-
Robot Air Hockey: A Manipulation Testbed for Robot Learning with Reinforcement Learning
Authors:
Caleb Chuck,
Carl Qi,
Michael J. Munje,
Shuozhe Li,
Max Rudolph,
Chang Shi,
Siddhant Agarwal,
Harshit Sikchi,
Abhinav Peri,
Sarthak Dayal,
Evan Kuo,
Kavan Mehta,
Anthony Wang,
Peter Stone,
Amy Zhang,
Scott Niekum
Abstract:
Reinforcement Learning is a promising tool for learning complex policies even in fast-moving and object-interactive domains where human teleoperation or hard-coded policies might fail. To effectively reflect this challenging category of tasks, we introduce a dynamic, interactive RL testbed based on robot air hockey. By augmenting air hockey with a large family of tasks ranging from easy tasks like…
▽ More
Reinforcement Learning is a promising tool for learning complex policies even in fast-moving and object-interactive domains where human teleoperation or hard-coded policies might fail. To effectively reflect this challenging category of tasks, we introduce a dynamic, interactive RL testbed based on robot air hockey. By augmenting air hockey with a large family of tasks ranging from easy tasks like reaching, to challenging ones like pushing a block by hitting it with a puck, as well as goal-based and human-interactive tasks, our testbed allows a varied assessment of RL capabilities. The robot air hockey testbed also supports sim-to-real transfer with three domains: two simulators of increasing fidelity and a real robot system. Using a dataset of demonstration data gathered through two teleoperation systems: a virtualized control environment, and human shadowing, we assess the testbed with behavior cloning, offline RL, and RL from scratch.
△ Less
Submitted 5 May, 2024;
originally announced May 2024.
-
MRQ:Support Multiple Quantization Schemes through Model Re-Quantization
Authors:
Manasa Manohara,
Sankalp Dayal,
Tariq Afzal,
Rahul Bakshi,
Kahkuen Fu
Abstract:
Despite the proliferation of diverse hardware accelerators (e.g., NPU, TPU, DPU), deploying deep learning models on edge devices with fixed-point hardware is still challenging due to complex model quantization and conversion. Existing model quantization frameworks like Tensorflow QAT [1], TFLite PTQ [2], and Qualcomm AIMET [3] supports only a limited set of quantization schemes (e.g., only asymmet…
▽ More
Despite the proliferation of diverse hardware accelerators (e.g., NPU, TPU, DPU), deploying deep learning models on edge devices with fixed-point hardware is still challenging due to complex model quantization and conversion. Existing model quantization frameworks like Tensorflow QAT [1], TFLite PTQ [2], and Qualcomm AIMET [3] supports only a limited set of quantization schemes (e.g., only asymmetric per-tensor quantization in TF1.x QAT [4]). Accordingly, deep learning models cannot be easily quantized for diverse fixed-point hardwares, mainly due to slightly different quantization requirements. In this paper, we envision a new type of model quantization approach called MRQ (model re-quantization), which takes existing quantized models and quickly transforms the models to meet different quantization requirements (e.g., asymmetric -> symmetric, non-power-of-2 scale -> power-of-2 scale). Re-quantization is much simpler than quantizing from scratch because it avoids costly re-training and provides support for multiple quantization schemes simultaneously. To minimize re-quantization error, we developed a new set of re-quantization algorithms including weight correction and rounding error folding. We have demonstrated that MobileNetV2 QAT model [7] can be quickly re-quantized into two different quantization schemes (i.e., symmetric and symmetric+power-of-2 scale) with less than 0.64 units of accuracy loss. We believe our work is the first to leverage this concept of re-quantization for model quantization and models obtained from the re-quantization process have been successfully deployed on NNA in the Echo Show devices.
△ Less
Submitted 3 August, 2023; v1 submitted 1 August, 2023;
originally announced August 2023.
-
Targeting Learning: Robust Statistics for Reproducible Research
Authors:
Jeremy R. Coyle,
Nima S. Hejazi,
Ivana Malenica,
Rachael V. Phillips,
Benjamin F. Arnold,
Andrew Mertens,
Jade Benjamin-Chung,
Weixin Cai,
Sonali Dayal,
John M. Colford Jr.,
Alan E. Hubbard,
Mark J. van der Laan
Abstract:
Targeted Learning is a subfield of statistics that unifies advances in causal inference, machine learning and statistical theory to help answer scientifically impactful questions with statistical confidence. Targeted Learning is driven by complex problems in data science and has been implemented in a diversity of real-world scenarios: observational studies with missing treatments and outcomes, per…
▽ More
Targeted Learning is a subfield of statistics that unifies advances in causal inference, machine learning and statistical theory to help answer scientifically impactful questions with statistical confidence. Targeted Learning is driven by complex problems in data science and has been implemented in a diversity of real-world scenarios: observational studies with missing treatments and outcomes, personalized interventions, longitudinal settings with time-varying treatment regimes, survival analysis, adaptive randomized trials, mediation analysis, and networks of connected subjects. In contrast to the (mis)application of restrictive modeling strategies that dominate the current practice of statistics, Targeted Learning establishes a principled standard for statistical estimation and inference (i.e., confidence intervals and p-values). This multiply robust approach is accompanied by a guiding roadmap and a burgeoning software ecosystem, both of which provide guidance on the construction of estimators optimized to best answer the motivating question. The roadmap of Targeted Learning emphasizes tailoring statistical procedures so as to minimize their assumptions, carefully grounding them only in the scientific knowledge available. The end result is a framework that honestly reflects the uncertainty in both the background knowledge and the available data in order to draw reliable conclusions from statistical analyses - ultimately enhancing the reproducibility and rigor of scientific findings.
△ Less
Submitted 12 June, 2020;
originally announced June 2020.
-
A Survey of Human Activity Recognition Using WiFi CSI
Authors:
Siamak Yousefi,
Hirokazu Narui,
Sankalp Dayal,
Stefano Ermon,
Shahrokh Valaee
Abstract:
In this article, we present a survey of recent advances in passive human behaviour recognition in indoor areas using the channel state information (CSI) of commercial WiFi systems. Movement of human body causes a change in the wireless signal reflections, which results in variations in the CSI. By analyzing the data streams of CSIs for different activities and comparing them against stored models,…
▽ More
In this article, we present a survey of recent advances in passive human behaviour recognition in indoor areas using the channel state information (CSI) of commercial WiFi systems. Movement of human body causes a change in the wireless signal reflections, which results in variations in the CSI. By analyzing the data streams of CSIs for different activities and comparing them against stored models, human behaviour can be recognized. This is done by extracting features from CSI data streams and using machine learning techniques to build models and classifiers. The techniques from the literature that are presented herein have great performances, however, instead of the machine learning techniques employed in these works, we propose to use deep learning techniques such as long-short term memory (LSTM) recurrent neural network (RNN), and show the improved performance. We also discuss about different challenges such as environment change, frame rate selection, and multi-user scenario, and suggest possible directions for future work.
△ Less
Submitted 23 August, 2017;
originally announced August 2017.
-
Characterization and Characteristics of mechanochemically synthesized amorphous fast ionic conductor 50 SISOMO (50AgI-25Ag2O-25MoO3)
Authors:
Saurabh Dayal,
K. Shahi
Abstract:
Mechanochemically synthesized amorphous 50SISOMO [50AgI-25Ag_2O-25MoO_3] fast ionic conductor shows high ionic conductivity of ~ 6x10^-3 Ω^-1 cm-1 at room temperature. The highest ionic conductivity is achieved for 36 h milled sample, which is more than three orders of magnitude higher than that of crystalline AgI at room temperature. The samples are thermally stable at least up to ~70 °C. Thermoe…
▽ More
Mechanochemically synthesized amorphous 50SISOMO [50AgI-25Ag_2O-25MoO_3] fast ionic conductor shows high ionic conductivity of ~ 6x10^-3 Ω^-1 cm-1 at room temperature. The highest ionic conductivity is achieved for 36 h milled sample, which is more than three orders of magnitude higher than that of crystalline AgI at room temperature. The samples are thermally stable at least up to ~70 °C. Thermoelectric power studies on 50 SISOMO amorphous fast ionic conductors (a-SIC) have been carried out in the temperature range 300-330K. Thermoelectric power (S) is found to vary linearly with the inverse of the absolute temperature, and can be expressed by the equation -S = [(0.19 \times 10^3/T) + 0.25] mV/K. The heat of transport (q*) of Ag+ ion i.e. 0.19 eV is nearly equal to the activation energy (E) i.e. 0.20 eV of Ag+ ion migration calculated from the conductivity plots indicating that the material has an average structure. This is also in consonance with earlier theories on heats of transport of ions in ionic solids.
△ Less
Submitted 7 April, 2012;
originally announced April 2012.
-
Absence of long-range superconducting correlations in the frustrated 1/2-filled band Hubbard model
Authors:
S. Dayal,
R. T. Clay,
S. Mazumdar
Abstract:
We present many-body calculations of superconducting pair-pair correlations in the ground state of the half-filled band Hubbard model on large anisotropic triangular lattices. Our calculations cover nearly the complete range of anisotropies between the square and isotropic triangular lattice limits. We find that the superconducting pair-pair correlations decrease monotonically with increasing onsi…
▽ More
We present many-body calculations of superconducting pair-pair correlations in the ground state of the half-filled band Hubbard model on large anisotropic triangular lattices. Our calculations cover nearly the complete range of anisotropies between the square and isotropic triangular lattice limits. We find that the superconducting pair-pair correlations decrease monotonically with increasing onsite Hubbard interaction U for inter-pair distances greater than nearest neighbor. For the large lattices of interest here the distance dependence of the correlations approaches that for noninteracting electrons. Both these results are consistent with the absence of superconductivity in this model in the thermodynamic limit. We conclude that the effective 1/2-filled band Hubbard model, suggested by many authors to be appropriate for the kappa-(BEDT-TTF)-based organic charge-transfer solids, does not explain the superconducting transition in these materials.
△ Less
Submitted 13 April, 2012; v1 submitted 24 January, 2012;
originally announced January 2012.
-
Beyond the quantum spin liquid concept in frustrated two dimensional organic superconductors
Authors:
R. T. Clay,
S. Dayal,
H. Li,
S. Mazumdar
Abstract:
The occurrence of antiferromagnetism in kappa-(ET)_2X can be understood within an effective 1/2-filled band with dimers of ET molecules containing one hole each. We argue that while this effective model can describe the presence of antiferromagnetism, a complete description for these materials requires the correct carrier density of one-half per molecule. For dimerized and strongly frustrated 1/4-…
▽ More
The occurrence of antiferromagnetism in kappa-(ET)_2X can be understood within an effective 1/2-filled band with dimers of ET molecules containing one hole each. We argue that while this effective model can describe the presence of antiferromagnetism, a complete description for these materials requires the correct carrier density of one-half per molecule. For dimerized and strongly frustrated 1/4-filled lattices we show that a singlet-paired state coexisting with charge ordering occurs that we have termed the Paired Electron Crystal (PEC). Here we investigate the 1/4-filled model on a dimerized lattice, showing regions where AFM, PEC, and the Wigner-crystal occur. We point out the need to go beyond quantum spin liquid concepts for highly frustrated materials such as kappa-(ET)_2Cu_2(CN)_3 and beta'-EtMe_3Sb[Pd(dmit)_2]_2 which we believe are PECs at low temperatures.
△ Less
Submitted 23 November, 2011;
originally announced November 2011.
-
Ground state and finite temperature behavior of 1/4-filled band zigzag ladders
Authors:
R. T. Clay,
J. P. Song,
S. Dayal,
S. Mazumdar
Abstract:
We consider the simplest example of lattice frustration in the 1/4-filled band, a one-dimensional chain with next-nearest neighbor interactions. For this zigzag ladder with electron-electron as well as electron-phonon interactions we present numerical results for ground state as well as thermodynamic properties. In this system the ground state bond distortion pattern is independent of electron-ele…
▽ More
We consider the simplest example of lattice frustration in the 1/4-filled band, a one-dimensional chain with next-nearest neighbor interactions. For this zigzag ladder with electron-electron as well as electron-phonon interactions we present numerical results for ground state as well as thermodynamic properties. In this system the ground state bond distortion pattern is independent of electron-electron interaction strength. The spin gap in the ground state of the zigzag ladder increases with the degree of frustration. Unlike in one-dimension, where the spin-gap and charge ordering transitions can be distinct, we show that in the ladder they occur simultaneously. We discuss spin gap and charge ordering transitions in 1/4-filled materials with one, two, or three dimensional crystal structures. We show empirically that regardless of dimensionality the occurrence of simultaneous or distinct charge and magnetic transitions can be correlated with the ground state bond distortion pattern.
△ Less
Submitted 10 May, 2012; v1 submitted 21 August, 2011;
originally announced August 2011.
-
The Paired Electron Crystal: order from frustration in the quarter-filled band
Authors:
S. Dayal,
R. T. Clay,
H. Li,
S. Mazumdar
Abstract:
We present a study of the effects of simultaneous charge- and spin-frustration on the two-dimensional strongly correlated quarter-filled band on an anisotropic triangular lattice. The broken-symmetry states that dominate in the weakly frustrated region near the rectangular lattice limit are the well known antiferromagnetic state with in-phase lattice dimerization along one direction, and the Wigne…
▽ More
We present a study of the effects of simultaneous charge- and spin-frustration on the two-dimensional strongly correlated quarter-filled band on an anisotropic triangular lattice. The broken-symmetry states that dominate in the weakly frustrated region near the rectangular lattice limit are the well known antiferromagnetic state with in-phase lattice dimerization along one direction, and the Wigner crystal state with the checkerboard charge order. For moderate to strong frustration, however, the dominant phase is a novel spin-singlet paired-electron crystal (PEC), consisting of pairs of charge-rich sites separated by pairs of charge-poor sites. The PEC, with coexisting charge-order and spin-gap in two dimension, is the quarter-filled band equivalent of the valence bond solid (VBS) that can appear in the frustrated half-filled band within antiferromagnetic spin Hamiltonians. We discuss the phase diagram as a function of on-site and intersite Coulomb interactions as well as electron-phonon coupling strength. We speculate that the spin-bonded pairs of the PEC can become mobile for even stronger frustration, giving rise to a paired-electron liquid. We discuss the implications of the PEC concept for understanding several classes of quarter-filled band materials that display unconventional superconductivity, focusing in particular on organic charge transfer solids. Our work points out the need to go beyond quantum spin liquid (QSL) concepts for highly frustrated organic charge-transfer solids such as kappa-(BEDT-TTF)_2Cu_2(CN)_3 and EtMe_3Sb[Pd(dmit)_2]_2, which we believe show frustration-induced charge disproportionation at low temperatures. We discuss possible application to layered cobaltates and 1/4-filled band spinels.
△ Less
Submitted 8 February, 2011;
originally announced February 2011.
