extent, we incorporate non-linear structural probes. We reformulate the design of non-linear structural probes introduced by White et al. making its design simpler yet effective. We also design a visualization framework that lets us qualitatively assess how strongly two words in a sentence are connected in the predicted dependency tree. We use this technique to understand which non-linear probe variant is good at encoding syntactical information. Additionally, we also use it to qualitatively investigate the structure of dependency trees that BERT encodes in each of its layers. We find that the radial basis function (RBF) is an effective non-linear probe for the BERT model than the linear probe.
△ Less
Submitted 25 February, 2024;
originally announced February 2024.
-
Distributed Deep Reinforcement Learning for Collaborative Spectrum Sharing
Authors:
Pranav M. Pawar,
Amir Leshem
Abstract:
Spectrum sharing among users is a fundamental problem in the management of any wireless network. In this paper, we discuss the problem of distributed spectrum collaboration without central management under general unknown channels. Since the cost of communication, coordination and control is rapidly increasing with the number of devices and the expanding bandwidth used there is an obvious need to…
▽ More
Spectrum sharing among users is a fundamental problem in the management of any wireless network. In this paper, we discuss the problem of distributed spectrum collaboration without central management under general unknown channels. Since the cost of communication, coordination and control is rapidly increasing with the number of devices and the expanding bandwidth used there is an obvious need to develop distributed techniques for spectrum collaboration where no explicit signaling is used. In this paper, we combine game-theoretic insights with deep Q-learning to provide a novel asymptotically optimal solution to the spectrum collaboration problem. We propose a deterministic distributed deep reinforcement learning(D3RL) mechanism using a deep Q-network (DQN). It chooses the channels using the Q-values and the channel loads while limiting the options available to the user to a few channels with the highest Q-values and among those, it selects the least loaded channel. Using insights from both game theory and combinatorial optimization we show that this technique is asymptotically optimal for large overloaded networks. The selected channel and the outcome of the successful transmission are fed back into the learning of the deep Q-network to incorporate it into the learning of the Q-values. We also analyzed performance to understand the behavior of D3RL in differ
△ Less
Submitted 6 April, 2021;
originally announced April 2021.
-
Large Area X-ray Proportional Counter (LAXPC) Instrument on AstroSat and Some Preliminary Results from its performance in the orbit
Authors:
P. C. Agrawal,
J. S. Yadav,
H. M. Antia,
Dhiraj Dedhia,
P. Shah,
Jai Verdhan Chauhan,
R. K. Manchanda,
V. R. Chitnis,
V. M. Gujar,
Tilak Katoch,
V. N. Kurhade,
P. Madhwani,
T. K. Manojkumar,
V. A. Nikam,
A. S. Pandya,
J. V. Parmar,
D. M. Pawar,
Jayashree Roy,
B. Paul,
Mayukh Pahari,
Ranjeev Misra,
M. H. Ravichandran,
K. Anilkumar,
C. C. Joseph,
K. H. Navalgund
, et al. (3 additional authors not shown)
Abstract:
Large Area X-ray Propositional Counter (LAXPC) instrument on AstroSat is aimed at providing high time resolution X-ray observations in 3 to 80 keV energy band with moderate energy resolution. To achieve large collecting area, a cluster of three co-aligned identical LAXPC detectors, is used to realize an effective area in access of about 6000 cm2 at 15 keV. The large detection volume of the LAXPC d…
▽ More
Large Area X-ray Propositional Counter (LAXPC) instrument on AstroSat is aimed at providing high time resolution X-ray observations in 3 to 80 keV energy band with moderate energy resolution. To achieve large collecting area, a cluster of three co-aligned identical LAXPC detectors, is used to realize an effective area in access of about 6000 cm2 at 15 keV. The large detection volume of the LAXPC detectors, filled with xenon gas at about 2 atmosphere pressure, results in detection efficiency greater than 50%, above 30 keV. In this article, we present salient features of the LAXPC detectors, their testing and characterization in the laboratory prior to launch and calibration in the orbit. Some preliminary results on timing and spectral characteristics of a few X-ray binaries and other type of sources, are briefly discussed to demonstrate that the LAXPC instrument is performing as planned in the orbit.
△ Less
Submitted 25 May, 2017; v1 submitted 18 May, 2017;
originally announced May 2017.
-
Calibration of the Large Area X-ray Proportional Counter (LAXPC) instrument on-board AstroSat
Authors:
H. M. Antia,
J. S. Yadav,
P. C. Agrawal,
Jai Verdhan Chauhan,
R. K. Manchanda,
Varsha Chitnis,
Biswajit Paul,
Dhiraj Dedhia,
Parag Shah,
V. M. Gujar,
Tilak Katoch,
V. N. Kurhade,
Pankaj Madhwani,
T. K. Manojkumar,
V. A. Nikam,
A. S. Pandya,
J. V. Parmar,
D. M. Pawar,
Mayukh Pahari,
Ranjeev Misra,
K. H. Navalgund,
R. Pandiyan,
K. S. Sharma,
K. Subbarao
Abstract:
We present the calibration and background model for the Large Area X-ray Proportional Counter (LAXPC) detectors on-board AstroSat. LAXPC instrument has three nominally identical detectors to achieve large collecting area. These detectors are independent of each other and in the event analysis mode, they record the arrival time and energy of each photon that is detected. The detectors have a time-r…
▽ More
We present the calibration and background model for the Large Area X-ray Proportional Counter (LAXPC) detectors on-board AstroSat. LAXPC instrument has three nominally identical detectors to achieve large collecting area. These detectors are independent of each other and in the event analysis mode, they record the arrival time and energy of each photon that is detected. The detectors have a time-resolution of 10 $μ$s and a dead-time of about 42 $μ$s. This makes LAXPC ideal for timing studies. The energy resolution and peak channel to energy map** were obtained from calibration on ground using radioactive sources coupled with GEANT4 simulations of the detectors. The response matrix was further refined from observations of the Crab X-ray source after launch. At around 20 keV the energy resolution of detector is about 10--15\%, while the combined effective area of the 3 detectors is about 6000 cm$^2$.
△ Less
Submitted 18 June, 2017; v1 submitted 27 February, 2017;
originally announced February 2017.
-
Performance of Large Area X-ray Proportional Counters in a Balloon Experiment
Authors:
J. Roy,
P. C. Agrawal,
D. K. Dedhia,
R. K. Manchanda,
P. B. Shah,
V. R. Chitnis,
V. M. Gujar,
J. V. Parmar,
D. M. Pawar,
V. B. Kurhade
Abstract:
ASTROSAT is India's first satellite fully devoted to astronomical observations covering a wide spectral band from optical to hard X-rays by a complement of 4 co-aligned instruments and a Scanning Sky X-ray Monitor. One of the instruments is Large Area X-ray Proportional Counter with 3 identical detectors. In order to assess the performance of this instrument, a balloon experiment with two prototyp…
▽ More
ASTROSAT is India's first satellite fully devoted to astronomical observations covering a wide spectral band from optical to hard X-rays by a complement of 4 co-aligned instruments and a Scanning Sky X-ray Monitor. One of the instruments is Large Area X-ray Proportional Counter with 3 identical detectors. In order to assess the performance of this instrument, a balloon experiment with two prototype Large Area X-ray Proportional Counters (LAXPC) was carried out on 2008 April 14. The design of these LAXPCs was similar to those on the ASTROSAT except that their field of view (FOV) was 3$^{\circ}$ $\times$ 3$^{\circ}$ versus FOV of 1$^{\circ}$ $\times$ 1$^{\circ}$ for the LAXPCs on the ASTROSAT. The LAXPCs are aimed at the timing and spectral studies of X-ray sources in 3-80 keV region. In the balloon experiment, the LAXPC, associated electronics and support systems were mounted on an oriented platform which could be pre-programmed to track any source in the sky. A brief description of the LAXPC design, laboratory tests, calibration and the detector characteristics is presented here. The details of the experiment and background counting rates of the 2 LAXPCs at the float altitude of about 41 km are presented in different energy bands. The bright black hole X-ray binary Cygnus X-1 (Cyg X-1) was observed in the experiment for $\sim$ 3 hours. Details of Cyg X-1 observations, count rates measured from it in different energy intervals and the intensity variations of Cyg X-1 detected during the observations are presented and briefly discussed.
△ Less
Submitted 12 September, 2016;
originally announced September 2016.
