-
Large Language Model Agent as a Mechanical Designer
Authors:
Yayati Jadhav,
Amir Barati Farimani
Abstract:
Conventional mechanical design paradigms rely on experts systematically refining concepts through experience-guided modification and FEA to meet specific requirements. However, this approach can be time-consuming and heavily dependent on prior knowledge and experience. While numerous machine learning models have been developed to streamline this intensive and expert-driven iterative process, these…
▽ More
Conventional mechanical design paradigms rely on experts systematically refining concepts through experience-guided modification and FEA to meet specific requirements. However, this approach can be time-consuming and heavily dependent on prior knowledge and experience. While numerous machine learning models have been developed to streamline this intensive and expert-driven iterative process, these methods typically demand extensive training data and considerable computational resources. Furthermore, methods based on deep learning are usually restricted to the specific domains and tasks for which they were trained, limiting their applicability across different tasks. This creates a trade-off between the efficiency of automation and the demand for resources. In this study, we present a novel approach that integrates pre-trained LLMs with a FEM module. The FEM module evaluates each design and provides essential feedback, guiding the LLMs to continuously learn, plan, generate, and optimize designs without the need for domain-specific training. We demonstrate the effectiveness of our proposed framework in managing the iterative optimization of truss structures, showcasing its capability to reason about and refine designs according to structured feedback and criteria. Our results reveal that these LLM-based agents can successfully generate truss designs that comply with natural language specifications with a success rate of up to 90%, which varies according to the applied constraints. By employing prompt-based optimization techniques we show that LLM based agents exhibit optimization behavior when provided with solution-score pairs to iteratively refine designs to meet specifications. This ability of LLM agents to produce viable designs and optimize them based on their inherent reasoning capabilities highlights their potential to develop and implement effective design strategies autonomously.
△ Less
Submitted 9 May, 2024; v1 submitted 26 April, 2024;
originally announced April 2024.
-
The Seoul National University AGN Monitoring Project III: H$β$ lag measurements of 32 luminous AGNs and the high-luminosity end of the size--luminosity relation
Authors:
Jong-Hak Woo,
Shu Wang,
Suvendu Rakshit,
Ho** Cho,
Donghoon Son,
Vardha N. Bennert,
Elena Gallo,
Edmund Hodges-Kluck,
Tommaso Treu,
Aaron J. Barth,
Wan** Cho,
Adi Foord,
Jaehyuk Geum,
Hengxiao Guo,
Yashashree Jadhav,
Yiseul Jeon,
Kyle M. Kabasares,
Won-Suk Kang,
Changseok Kim,
Min** Kim,
Tae-Woo Kim,
Huynh Anh N. Le,
Matthew A. Malkan,
Amit Kumar Mandal,
Daeseong Park
, et al. (6 additional authors not shown)
Abstract:
We present the main results from a long-term reverberation map** campaign carried out for the Seoul National University Active Galactic Nuclei (AGN) Monitoring Project. High-quality data were obtained during 2015-2021 for 32 luminous AGNs (i.e., continuum luminosity in the range of $10^{44-46}$ erg s$^{-1}$) at a regular cadence, of 20-30 days for spectroscopy and 3-5 days for photometry. We obt…
▽ More
We present the main results from a long-term reverberation map** campaign carried out for the Seoul National University Active Galactic Nuclei (AGN) Monitoring Project. High-quality data were obtained during 2015-2021 for 32 luminous AGNs (i.e., continuum luminosity in the range of $10^{44-46}$ erg s$^{-1}$) at a regular cadence, of 20-30 days for spectroscopy and 3-5 days for photometry. We obtain time lag measurements between the variability in the H$β$ emission and the continuum for 32 AGNs; twenty-five of those have the best lag measurements based on our quality assessment, examining correlation strength, and the posterior lag distribution. Our study significantly increases the current sample of reverberation-mapped AGNs, particularly at the moderate to high luminosity end. Combining our results with literature measurements, we derive a H$β$ broad line region size--luminosity relation with a shallower slope than reported in the literature. For a given luminosity, most of our measured lags are shorter than the expectation, implying that single-epoch black hole mass estimators based on previous calibrations could suffer large systematic uncertainties.
△ Less
Submitted 26 November, 2023;
originally announced November 2023.
-
GPT-MolBERTa: GPT Molecular Features Language Model for molecular property prediction
Authors:
Suryanarayanan Balaji,
Rishikesh Magar,
Yayati Jadhav,
Amir Barati Farimani
Abstract:
With the emergence of Transformer architectures and their powerful understanding of textual data, a new horizon has opened up to predict the molecular properties based on text description. While SMILES are the most common form of representation, they are lacking robustness, rich information and canonicity, which limit their effectiveness in becoming generalizable representations. Here, we present…
▽ More
With the emergence of Transformer architectures and their powerful understanding of textual data, a new horizon has opened up to predict the molecular properties based on text description. While SMILES are the most common form of representation, they are lacking robustness, rich information and canonicity, which limit their effectiveness in becoming generalizable representations. Here, we present GPT-MolBERTa, a self-supervised large language model (LLM) which uses detailed textual descriptions of molecules to predict their properties. A text based description of 326000 molecules were collected using ChatGPT and used to train LLM to learn the representation of molecules. To predict the properties for the downstream tasks, both BERT and RoBERTa models were used in the finetuning stage. Experiments show that GPT-MolBERTa performs well on various molecule property benchmarks, and approaching state of the art performance in regression tasks. Additionally, further analysis of the attention mechanisms show that GPT-MolBERTa is able to pick up important information from the input textual data, displaying the interpretability of the model.
△ Less
Submitted 10 October, 2023; v1 submitted 20 September, 2023;
originally announced October 2023.
-
The Seoul National University AGN Monitoring Project IV: H$α$ reverberation map** of 6 AGNs and the H$α$ Size-Luminosity Relation
Authors:
Ho** Cho,
Jong-Hak Woo,
Shu Wang,
Donghoon Son,
Jae** Shin,
Suvendu Rakshit,
Aaron J. Barth,
Vardha N. Bennert,
Elena Gallo,
Edmund Hodges-Kluck,
Tommaso Treu,
Hyun-** Bae,
Wan** Cho,
Adi Foord,
Jaehyuk Geum,
Yashashree Jadhav,
Yiseul Jeon,
Kyle M. Kabasares,
Daeun Kang,
Wonseok Kang,
Changseok Kim,
Donghwa Kim,
Min** Kim,
Taewoo Kim,
Huynh Anh N. Le
, et al. (7 additional authors not shown)
Abstract:
The broad line region (BLR) size-luminosity relation has paramount importance for estimating the mass of black holes in active galactic nuclei (AGNs). Traditionally, the size of the H$β$ BLR is often estimated from the optical continuum luminosity at 5100\angstrom{} , while the size of the H$α$ BLR and its correlation with the luminosity is much less constrained. As a part of the Seoul National Un…
▽ More
The broad line region (BLR) size-luminosity relation has paramount importance for estimating the mass of black holes in active galactic nuclei (AGNs). Traditionally, the size of the H$β$ BLR is often estimated from the optical continuum luminosity at 5100\angstrom{} , while the size of the H$α$ BLR and its correlation with the luminosity is much less constrained. As a part of the Seoul National University AGN Monitoring Project (SAMP) which provides six-year photometric and spectroscopic monitoring data, we present our measurements of the H$α$ lags of 6 high-luminosity AGNs. Combined with the measurements for 42 AGNs from the literature, we derive the size-luminosity relations of H$α$ BLR against broad H$α$ and 5100\angstrom{} continuum luminosities. We find the slope of the relations to be $0.61\pm0.04$ and $0.59\pm0.04$, respectively, which are consistent with the \hb{} size-luminosity relation. Moreover, we find a linear relation between the 5100\angstrom{} continuum luminosity and the broad H$α$ luminosity across 7 orders of magnitude. Using these results, we propose a new virial mass estimator based on the H$α$ broad emission line, finding that the previous mass estimates based on the scaling relations in the literature are overestimated by up to 0.7 dex at masses lower than $10^7$~M$_{\odot}$.
△ Less
Submitted 29 June, 2023;
originally announced June 2023.
-
Determining star formation rates of active galactic nuclei host galaxies based on SED fitting with sub-mm data
Authors:
Changseok Kim,
Jong-Hak Woo,
Yashashree Jadhav,
Aeree Chung,
Junhyun Baek,
Jeong Ae Lee,
Jae** Shin,
Ho Seong Hwang,
Rongxin Luo,
Donghoon Son,
HyunGi Kim,
Hyuk Woo
Abstract:
We present the star formation rate (SFR) measurements based on the spectral energy distribution (SED) analysis with new sub-mm fluxes combined with archival multi-wavelength data for a sample of 52 AGN host galaxies at z $<0.2$. We carried out sub-mm observations using the SCUBA-2 camera at the James Clerk Maxwell Telescope, and obtained flux or an upper limit at 450 and 850 $μ$m for each target.…
▽ More
We present the star formation rate (SFR) measurements based on the spectral energy distribution (SED) analysis with new sub-mm fluxes combined with archival multi-wavelength data for a sample of 52 AGN host galaxies at z $<0.2$. We carried out sub-mm observations using the SCUBA-2 camera at the James Clerk Maxwell Telescope, and obtained flux or an upper limit at 450 and 850 $μ$m for each target. By experimenting the effect of the AGN dust component in the SED fit, we find that dust luminosity can be overestimated if AGN contribution is ignored. While the SFR based on 4000Å break shows a significant offset compared to dust luminosity based SFR, the SFR obtained by the artificial neural network (Ellison et al. 2016) generally shows consistency albeit with a large scatter. We find that SFR correlates with AGN outflow strength manifested by the [OIII] $λ5007$ emission line, and that AGNs with higher Eddington ratios and stronger outflows are in general hosted by galaxies with higher SFR, which is consistent with the correlation reported by Woo et al. (2020). This suggests no instantaneous quenching of star formation due to AGN feedback.
△ Less
Submitted 21 February, 2022;
originally announced February 2022.
-
The spatially offset quasar E1821+643: New evidence for gravitational recoil
Authors:
Yashashree Jadhav,
Andrew Robinson,
Triana Almeyda,
Rachel Curran,
Alessandro Marconi
Abstract:
A galaxy merger is expected to cause the formation of a supermassive black hole (SMBH) binary, which itself eventually coalesces through the anisotropic emission of gravitational waves. This may result in the merged SMBH receiving a recoil kick velocity ~100 - 1000 km/s, causing it to oscillate in the gravitational potential of the host galaxy. The luminous quasar E1821+643, identified as an SMBH…
▽ More
A galaxy merger is expected to cause the formation of a supermassive black hole (SMBH) binary, which itself eventually coalesces through the anisotropic emission of gravitational waves. This may result in the merged SMBH receiving a recoil kick velocity ~100 - 1000 km/s, causing it to oscillate in the gravitational potential of the host galaxy. The luminous quasar E1821+643, identified as an SMBH recoil candidate via spectropolarimetry observations, shows Doppler shifting of the broad emission lines in direct and scattered light, consistent with a relative velocity of 2100 km/s between the quasar nucleus and host galaxy. In this paper, we attempt to detect the expected spatial displacement using a combination of optical spectroastrometry and Hubble Space Telescope (HST) narrow band images. The spectroastrometry reveals a relative spatial displacement between the quasar nucleus and the gas emitting the [OIII]4959,5007 lines of ~130mas (~580pc) to the North-West. Our HST images resolve the [OIII] emission on sub-arcsecond scales, showing that it is asymmetrically distributed, extending to radial distances ~0.5 - 0.6" from the nucleus in a wide arc running from the North-East around to the West. A simulated spectroastrometry observation based on the HST [OIII] image indicates that only a small fraction of the measured displacement can be attributed to the asymmetric [OIII] emission. This displacement therefore appears to be a real spatial offset of the quasar nucleus with respect to the narrow-line region, presumed to be located at the host galaxy center, further supporting the interpretation that a post-merger gravitational recoil of the SMBH has occurred in E1821+643.
△ Less
Submitted 30 July, 2021;
originally announced July 2021.
-
Dominant motion identification of multi-particle system using deep learning from video
Authors:
Yayati Jadhav,
Amir Barati Farimani
Abstract:
Identifying underlying governing equations and physical relevant information from high-dimensional observable data has always been a challenge in physical sciences. With the recent advances in sensing technology and available datasets, various machine learning techniques have made it possible to distill underlying mathematical models from sufficiently clean and usable datasets. However, most of th…
▽ More
Identifying underlying governing equations and physical relevant information from high-dimensional observable data has always been a challenge in physical sciences. With the recent advances in sensing technology and available datasets, various machine learning techniques have made it possible to distill underlying mathematical models from sufficiently clean and usable datasets. However, most of these techniques rely on prior knowledge of the system and noise-free data obtained by simulation of physical system or by direct measurements of the signals. Hence, the inference obtained by using these techniques is often unreliable to be used in the real world where observed data is noisy and requires feature engineering to extract relevant features. In this work, we provide a deep-learning framework that extracts relevant information from real-world videos of highly stochastic systems, with no prior knowledge and distills the underlying governing equation representing the system. We demonstrate this approach on videos of confined multi-agent/particle systems of ants, termites, fishes as well as a simulated confined multi-particle system with elastic collision interactions. Furthermore, we explore how these seemingly diverse systems have predictable underlying behavior. In this study, we have used computer vision and motion tracking to extract spatial trajectories of individual agents/particles in a system, and by using LSTM VAE we projected these features on a low-dimensional latent space from which the underlying differential equation representing the data was extracted using SINDy framework.
△ Less
Submitted 26 April, 2021;
originally announced April 2021.
-
Pre-outburst observations of Nova Del 2013 from Pan-STARRS 1
Authors:
N. R. Deacon,
D. W. Hoard,
E. A. Magnier,
Y. S. Jadhav,
M. Huber,
K. C. Chambers,
H. Flewelling,
K. W. Hodapp,
N. Kaiser,
R. P. Kudritzki,
N. Metcalfe,
C. Waters
Abstract:
Nova Delphini 2013 was identified on the 14th of August 2013 and eventually rose to be a naked eye object. We sought to study the behaviour of the object in the run-up to outburst and to compare it to the pre-outburst photometric characteristics of other novae. We searched the Pan-STARRS 1 datastore to identify pre-outburst photometry of Nova Del 2013 and identified twenty-four observations in the…
▽ More
Nova Delphini 2013 was identified on the 14th of August 2013 and eventually rose to be a naked eye object. We sought to study the behaviour of the object in the run-up to outburst and to compare it to the pre-outburst photometric characteristics of other novae. We searched the Pan-STARRS 1 datastore to identify pre-outburst photometry of Nova Del 2013 and identified twenty-four observations in the 1.2 years before outburst. The progenitor of Nova Delphini showed variability of a few tenths of a magnitude but did not brighten significantly in comparison with archival plate photometry. We also found that the object did not vary significantly on the approximately half hour timescale between pairs of Pan-STARRS 1 observations.
△ Less
Submitted 24 February, 2014;
originally announced February 2014.
