Assessing the Performance of Analog Training for Transfer Learning

Corey Lammie; Malte J. Rasch; Omobayode Fagbohungbe; Takashi Ando; Tayfun Gokmen; Vijay Narayanan

REVIEW 3 major objections 6 minor 33 references

Assessing the Performance of Analog Training for Transfer Learning

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read c-TTv2 brings analog transfer learning within 2% of digital fine-tuning.

desk verdict First simulation of c-TTv2 analog training for a ViT under transfer learning, landing within ~2% of digital fine-tuning; the robustness sweep is the real contribution, but reporting gaps need work. read the letter →

arxiv 2505.11067 v1 pith:WI6B73RS submitted 2025-05-16 cs.LG cs.AIcs.ARcs.CVcs.DCcs.NE

Omobayode Fagbohungbe , Corey Lammie , Malte J. Rasch , Takashi Ando , Tayfun Gokmen , Vijay Narayanan This is my paper · ORCID

classification cs.LGcs.AIcs.ARcs.CVcs.DCcs.NE

keywords analogin-memorycomputingtransferlearningc-TTv2Swin-ViTReRAMweightnoisefine-tuningrobustness

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Analog in-memory computing promises to move neural-network training onto resistive crossbar arrays, but the asymmetric, nonlinear switching of real memory devices breaks ordinary training algorithms. This paper evaluates whether an algorithm called chopped TTv2 (c-TTv2), which cancels reference-value offsets by periodically flipping the sign of gradient accumulation, can fine-tune a pretrained Swin-ViT transformer on analog hardware. Simulating HfOx ReRAM arrays, the authors find that analog transfer learning beats both analog and digital training from scratch and lands within roughly 2% of digital transfer learning on 2-class and 5-class subsets of CIFAR100. They also find that one-shot weight-transfer noise is harmless up to about 15% for the simpler task and 10% for the harder task, and that the algorithm tolerates variation in symmetry point, pulse update noise, and device-to-device spread. If these results hold on physical chips, analog fine-tuning could be a low-energy path for adapting pretrained models at the edge.

What carries the argument

The load-bearing mechanism is the c-TTv2 algorithm, a chopped variant of the TTv2 analog training scheme. TTv2 separates weight and gradient accumulation onto dedicated devices around a programmed reference value and applies digital low-pass filtering; c-TTv2 adds periodic or random sign changes ('chopping') to that accumulation so that any offset from a drifting reference value cancels instead of building up. That sign-chopping is what permits training on devices with asymmetric, nonlinear, and variable switching behavior. Around that update rule, the paper builds a fine-tuning pipeline: convert the pretrained digital weights to analog conductances with one-shot additive noise $W_{\text{noise}} = W_{\text{pre-trained}} + \tau \mathcal{N}(0,1)$, then run analog fine-tuning with the same hyperparameter schedule used for digital training. Device physics enter through the softbound HfOx ReRAM model, whose symmetry-point skew, symmetry-point variability, pulse update noise, and device-to-device variation are the knobs swept in the robustness experiments.

What would settle it

On a physical HfOx ReRAM crossbar array, program a pretrained Swin-ViT, measure the actual distribution of weight-transfer error, and fine-tune with c-TTv2 on the same 2-class and 5-class CIFAR100 subsets; if the measured error is correlated, non-Gaussian, or large enough to push the accuracy gap to digital transfer learning well beyond the simulated ~2%, the central claim fails.

Watch

Extended reading notes

Core claim

At its core, the paper claims that c-TTv2 makes analog transfer learning practical at transformer scale. A digitally pretrained Swin-ViT is converted to analog conductances with one-shot additive white Gaussian noise, $\mathcal{N}(0,1)$, scaled by a noise factor $\tau$ as in Eq. (1), and then fine-tuned on a smaller downstream task using the c-TTv2 update rule on simulated HfOx ReRAM arrays. On 2-class and 5-class subsets of CIFAR100, the resulting analog model outperforms both analog training from scratch and digital training from scratch, and it trails digital transfer learning by about 2%. The same simulations show that accuracy stays nearly flat until weight-transfer noise passes a critical threshold (approximately 15% for the 2-class task, 10% for the 5-class task), and that the training is robust against symmetry point skew, symmetry point variability, pulse update noise, and mean pulse device-to-device variation. The authors conclude that c-TTv2 is suitable for analog transfer learning and that the approach is competitive with digital transfer learning.

Load-bearing premise

The central results assume the simulated softbound HfOx ReRAM model, including the one-shot additive white Gaussian weight-transfer noise of Eq. (1), faithfully represents what real HfOx devices do during fine-tuning.

Editorial extensions

If this is right

Analog fine-tuning with c-TTv2 can deploy a pretrained transformer at the edge without a full digital retrain, and it beats training the same analog model from scratch.
Weight programming precision is not a hard requirement: transfer-noise-induced errors below roughly 10–15% of the weight scale do not measurably hurt downstream accuracy.
The algorithm's tolerance to pulse update noise and device-to-device variation means the same trained analog model can be used across arrays with relaxed per-device calibration.
This is the first reported simulation of analog in-memory training of a vision transformer, opening the same protocol to larger transformer and LLM-scale models.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the simulated robustness carries to hardware, the same chopping scheme should let even larger pretrained models be fine-tuned without bit-exact weight programming, since the noise tolerance relaxes the transfer step.
The critical-noise elbows suggest a concrete design rule: keep analog programming error below about 10% of the weight range, and accuracy should remain stable; an on-chip calibration loop that trims transfer noise to that bound could make the approach hardware-ready.
The paper attributes the analog advantage over digital from-scratch training to noise-induced regularization; that suggests deliberately injecting controlled noise during digital fine-tuning might reproduce part of the benefit without analog hardware.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The one thing to know: this is the first paper to simulate analog in-memory training of a Vision Transformer for transfer learning, using the c-TTv2 algorithm, and it reports that analog fine-tuning lands within about 2% of digital fine-tuning on 2-class and 5-class CIFAR subsets. It also sweeps device non-idealities and finds the algorithm tolerates weight-transfer noise up to 10–15% and is fairly robust to symmetry and pulse variability. If those numbers hold up on physical hardware, it is a useful data point for edge fine-tuning.

What the paper does well: it is an honest application and robustness study, not a new algorithm. It sets up the right baseline comparisons—digital TL, analog from scratch, digital from scratch—and the central trend (analog TL beats analog from scratch and roughly matches digital TL) is plausible. The use of a 26M-parameter Swin-ViT is a step beyond the toy models in earlier work. The robustness sweeps are the real contribution; figures like the weight-transfer-noise elbow are exactly what hardware developers need.

The stress-test note about Eq. (1) is fair but not fatal. The paper models weight transfer error as one-shot additive white Gaussian noise, scaled by tau. That is a simplification, and it could be optimistic for real HfOx devices with column-correlated programming errors or drift. But the manuscript is explicitly a simulation study; it does not claim hardware validation. The right fix is to state that caveat clearly and, ideally, calibrate the noise model to the 14nm ReRAM data they cite. As it stands, the robustness numbers should be read as simulator behavior, not guaranteed chip behavior.

The soft spots are mostly reporting rigor. There are no error bars or multiple seeds; the figures appear to be single runs. Section III.A contains a direct contradiction: one sentence says the analog TL model outperforms the digital model, the next says digital TL outperforms analog by about 2%. That needs a clean rewrite. They also don't include a direct TTv2 or AGAD baseline in this paper, even though they reference AGAD in the conclusion. Per-configuration learning-rate tuning is fine, but it should be reported systematically. No code or data is released, which makes independent checking harder.

Verdict: the central claim holds up as a simulation result. The flaws are addressable and do not undermine the main trend. A serious referee should engage with this; the paper is worth peer review with major revisions. I would cite it only if I were working on analog training algorithms, which I am not, but I would bring it to a reading group focused on AI hardware or transfer learning at the edge.

Referee Report

3 major / 6 minor

Summary. The paper reports aihwkit-based simulations of the c-TTv2 training algorithm for analog in-memory transfer learning using a Swin-ViT model pretrained on CIFAR-10 and fine-tuned on 2-class and 5-class subsets of CIFAR-100. The study compares analog transfer learning (TL) against digital TL and against analog and digital training from scratch, and investigates robustness to weight-transfer noise, symmetry point skew, symmetry point variability, pulse update noise, and device-to-device variability. The authors conclude that c-TTv2 is suitable for analog TL, that analog TL outperforms both analog and digital training from scratch, and that it comes within roughly 2% of digital TL, with critical weight-transfer noise at about 15% (2-class) and 10% (5-class).

Significance. If the results are reproducible, this is a useful step toward analog in-memory fine-tuning of transformer-scale models, which has mostly been demonstrated only on small networks. The paper's strengths include the use of the open-source aihwkit simulator, a nontrivial model (Swin-ViT with ~26.6M parameters), four comparison baselines (analog/digital TL; analog/digital from scratch), and robustness sweeps over several device parameters. The work is clearly relevant to the AIMC community and provides a baseline for future algorithm development. However, because the results are simulation-only, based on single-run traces, and contain an internal contradiction in the headline comparison, the quantitative claims should be treated with caution until strengthened.

major comments (3)

[III.A, Fig. 2] Section III.A contains a direct contradiction about the central result. The paragraph first states 'the analog TL model outperforms the digital model’s performance' and then states 'digital TL outperforms the analog TL by about 2%.' These statements cannot both be true. The subsequent explanation in terms of noise-based regularization is predicated on the first (incorrect) statement. This must be corrected and the intended claim stated unambiguously, since the paper's main conclusion depends on it.
[II, III.B and all figures] All reported results appear to be single-run traces: no seeds, no repeated trials, and no error bars are given for any of the comparisons or robustness sweeps. The claims of a '~2% gap' between analog and digital TL and the 'critical' weight-transfer-noise elbows at ~15% and ~10% are quantitative statements that need uncertainty estimates. Please report means and standard deviations (or at least show runs from multiple seeds) for the main traces and for the elbow points in Fig. 3.
[II, Eq. (1) and Fig. 3] The weight-transfer noise model in Eq. (1) is one-shot additive iid Gaussian noise with a single scalar tau, applied once before fine-tuning. This model is not calibrated to the 14nm HfOx ReRAM devices cited as motivation, and no justification is given for assuming spatially white, homoscedastic, stationary transfer error. If real programming errors have column correlations, low-rank structure, or drift, the robustness elbows in Fig. 3 could shrink substantially. At minimum, the softbound/aihwkit default parameters (e.g., Wmax, Wmin, update noise scales, symmetry point distribution) should be stated explicitly; preferably, add a correlated-noise ablation or refer to measured device statistics to support the white-noise assumption.

minor comments (6)

[I, Introduction] There is a typo: 'the TTv2 algorithm algorithm was used' should read 'the TTv2 algorithm was used.'
[III.A, last paragraph] The sentence 'the performance difference is insignificant compared to the earlier algorithm [20]' is confusing because reference [20] is the c-TTv2 paper itself, not an earlier baseline. Please clarify which algorithm is being compared and cite the appropriate prior work.
[IV, Conclusion] The phrase 'as expected' is not scientifically neutral; remove it or replace with a quantitative comparison to the baselines.
[II, Methodology] The manuscript does not state whether the fine-tuning and from-scratch runs share the same random seed or initialization. This information is necessary for reproducing the comparison.
[All figures] Figure 4's vertical axis appears to be unlabeled; please add an axis label. Also, figure legends use underscored names ('Dig_TL'), which should be typeset as readable labels.
[II, Data availability] There is no code or data availability statement. For a simulation-based study, including the configuration scripts or a link to the aihwkit configuration used would greatly improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the c-TTv2 transfer-learning results are benchmarked against independent digital and analog baselines, and the inherited c-TTv2 algorithm is used as a tool rather than as the evidence for the paper's conclusions.

full rationale

The paper's primary empirical claim is that c-TTv2 enables analog transfer learning on a Swin-ViT, performing better than analog training from scratch and competitively with digital TL (within roughly 2%). These claims are evaluated against independent baselines (digital TL, analog/digital training from scratch) in Figs. 2-5, so the central result is not defined in terms of a fitted constant or constructed from its own output. Eq. (1) introduces an assumed one-shot additive Gaussian weight-transfer noise model (W_noise = W_pre_trained + tau N(0,1)); Fig. 3 then sweeps tau. This is a simulation under an explicit input assumption, not a prediction derived from the assumption by renaming, and it is not used to fit the model. The c-TTv2 algorithm itself is taken from the authors' prior work [20]; this is a normal method citation, and the paper does not invoke [20] as the proof of its transfer-learning result, nor does it import a uniqueness theorem from [20]. The conclusion's statement that AGAD benchmarking remains future work is an acknowledged limitation, not a circular load-bearing step. The Sec. III.A inconsistency ('analog TL model outperforms the digital model's performance' vs. 'digital TL outperforms the analog TL by about 2%') is a correctness or clarity issue, but it does not constitute a derivation that reduces to its inputs. No circular step satisfying the quoting-and-reduction standard was found.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

All central claims are empirical and simulation-based. The analysis adds no fitted theory constants, so the free-parameter count is mostly swept device and algorithm settings. The load-bearing axioms are the fidelity of Eq. (1)'s Gaussian weight-transfer model and of the softbound ReRAM simulation, plus the assumption that the aihwkit implementation of c-TTv2 matches [20]. No invented entities are introduced.

free parameters (6)

Weight transfer noise factor tau = swept through elbow at about 15% for 2-class and 10% for 5-class
Eq. (1), W_noise = W_pre_trained + tau N(0,1), defines the magnitude of weight transfer error. The central robustness claim depends on this user-chosen scale; tau is not measured from hardware and is treated as a knob in Fig. 3.
Symmetry point skew (SpS) = swept as a percentage; exact range not stated
Eq. (2) defines SpS as Wmax/(Wmax-Wmin)*100. Fig. 4's robustness conclusion is conditioned on the authors' chosen sweep range and the softbound device model.
Symmetry point variability (SpV) = swept; 5% SpV equals a variability of 0.10 in the +/-1 weight range
Section III.C states that 5% SpV equals 0.10 of the weight range; the robustness thresholds, such as about 10% SpV for the 5-class task, are based on this scale.
Pulse update noise and mean pulse device-to-device variability = swept over unreported ranges in Fig. 5
Fig. 5 claims near-constant test error as these scalars increase, but the text does not list the numerical levels used, so the sweep is not independently reproducible.
Learning rate = 0.01 in Table I, but varied to select the best model per configuration
The methodology states the learning rate is varied to select the model with the best inference performance for each device setting. The per-condition search can inflate apparent robustness, and the selected values are not reported.
c-TTv2 algorithm hyperparameters = transfer_every=1.0, autogranularity=10000, device momentum=0, in_chop probability=0.10, auto_momentum=0.99
Table I fixes these values and they are inherited from the authors' prior c-TTv2 paper [20]; the sensitivity of the transfer-learning conclusions to these hyperparameters is not studied.

assumptions (4)

domain assumption Equation (1) models weight transfer from digital to analog as W_noise = W_pre_trained + tau N(0,1), with zero-mean unit-variance Gaussian noise applied once before fine-tuning.
Section II. If actual programming error is not i.i.d. Gaussian or appears gradually during training, the critical weight-transfer-noise values from Fig. 3 do not generalize to hardware.
domain assumption The softbound device model in aihwkit faithfully represents the switching behavior of HfOx ReRAM devices.
Section III.C invokes the softbound model as more representative of the device behavior. The robustness claims for symmetry and pulse noise inherit this simulator's fidelity.
domain assumption The aihwkit implementation of c-TTv2 matches the algorithm described in reference [20].
The paper does not derive or re-verify c-TTv2; it cites [20] and applies the algorithm. Errors in the inherited implementation would propagate into all results.
domain assumption The 2-class and 5-class CIFAR-100 subsets are valid lower-complexity transfer tasks relative to CIFAR-10 pretraining.
Section II states that transfer learning requires the downstream task complexity to be lower than the pretraining task; the paper does not quantify task complexity and simply assumes the subsets satisfy the condition.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Assessing the Performance of Analog Training for Transfer Learning." pith.science (2026). https://pith.science/paper/WI6B73RS

@misc{pith2026250511067,
  author       = {Pith},
  title        = {Pith review of: Assessing the Performance of Analog Training for Transfer Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WI6B73RS}},
  note         = {Machine review of arXiv:2505.11067}
}

read the original abstract

Analog in-memory computing is a next-generation computing paradigm that promises fast, parallel, and energy-efficient deep learning training and transfer learning (TL). However, achieving this promise has remained elusive due to a lack of suitable training algorithms. Analog memory devices exhibit asymmetric and non-linear switching behavior in addition to device-to-device variation, meaning that most, if not all, of the current off-the-shelf training algorithms cannot achieve good training outcomes. Also, recently introduced algorithms have enjoyed limited attention, as they require bi-directionally switching devices of unrealistically high symmetry and precision and are highly sensitive. A new algorithm chopped TTv2 (c-TTv2), has been introduced, which leverages the chopped technique to address many of the challenges mentioned above. In this paper, we assess the performance of the c-TTv2 algorithm for analog TL using a Swin-ViT model on a subset of the CIFAR100 dataset. We also investigate the robustness of our algorithm to changes in some device specifications, including weight transfer noise, symmetry point skew, and symmetry point variability

Figures

Figures reproduced from arXiv: 2505.11067 by the authors.

**Figure 2.** Training results of the analog and digital Swin [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Relationship between the inference accuracy and [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Relationship between the model inference accuracy [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]

**Figure 5.** Figure 5: Relationship between the model inference accuracy, [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 23 canonical work pages

[1]

A survey on deep transfer learning,

C. Tan, F. Sun, T. Kong, W. Zhang, C. Yang, and C. Liu, “A survey on deep transfer learning,” in Artificial Neural Networks and Machine Learning–ICANN 2018: 27th International Conference on Artificial Neural Networks, Rhodes, Greece, October 4-7, 2018, Proceedings, Part III 27. Springer, 2018, pp. 270–279

work page 2018
[2]

On the opportunities and risks of foundation models,

R. Bommasani, D. A. Hudson, E. Adeli, R. Altman, S. Arora, S. von Arx, M. S. Bernstein, J. Bohg, A. Bosselut, E. Brunskill et al. , “On the opportunities and risks of foundation models,” arXiv preprint arXiv:2108.07258, 2021

arXiv 2021
[3]

Detecting llm-generated text in computing education: Comparative study for chatgpt cases,

M. S. Orenstrakh, O. Karnalim, C. A. Suarez, and M. Liut, “Detecting llm-generated text in computing education: Comparative study for chatgpt cases,” in 2024 IEEE 48th Annual Computers, Software, and Applications Conference (COMPSAC) . IEEE, 2024, pp. 121–126

2024
[4]

Large language models in medicine,

A. J. Thirunavukarasu, D. S. J. Ting, K. Elangovan, L. Gutierrez, T. F. Tan, and D. S. W. Ting, “Large language models in medicine,” Nature medicine, vol. 29, no. 8, pp. 1930–1940, 2023

1930
[5]

Driving with llms: Fusing object- level vector modality for explainable autonomous driving,

L. Chen, O. Sinavski, J. Hünermann, A. Karnsund, A. J. Willmott, D. Birch, D. Maund, and J. Shotton, “Driving with llms: Fusing object- level vector modality for explainable autonomous driving,” in2024 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2024, pp. 14 093–14 100

work page 2024
[6]

Large language models in finance: A survey,

Y . Li, S. Wang, H. Ding, and H. Chen, “Large language models in finance: A survey,” in Proceedings of the fourth ACM international conference on AI in finance , 2023, pp. 374–382

work page 2023
[7]

Emergent abilities of large language models,

J. Wei, Y . Tay, R. Bommasani, C. Raffel, B. Zoph, S. Borgeaud, D. Yogatama, M. Bosma, D. Zhou, D. Metzler et al., “Emergent abilities of large language models,” arXiv preprint arXiv:2206.07682 , 2022

arXiv 2022
[8]

A comparison review of transfer learning and self-supervised learning: Definitions, applications, advantages and limitations,

Z. Zhao, L. Alzubaidi, J. Zhang, Y . Duan, and Y . Gu, “A comparison review of transfer learning and self-supervised learning: Definitions, applications, advantages and limitations,” Expert Systems with Appli- cations, p. 122807, 2023

work page 2023

Show all 33 references

[9]

A survey on multimodal large language models,

S. Yin, C. Fu, S. Zhao, K. Li, X. Sun, T. Xu, and E. Chen, “A survey on multimodal large language models,” arXiv preprint arXiv:2306.13549 , 2023

2023 arXiv
[10]

A joint energy and latency framework for transfer learning over 5g industrial edge networks,

B. Yang, O. Fagbohungbe, X. Cao, C. Yuen, L. Qian, D. Niyato, and Y . Zhang, “A joint energy and latency framework for transfer learning over 5g industrial edge networks,” IEEE Transactions on Industrial Informatics, vol. 18, no. 1, pp. 531–541, 2021

2021
[11]

Efficient privacy preserving edge intelligent computing framework for image classifica- tion in iot,

O. Fagbohungbe, S. R. Reza, X. Dong, and L. Qian, “Efficient privacy preserving edge intelligent computing framework for image classifica- tion in iot,” IEEE Transactions on Emerging Topics in Computational Intelligence, vol. 6, no. 4, pp. 941–956, 2021

2021
[12]

Demonstration of transfer learning using 14 nm technology analog reram array,

F. F. Athena, O. Fagbohungbe, N. Gong, M. J. Rasch, J. Penaloza, S. Seo, A. Gasasira, P. Solomon, V . Bragaglia, S. Consiglio et al. , “Demonstration of transfer learning using 14 nm technology analog reram array,” Frontiers in Electronics, vol. 4, p. 1331280, 2024

2024
[13]

Ohm’s law + kirchhoff’s current law = better ai: Neural-network processing done in memory with analog circuits will save energy,

G. W. Burr, A. Sebastian, T. Ando, and W. Haensch, “Ohm’s law + kirchhoff’s current law = better ai: Neural-network processing done in memory with analog circuits will save energy,” IEEE Spectrum, vol. 58, no. 12, pp. 44–49, 2021

2021
[14]

A heterogeneous and programmable compute-in-memory accelerator architecture for analog-ai using dense 2-d mesh,

S. Jain, H. Tsai, C.-T. Chen, R. Muralidhar, I. Boybat, M. M. Frank, S. Wo´ zniak, M. Stanisavljevic, P. Adusumilli, P. Narayanan et al. , “A heterogeneous and programmable compute-in-memory accelerator architecture for analog-ai using dense 2-d mesh,” IEEE Transactions on Ver...

2022
[15]

Bottom-up and top-down ap- proaches for the design of neuromorphic processing systems: tradeoffs and synergies between natural and artificial intelligence,

C. Frenkel, D. Bol, and G. Indiveri, “Bottom-up and top-down ap- proaches for the design of neuromorphic processing systems: tradeoffs and synergies between natural and artificial intelligence,” Proceedings of the IEEE , vol. 111, no. 6, pp. 623–652, 2023

2023
[16]

Computational phase- change memory: Beyond von neumann computing,

A. Sebastian, M. Le Gallo, and E. Eleftheriou, “Computational phase- change memory: Beyond von neumann computing,” Journal of Physics D: Applied Physics , vol. 52, no. 44, p. 443002, 2019

2019
[17]

Using the ibm analog in-memory hardware acceleration kit for neural network training and inference,

M. Le Gallo, C. Lammie, J. Büchel, F. Carta, O. Fagbohungbe, C. Mackin, H. Tsai, V . Narayanan, A. Sebastian, K. El Maghraoui et al., “Using the ibm analog in-memory hardware acceleration kit for neural network training and inference,” APL Machine Learning , vol. 1, no. 4, 2023

2023
[18]

Analog-memory- based 14nm hardware accelerator for dense deep neural networks includ- ing transformers,

A. Okazaki, P. Narayanan, S. Ambrogio, K. Hosokawa, H. Tsai, A. No- mura, T. Yasuda, C. Mackin, A. Friz, M. Ishii et al., “Analog-memory- based 14nm hardware accelerator for dense deep neural networks includ- ing transformers,” in 2022 IEEE International Symposium on Circuits ...

2022
[19]

Impact of asymmetric weight update on neural network training with tiki-taka algorithm,

C. Lee, K. Noh, W. Ji, T. Gokmen, and S. Kim, “Impact of asymmetric weight update on neural network training with tiki-taka algorithm,” Frontiers in neuroscience, vol. 15, p. 767953, 2022

2022
[20]

Fast and robust analog in-memory deep neural network training,

M. J. Rasch, F. Carta, O. Fagbohungbe, and T. Gokmen, “Fast and robust analog in-memory deep neural network training,” Nature Communica- tions, vol. 15, no. 1, p. 7133, 2024

2024
[21]

Neural network learning using non-ideal resistive memory devices,

Y . Kim, T. Gokmen, H. Miyazoe, P. Solomon, S. Kim, A. Ray, J. Doevenspeck, R. S. Khan, V . Narayanan, and T. Ando, “Neural network learning using non-ideal resistive memory devices,” Frontiers in Nanotechnology, vol. 4, p. 1008266, 2022

2022
[22]

Training large-scale artificial neural networks on simulated resistive crossbar arrays,

M. J. Rasch, T. Gokmen, and W. Haensch, “Training large-scale artificial neural networks on simulated resistive crossbar arrays,” IEEE Design & Test, vol. 37, no. 2, pp. 19–29, 2019

2019
[23]

Mixed-precision architecture based on computational memory for training deep neural networks,

S. Nandakumar, M. Le Gallo, I. Boybat, B. Rajendran, A. Sebastian, and E. Eleftheriou, “Mixed-precision architecture based on computational memory for training deep neural networks,” in 2018 IEEE International Symposium on Circuits and Systems (ISCAS) . IEEE, 2018, pp. 1–5

2018
[24]

Acceleration of deep neural network training with resistive cross-point devices,

G. Tayfun and Y . Vlasov, “Acceleration of deep neural network training with resistive cross-point devices,” CoRR, 2016

2016
[25]

Algorithm for training neural networks on resistive device arrays,

T. Gokmen and W. Haensch, “Algorithm for training neural networks on resistive device arrays,” Frontiers in neuroscience , vol. 14, p. 103, 2020

2020
[26]

Enabling training of neural networks on noisy hardware,

T. Gokmen, “Enabling training of neural networks on noisy hardware,” Frontiers in Artificial Intelligence , vol. 4, p. 699148, 2021

2021
[27]

Deep learning acceleration in 14nm cmos compatible reram array: device, material and algorithm co-optimization,

N. Gong, M. J. Rasch, S.-C. Seo, A. Gasasira, P. Solomon, V . Bragaglia, S. Consiglio, H. Higuchi, C. Park, K. Brew et al. , “Deep learning acceleration in 14nm cmos compatible reram array: device, material and algorithm co-optimization,” in 2022 International Electron Devices...

2022
[28]

Circuit techniques for reducing the effects of op-amp imperfections: autozeroing, correlated double sampling, and chopper stabilization,

O.-A. Imperfections, “Circuit techniques for reducing the effects of op-amp imperfections: autozeroing, correlated double sampling, and chopper stabilization,” Proceedings of the IEEE , vol. 84, no. 11, 1996

1996
[29]

Swin transformer: Hierarchical vision transformer using shifted windows,

Z. Liu, Y . Lin, Y . Cao, H. Hu, Y . Wei, Z. Zhang, S. Lin, and B. Guo, “Swin transformer: Hierarchical vision transformer using shifted windows,” in Proceedings of the IEEE/CVF international conference on computer vision, 2021, pp. 10 012–10 022

2021
[30]

Attention is all you need,

A. Vaswani, “Attention is all you need,” Advances in Neural Information Processing Systems, 2017

2017
[31]

A survey on vision transformer,

K. Han, Y . Wang, H. Chen, X. Chen, J. Guo, Z. Liu, Y . Tang, A. Xiao, C. Xu, Y . Xuet al., “A survey on vision transformer,” IEEE transactions on pattern analysis and machine intelligence, vol. 45, no. 1, pp. 87–110, 2022

2022
[32]

Learning multiple layers of features from tiny images,

A. Krizhevsky, G. Hinton et al. , “Learning multiple layers of features from tiny images,” 2009

2009
[33]

Impact of l 1 batch normalization on analog noise resistant property of deep learning models,

O. Fagbohungbe and L. Qian, “Impact of l 1 batch normalization on analog noise resistant property of deep learning models,” in 2022 International Joint Conference on Neural Networks (IJCNN) . IEEE, 2022, pp. 1–9

2022

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

A survey on deep transfer learning,

C. Tan, F. Sun, T. Kong, W. Zhang, C. Yang, and C. Liu, “A survey on deep transfer learning,” in Artificial Neural Networks and Machine Learning–ICANN 2018: 27th International Conference on Artificial Neural Networks, Rhodes, Greece, October 4-7, 2018, Proceedings, Part III 27. Springer, 2018, pp. 270–279

work page 2018

[2] [2]

On the opportunities and risks of foundation models,

R. Bommasani, D. A. Hudson, E. Adeli, R. Altman, S. Arora, S. von Arx, M. S. Bernstein, J. Bohg, A. Bosselut, E. Brunskill et al. , “On the opportunities and risks of foundation models,” arXiv preprint arXiv:2108.07258, 2021

arXiv 2021

[3] [3]

Detecting llm-generated text in computing education: Comparative study for chatgpt cases,

M. S. Orenstrakh, O. Karnalim, C. A. Suarez, and M. Liut, “Detecting llm-generated text in computing education: Comparative study for chatgpt cases,” in 2024 IEEE 48th Annual Computers, Software, and Applications Conference (COMPSAC) . IEEE, 2024, pp. 121–126

2024

[4] [4]

Large language models in medicine,

A. J. Thirunavukarasu, D. S. J. Ting, K. Elangovan, L. Gutierrez, T. F. Tan, and D. S. W. Ting, “Large language models in medicine,” Nature medicine, vol. 29, no. 8, pp. 1930–1940, 2023

1930

[5] [5]

Driving with llms: Fusing object- level vector modality for explainable autonomous driving,

L. Chen, O. Sinavski, J. Hünermann, A. Karnsund, A. J. Willmott, D. Birch, D. Maund, and J. Shotton, “Driving with llms: Fusing object- level vector modality for explainable autonomous driving,” in2024 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2024, pp. 14 093–14 100

work page 2024

[6] [6]

Large language models in finance: A survey,

Y . Li, S. Wang, H. Ding, and H. Chen, “Large language models in finance: A survey,” in Proceedings of the fourth ACM international conference on AI in finance , 2023, pp. 374–382

work page 2023

[7] [7]

Emergent abilities of large language models,

J. Wei, Y . Tay, R. Bommasani, C. Raffel, B. Zoph, S. Borgeaud, D. Yogatama, M. Bosma, D. Zhou, D. Metzler et al., “Emergent abilities of large language models,” arXiv preprint arXiv:2206.07682 , 2022

arXiv 2022

[8] [8]

A comparison review of transfer learning and self-supervised learning: Definitions, applications, advantages and limitations,

Z. Zhao, L. Alzubaidi, J. Zhang, Y . Duan, and Y . Gu, “A comparison review of transfer learning and self-supervised learning: Definitions, applications, advantages and limitations,” Expert Systems with Appli- cations, p. 122807, 2023

work page 2023