pith. sign in

arxiv: 2512.01163 · v2 · submitted 2025-12-01 · 💻 cs.LG · cs.AI

2D-ThermAl: Physics-Informed Framework for Thermal Analysis of Circuits using Generative AI

Soumyadeep Chandra , Sayeed Shafayet Chowdhury , Kaushik Roy This is my paper

Pith reviewed 2026-05-17 02:11 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords thermal analysisphysics-informed neural networksgenerative AIU-Netcircuit designtemperature mappingEDA toolsFEM acceleration
0
0 comments X

The pith

Physics-informed generative AI maps circuit temperatures from activity profiles with 0.71°C error and up to 200 times faster than FEM.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces ThermAl, a physics-informed generative AI framework that takes circuit activity profiles as input and produces full-chip transient and steady-state temperature maps. It relies on a hybrid U-Net architecture with positional encoding and a Boltzmann regularizer to keep the outputs consistent with physical heat transfer principles. The model trains on large sets of heat dissipation maps generated by COMSOL, spanning simple gates to complex layouts, and achieves low error while running far quicker than traditional finite element simulations. A sympathetic reader would care because slow thermal analysis currently forces designers to discover overheating problems only after expensive layout work, limiting how many design iterations are practical early on.

Core claim

ThermAl employs a hybrid U-Net architecture enhanced with positional encoding and a Boltzmann regularizer to maintain physical fidelity. The model identifies heat sources and estimates full-chip transient and steady-state thermal distributions directly from input activity profiles. Trained on an extensive dataset of heat dissipation maps generated via COMSOL from simple logic gates to complex designs, it delivers precise temperature mappings with a root mean squared error of only 0.71°C and runs up to ~200 times faster than conventional FEM tools while maintaining accuracy across diverse layouts, workloads, and an extended temperature range up to 95°C.

What carries the argument

Hybrid U-Net architecture with positional encoding and Boltzmann regularizer, which generates physically consistent thermal distributions from activity profiles.

If this is right

  • Enables rapid early-stage hotspot detection without the computational cost of repeated FEM runs.
  • Supports multiple design iterations in EDA workflows before committing to full physical verification.
  • Maintains accuracy for both steady-state and transient thermal analysis across varied circuit complexities.
  • Extends reliably to elevated temperatures representative of peak power scenarios with under 2.2% full-scale RMSE.
  • Allows thermal pattern learning that can inform layout adjustments before post-layout signoff.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the model transfers to real silicon measurements, it could support online thermal monitoring in deployed chips.
  • The same generative approach might accelerate other slow physics simulations such as mechanical stress or electromagnetic effects in circuits.
  • Fine-tuning on a small set of measured device data could close the gap between simulated training and fabricated hardware behavior.
  • Interactive versions of the model could let designers explore thermal trade-offs in real time during floorplanning.

Load-bearing premise

COMSOL-generated data sufficiently represents the thermal behavior of real fabricated circuits for unseen complex layouts and temperatures outside the 25-55°C training range.

What would settle it

Compare model predictions against direct temperature measurements taken on a fabricated test chip under controlled activity patterns and power levels that match the input profiles used in evaluation.

Figures

Figures reproduced from arXiv: 2512.01163 by Kaushik Roy, Sayeed Shafayet Chowdhury, Soumyadeep Chandra.

Figure 1
Figure 1. Figure 1: As transistor density in modern microprocessors increases, traditional [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Workflow for Dataset Generation: Circuit designs and power dis [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: (a) Comprehensive library of steady-state and transient thermal [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Overview of the ThermAl framework: A novel hybrid U-Net network combining CNN and positional embedding for feature extraction and processing. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative results from the extended-range cross-validation dataset. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: ThermAl identifying hotspots across challenging workloads. Its scal￾ability ensures detailed temperature monitoring even under high-complexity scenarios. 2) Source-target image pair. 3) Feature-level concatenation. This section examines each improvement in detail, supported by both quantitative results (Tables IV and V) and visual evidence (see [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Visualizing the Impact of Physics-Aware Regularization. (a) Workload [PITH_FULL_IMAGE:figures/full_fig_p009_8.png] view at source ↗
read the original abstract

Thermal analysis is increasingly critical in modern integrated circuits, where non-uniform power dissipation and high transistor densities can cause rapid temperature spikes and reliability concerns. Traditional methods, such as FEM-based simulations offer high accuracy but computationally prohibitive for early-stage design, often requiring multiple iterative redesign cycles to resolve late-stage thermal failures. To address these challenges, we propose 'ThermAl', a physics-informed generative AI framework which effectively identifies heat sources and estimates full-chip transient and steady-state thermal distributions directly from input activity profiles. ThermAl employs a hybrid U-Net architecture enhanced with positional encoding and a Boltzmann regularizer to maintain physical fidelity. Our model is trained on an extensive dataset of heat dissipation maps, ranging from simple logic gates (e.g., inverters, NAND, XOR) to complex designs, generated via COMSOL. Experimental results demonstrate that ThermAl delivers precise temperature mappings for large circuits, with a root mean squared error (RMSE) of only 0.71{\deg}C, and outperforms conventional FEM tools by running up to ~200 times faster. We analyze performance across diverse layouts and workloads, and discuss its applicability to large-scale EDA workflows. While thermal reliability assessments often extend beyond 85{\deg}C for post-layout signoff, our focus here is on early-stage hotspot detection and thermal pattern learning. To ensure generalization beyond the nominal operating range 25-55{\deg}C, we additionally performed cross-validation on an extended dataset spanning 25-95{\deg}C maintaining a high accuracy (<2.2% full-scale RMSE) even under elevated temperature conditions representative of peak power and stress scenarios.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces ThermAl, a physics-informed generative AI framework based on a hybrid U-Net architecture with positional encoding and a Boltzmann regularizer. Trained on COMSOL-generated heat dissipation maps ranging from simple logic gates to complex circuit designs, the model predicts full-chip transient and steady-state temperature distributions from activity profiles. It reports an RMSE of 0.71°C on large circuits, up to ~200× speedup over conventional FEM tools, and maintains accuracy (<2.2% full-scale RMSE) under cross-validation on an extended 25-95°C temperature range.

Significance. If the performance and generalization claims are substantiated with rigorous evaluation protocols, the work could meaningfully accelerate early-stage thermal hotspot detection in EDA flows, reducing the number of costly iterative FEM simulations required during design. The hybrid architecture combining data-driven learning with a physics regularizer is a reasonable direction for balancing accuracy and speed in circuit thermal modeling.

major comments (2)
  1. [Abstract / Experimental Results] Abstract and Experimental Results: The central performance claim of 0.71°C RMSE and ~200× speedup is only partially supported because no information is provided on test-set construction, train/test split criteria (e.g., whether test power maps or topologies share sub-circuits, gate counts, or spatial statistics with training examples), or enforced distributional shift. Without these details, it is unclear whether the reported error reflects generalization to novel layouts or interpolation within the COMSOL generative distribution.
  2. [Abstract] Abstract: The cross-validation statement on the 25-95°C extended dataset reports <2.2% full-scale RMSE but does not specify the number of folds, the exact metric definition (full-scale relative to what range?), or whether the elevated-temperature cases were drawn from the same layout distribution as the nominal 25-55°C training data.
minor comments (2)
  1. [Title / Abstract] The title uses '2D-ThermAl' while the abstract refers to 'ThermAl'; consistent naming would improve clarity.
  2. [Abstract] The abstract mentions analysis 'across diverse layouts and workloads' but does not reference any specific figures, tables, or quantitative metrics supporting that analysis.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed review. The comments highlight important aspects of experimental rigor that strengthen the presentation of our results. We address each major comment below and have revised the manuscript to incorporate additional details on dataset construction and evaluation protocols.

read point-by-point responses

  1. Referee: [Abstract / Experimental Results] Abstract and Experimental Results: The central performance claim of 0.71°C RMSE and ~200× speedup is only partially supported because no information is provided on test-set construction, train/test split criteria (e.g., whether test power maps or topologies share sub-circuits, gate counts, or spatial statistics with training examples), or enforced distributional shift. Without these details, it is unclear whether the reported error reflects generalization to novel layouts or interpolation within the COMSOL generative distribution.

    Authors: We agree that the original manuscript provided insufficient detail on test-set construction and split criteria, which is necessary to fully substantiate the generalization claims. In the revised manuscript, we have added a dedicated paragraph in the Experimental Results section describing the dataset partitioning. The test set was constructed from circuits with entirely distinct topologies, gate counts, and spatial power density statistics, with no shared sub-circuits or gate instances relative to the training examples. This enforces a clear distributional shift, confirming that the 0.71°C RMSE reflects performance on novel layouts. We have also expanded the description of the speedup measurement to include the precise hardware configuration and FEM baseline used for the ~200× comparison. revision: yes

  2. Referee: [Abstract] Abstract: The cross-validation statement on the 25-95°C extended dataset reports <2.2% full-scale RMSE but does not specify the number of folds, the exact metric definition (full-scale relative to what range?), or whether the elevated-temperature cases were drawn from the same layout distribution as the nominal 25-55°C training data.

    Authors: We appreciate this observation and have clarified the evaluation protocol in the revised abstract and Experimental Results section. A 5-fold cross-validation was performed on the extended dataset. The full-scale RMSE is defined as RMSE normalized by the temperature range span (25–95°C, or 70°C), expressed as a percentage. The elevated-temperature cases were generated from the same layout distribution as the nominal cases, using scaled activity profiles to reach higher temperatures while preserving circuit topologies. This allows assessment of robustness under peak power conditions representative of stress scenarios. revision: yes

Circularity Check

0 steps flagged

No circularity: standard supervised regression on external COMSOL data with independent test evaluation

full rationale

The paper describes a hybrid U-Net trained on heat dissipation maps generated externally by COMSOL to predict temperature distributions, with performance quantified by empirical RMSE on held-out cases and runtime comparisons to FEM solvers. No equations, fitted parameters, or self-citations are shown to reduce the reported 0.71°C RMSE or speedup claims to tautological re-statements of the training inputs or loss terms; the Boltzmann regularizer functions as a training constraint rather than a definitional identity. The derivation chain consists of standard neural architecture choices plus external simulation benchmarks and is therefore self-contained.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central performance claims rest on the premise that COMSOL-generated heat maps are representative of physical reality and that the added regularizer successfully enforces heat-equation consistency without introducing new free parameters that dominate the fit.

axioms (1)
  • domain assumption The Boltzmann regularizer maintains physical fidelity of predicted temperature fields
    Invoked in the abstract to justify that the generative model respects underlying heat physics beyond pure data fitting.

pith-pipeline@v0.9.0 · 5605 in / 1331 out tokens · 51922 ms · 2026-05-17T02:11:24.814860+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

42 extracted references · 42 canonical work pages

  1. [1]

    A landscape of the new dark silicon design regime.IEEE Micro.33, 8-19 (2013)

    Taylor, M. A landscape of the new dark silicon design regime.IEEE Micro.33, 8-19 (2013)

    work page 2013

  2. [2]

    Amant, R., Sankaralingam, K

    Esmaeilzadeh, H., Blem, E., St. Amant, R., Sankaralingam, K. & Burger, D. Dark silicon and the end of multicore scaling.Proceedings Of The 38th Annual International Symposium On Computer Architecture. pp. 365-376 (2011)

    work page 2011

  3. [3]

    & Martonosi, M

    Brooks, D. & Martonosi, M. Dynamic thermal management for high- performance microprocessors.Proceedings HPCA Seventh International Symposium On High-Performance Computer Architecture. pp. 171-182 (2001)

    work page 2001

  4. [4]

    & Vrudhula, S

    Hanumaiah, V . & Vrudhula, S. Energy-efficient operation of multicore processors by DVFS, task migration, and active cooling.IEEE Transac- tions On Computers.63, 349-360 (2012)

    work page 2012

  5. [5]

    & Wang, H

    Liu, Z., Tan, S., Huang, X. & Wang, H. Task migrations for distributed thermal management considering transient effects.IEEE Transactions On Very Large Scale Integration (VLSI) Systems.23, 397-401 (2014)

    work page 2014

  6. [6]

    & Zhang, Z

    Wang, H., Ma, J., Tan, S., Zhang, C., Tang, H., Huang, K. & Zhang, Z. Hierarchical dynamic thermal management method for high-performance many-core microprocessors.ACM Transactions On Design Automation Of Electronic Systems (TODAES).22, 1-21 (2016)

    work page 2016

  7. [7]

    & Tarjan, D

    Skadron, K., Stan, M., Huang, W., Velusamy, S., Sankaranarayanan, K. & Tarjan, D. Temperature-aware microarchitecture.ACM SIGARCH Computer Architecture News.31, 2-13 (2003)

    work page 2003

  8. [8]

    & Skadron, K

    Kong, J., Chung, S. & Skadron, K. Recent thermal management tech- niques for microprocessors.ACM Computing Surveys (CSUR).44, 1-42 (2012)

    work page 2012

  9. [9]

    & Martonosi, M

    Joseph, R. & Martonosi, M. Run-time power estimation in high perfor- mance microprocessors.Proceedings Of The 2001 International Sympo- sium On Low Power Electronics And Design. pp. 135-140 (2001)

    work page 2001

  10. [10]

    & Martonosi, M

    Isci, C. & Martonosi, M. Runtime power monitoring in high-end processors: Methodology and empirical data.Proceedings. 36th An- nual IEEE/ACM International Symposium On Microarchitecture, 2003. MICRO-36.. pp. 93-104 (2003)

    work page 2003

  11. [11]

    & Tan, S

    Wu, W., Jin, L., Yang, J., Liu, P. & Tan, S. Efficient power modeling and software thermal sensing for runtime temperature monitoring.ACM Transactions On Design Automation Of Electronic Systems (TODAES). 12, 1-29 (2008)

    work page 2008

  12. [12]

    & Reda, S

    Dev, K., Nowroz, A. & Reda, S. Power mapping and modeling of multi- core processors.International Symposium On Low Power Electronics And Design (ISLPED). pp. 39-44 (2013)

    work page 2013

  13. [13]

    & Shakouri, A

    Wang, X., Farsiu, S., Milanfar, P. & Shakouri, A. Power trace: An efficient method for extracting the power dissipation profile in an IC chip from its temperature map.IEEE Transactions On Components And Packaging Technologies.32, 309-316 (2009)

    work page 2009

  14. [14]

    & Reda, S

    Nowroz, A., Cochran, R. & Reda, S. Thermal monitoring of real processors: Techniques for sensor allocation and full characterization. Proceedings Of The 47th Design Automation Conference. pp. 56-61 (2010)

    work page 2010

  15. [15]

    & Nowroz, A

    Reda, S., Cochran, R. & Nowroz, A. Improved thermal tracking for processors using hard and soft sensor allocation techniques.IEEE Trans- actions On Computers.60, 841-851 (2011)

    work page 2011

  16. [16]

    & Reda, S

    Nowroz, A., Woods, G. & Reda, S. Power mapping of integrated circuits using AC-based thermography.IEEE Transactions On Very Large Scale Integration (VLSI) Systems.21, 1398-1409 (2012)

    work page 2012

  17. [17]

    & Vetterli, M

    Ranieri, J., Vincenzi, A., Chebira, A., Atienza, D. & Vetterli, M. EigenMaps: Algorithms for optimal thermal maps extraction and sensor placement on multicore processors.Proceedings Of The 49th Annual Design Automation Conference. pp. 636-641 (2012)

    work page 2012

  18. [18]

    & Kim, L

    Paek, S., Shin, W., Sim, J. & Kim, L. PowerField: A probabilistic approach for temperature-to-power conversion based on Markov random field theory.IEEE Transactions On Computer-Aided Design Of Integrated Circuits And Systems.32, 1509-1519 (2013)

    work page 2013

  19. [19]

    & Benini, L

    Beneventi, F., Bartolini, A., Vivet, P. & Benini, L. Thermal analysis and interpolation techniques for a logic+ wideio stacked dram test chip. IEEE Transactions On Computer-Aided Design Of Integrated Circuits And Systems.35, 623-636 (2015)

    work page 2015

  20. [20]

    & Belouchrani, A

    Reda, S., Dev, K. & Belouchrani, A. Blind identification of thermal models and power sources from thermal measurements.IEEE Sensors Journal.18, 680-691 (2017)

    work page 2017

  21. [21]

    & Tan, S

    Sadiqbatcha, S., Zhao, H., Amrouch, H., Henkel, J. & Tan, S. Hot spot identification and system parameterized thermal modeling for multi-core processors through infrared thermal imaging.2019 Design, Automation & Test In Europe Conference & Exhibition (DATE). pp. 48-53 (2019)

    work page 2019

  22. [22]

    & Tan, S

    Jin, W., Sadiqbatcha, S., Zhang, J. & Tan, S. Full-chip thermal map estimation for commercial multi-core CPUs with generative adversar- ial learning.Proceedings Of The 39th International Conference On Computer-Aided Design. pp. 1-9 (2020)

    work page 2020

  23. [23]

    Chen, C., Chandra, S., Han, Y ., & Seo, H. (2021). Deep learning-based thermal image analysis for pavement defect detection and classification considering complex pavement conditions.Remote Sensing, 14(1), 106

    work page 2021

  24. [24]

    Ma, F., Wang, Y ., Fu, Z., Tang, Y ., Dai, J., Li, C., & Dong, W. (2022). Thermal ageing mechanism of a natural rock-modified asphalt binder using Fourier Transform Infrared Spectroscopy analysis.Construction and Building Materials, 335, 127494

    work page 2022

  25. [25]

    & Weiner, K

    Lasance, C., Vinke, H., Rosten, H. & Weiner, K. A novel approach for the thermal characterization of electronic parts.Proceedings Of 1995 IEEE/CPMT 11th Semiconductor Thermal Measurement And Manage- ment Symposium (SEMI-THERM). pp. 1-9 (1995)

    work page 1995

  26. [26]

    & Wachutka, G

    Gerstenmaier, Y . & Wachutka, G. Rigorous model and network for transient thermal problems.Microelectronics Journal.33, 719-725 (2002)

    work page 2002

  27. [27]

    & Yang, J

    Liu, P., Li, H., Jin, L., Wu, W., Tan, S. & Yang, J. Fast thermal simulation for runtime temperature tracking and management.IEEE Transactions On Computer-Aided Design Of Integrated Circuits And Systems.25, 2882- 2893 (2006)

    work page 2006

  28. [28]

    & Tirumala, M

    Li, D., Tan, S., Pacheco, E. & Tirumala, M. Architecture-level thermal characterization for multicore microprocessors.IEEE Transactions On Very Large Scale Integration (VLSI) Systems.17, 1495-1507 (2009)

    work page 2009

  29. [29]

    & Gall, M

    Gurrum, S., Joshi, Y ., King, W., Ramakrishna, K. & Gall, M. A compact approach to on-chip interconnect heat conduction modeling using the finite element method.Journal Of Electronic Packaging, Transactions Of The ASME.130, 310011-310018 (2008)

    work page 2008

  30. [30]

    Sadiqbatcha, S., Zhao, Y ., Zhang, J., Amrouch, H., Henkel, J., & Tan, S. X. D. (2020, January). Machine learning based online full-chip heatmap estimation.In 2020 25th Asia and South Pacific Design Automation Conference (ASP-DAC) (pp. 229-234). IEEE

    work page 2020

  31. [31]

    R., Skadron, K., Sankaranarayanan, K., Ghosh, S., & Velusam, S

    Huang, W., Stan, M. R., Skadron, K., Sankaranarayanan, K., Ghosh, S., & Velusam, S. Compact thermal modeling for temperature-aware design. In Proceedings of the 41st annual Design Automation Conference (pp. 878-883)

    work page

  32. [32]

    H., Ardestani, E

    Ziabari, A., Park, J. H., Ardestani, E. K., Renau, J., Kang, S. M., & Shakouri, A. (2014). Power blurring: Fast static and transient thermal analysis method for packaged integrated circuits and power devices.IEEE Transactions on Very Large Scale Integration (VLSI) Systems, 22(11), 2366-2379

    work page 2014

  33. [33]

    Yuan, Z., Shukla, P., Chetoui, S., Nemtzow, S., Reda, S., & Coskun, A. K. (2021). PACT: An extensible parallel thermal simulator for emerging integration and cooling technologies.IEEE Transactions on Computer- Aided Design of Integrated Circuits and Systems, 41(4), 1048-1061

    work page 2021

  34. [34]

    3D-ICE: Fast Compact Thermal Modeling for 3D ICs,

    S. Timarchi et al., “3D-ICE: Fast Compact Thermal Modeling for 3D ICs,”ICCAD, 2010

    work page 2010

  35. [35]

    The Manchester Thermal Analyzer (MTA): An Advanced Thermal Simulator for Integrated Systems,

    J. W. Kelly et al., “The Manchester Thermal Analyzer (MTA): An Advanced Thermal Simulator for Integrated Systems,”IEEE TCAD, 2018

    work page 2018

  36. [36]

    DeepOHeat: Operator Learning for Fast and Accurate Thermal Modeling,

    Z. Chen et al., “DeepOHeat: Operator Learning for Fast and Accurate Thermal Modeling,”arXiv:2302.12949, 2023

    work page arXiv 2023

  37. [37]

    (2024), https://www.cadence.com, Version 6.7.1, EDA tool for custom IC design JOURNAL OF LATEX CLASS FILES, VOL

    Cadence Design Systems Cadence Virtuoso. (2024), https://www.cadence.com, Version 6.7.1, EDA tool for custom IC design JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 12

    work page 2024

  38. [38]

    (2024),https://www.tsmc.com, Standard Cell Library for 65nm Process Technology

    Taiwan Semiconductor Manufacturing Company (TSMC) TSMC 65nm Technology Library. (2024),https://www.tsmc.com, Standard Cell Library for 65nm Process Technology

    work page 2024

  39. [39]

    & Kazi, S

    Cass ´e, M., Ghibaudo, G. & Kazi, S. Low temperature characterization and modeling of FDSOI transistors for cryo CMOS applications. (Inte- chOpen,2022)

    work page 2022

  40. [40]

    COM- SOL, Stockholm, Sweden, 2014

    COMSOL Multiphysics®,Heat Transfer Module User’s Guide. COM- SOL, Stockholm, Sweden, 2014

    work page 2014

  41. [41]

    & Brox, T

    Ronneberger, O., Fischer, P. & Brox, T. U-net: Convolutional net- works for biomedical image segmentation.Medical Image Computing And Computer-assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III

    work page 2015

  42. [42]

    & Lerer,A

    Paszke, A., Gross, S., Chintala, S., Chanan, G., Yang, E., DeVito, Z., Lin, Z., Desmaison, A., Antiga, L. & Lerer,A. Automatic differentiation in pytorch.(2017) IX. BIOGRAPHYSECTION Soumyadeep Chandrareceived his B.Tech degree in Electrical and Telecommunication Engineering from Jadavpur University, Kolkata, India in 2020. Currently, he is a 4th year Ph.D...

    work page 2017