Pith. sign in

REVIEW 4 major objections 7 minor 26 references

M3: Mamba-assisted Multi-Circuit Optimization via MBRL with Effective Scheduling

T0 review · 4 major / 7 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read M3 claims that a single Mamba-based RL agent with scheduled model-based training can concurrently optimize multiple analog circuits of different topologies and specifications in roughly 10,000 simulator calls.

desk verdict Real new combination (Mamba + scheduled MBRL for multi-circuit analog sizing), but the 10x sample-efficiency claim overstates reliability because it rests on mean-based criteria that a minority of successful seeds can satisfy. read the letter →

arxiv 2411.16019 v1 pith:OVLBPPPO submitted 2024-11-25 cs.LG

Youngmin Oh , Jinje Park , Seunggeun Kim , Taejin Paik , David Pan , Bosun Hwang This is my paper · ORCID
classification cs.LG MSC 68T0768T05
keywords Mambamodel-basedreinforcementlearninganalogcircuitoptimizationtransistorsizingmulti-tasksampleefficiencyeffectiveschedulingstatespacemodel
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

M3 aims to show that a single reinforcement-learning agent can optimize several analog circuits at once—circuits with different topologies, parameter counts, and target specifications—rather than requiring a bespoke agent per circuit. The paper proposes combining the Mamba sequence architecture (used for the actor, critic, and a learned neural simulator) with a model-based RL loop whose three key training parameters are scheduled: the number of rollouts, the fraction of real versus synthetic data in each batch, and the number of agent update iterations per environment step. This scheduling starts synthetic-heavy for exploration and gradually shifts to real-data-heavy fine-tuning. If correct, the result matters because analog circuit sizing is dominated by expensive simulator calls, and the paper reports positive mean episode rewards for all four test circuits at 10,500 environment steps, roughly 10 times more sample-efficient than the best non-M3 baseline.

What carries the argument

The load-bearing component is the Mamba architecture used in all three networks—actor, critic, and environment model—where variable-length observations and actions from different circuits are padded to a single token sequence and read out through the final context vector. Around it, the effective scheduling mechanism adjusts three model-based RL hyperparameters as functions of environment step $t$: the rollout number $R(t)$, the real-synthetic data ratio $\alpha(t)$, and the actor-critic update iterations $T_a(t)$, each interpolated linearly between an initial and final value and then clipped. The reward is the figure-of-merit $\mathrm{FoM} = \sum_i \min\{d(m_{c,i}, n_{c,i}), 0\}$ with a sparse bonus of 10 when all normalized gaps are within tolerance. The scheduling is what converts a large volume of early, possibly inaccurate synthetic experience into exploration, then shifts to real data for exploitation.

What would settle it

Run the identical four-circuit benchmark with the same hyperparameters but require a stricter success criterion, e.g., that each of ten seeds reaches a positive episode reward for every circuit; or replicate with a different random ordering of the padded observation dimensions. If the reported 10,500-step positive mean reward is not reproduced within, say, 50,000 environment steps under either change, the central multi-circuit claim would be disproved.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that Mamba-assisted model-based RL with effective scheduling—called M3—is the first framework to solve multi-circuit optimization in a single online learning run. With four benchmark circuits (two-stage op-amp, its complementary variant, transimpedance amplifier, and comparator) that differ in target-specification count, parameter type, and dimensionality, M3 obtains a positive mean episode reward for all circuits before 12,000 environment steps (precisely at 10,500 steps), a criterion no non-Mamba baseline met within 200,000 steps. Against the closest competitor, SAC with Mamba, M3 reaches the criterion about 10 times faster, and in the reported success-rate table M3 trained for 15,000 steps is comparable or better than SAC with Mamba trained for 150,000 steps on three of four circuits. The authors attribute the gain to two mechanisms working together: Mamba's linear-time state-space modeling, which they argue aligns naturally with the Markov structure of circuit simulation, and the scheduled increase of rollout count, real-data ratio, and update iterations.

Load-bearing premise

The load-bearing premise is that, after padding the different circuits' observations and actions into one token sequence, the Mamba network's final context vector captures enough of each circuit's identity and target specification that a single policy trained this way transfers across all four circuits.

Editorial extensions

If this is right

  • If the result holds, analog circuit designers no longer need to train a separate RL agent for each topology; one Mamba-based agent can handle a set of circuits with different parameter and specification dimensions.
  • The reported ~10,000 simulator calls to reach positive mean reward across all circuits makes online MBRL practical for circuit sizing, where each simulation is the main computational bottleneck.
  • Mamba's linear-time inference (about 5.1 ms versus 44.1 ms for the transformer) means that, at equal or better sample efficiency, the wall-clock cost per environment step is an order of magnitude lower, favoring sequence models over attention for this task.
  • The scheduling principle—start with synthetic-heavy batches and more rollouts only later, then fine-tune with real data—is stated generally enough to apply to other multi-task MBRL problems where simulator calls are expensive.
  • A single shared network also removes the memory and bookkeeping overhead of managing separate per-circuit policies, which is the practical motivation the paper gives for multi-circuit optimization.

Reading between the lines

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

  • The paper's headline claim is about mean episode reward averaged over seeds; the per-run success rates in its Table IV are 5/10 to 10/10 per circuit at 15,000 steps, so the honest reading is that the agent typically—not always—sizes each circuit within the reported budget.
  • The padded-input representation is never stress-tested: no ablation permutes the ordering of padded dimensions or masks unused entries, so a natural follow-up is to test order invariance and compare against a graph-based encoder that explicitly knows each circuit's topology.
  • The authors hypothesize, but do not demonstrate, that Mamba's state-space model matches the MDP structure of circuit simulation; a matched-parameter-count ablation against an LSTM or a gated transformer would separate the architecture's inductive bias from the scheduling's contribution.
  • Because the scheduling schedule is hand-set (initial/final values and a scale of 15,000 steps), an adaptive version that schedules from model uncertainty or real-data error could further cut the simulator budget; the paper does not explore this.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 7 minor

Summary. The paper proposes M3, a model-based RL method that combines the Mamba sequence architecture with an effective scheduling scheme (real-synthetic data ratio, rollout count, and update iteration number) to optimize multiple analog circuits with a single agent. The authors evaluate M3 on four circuits (2SOA, R2SOA, 2STIA, Comp) against SAC/PPO variants with Mamba, Transformer, LSTM, and fully connected architectures. They report that M3 achieves positive mean episode rewards for all circuits at 10,500 environment steps and claim roughly 10x sample efficiency over SAC with Mamba, with per-seed success rates at 15,000 steps shown in Table IV.

Significance. If substantiated, the claimed result—a single MBRL agent handling multiple analog circuits with distinct parameters and target specifications using only O(10^4) simulator calls—would be a meaningful step for analog circuit optimization. The paper's strengths include a clear pseudocode (Algorithm 1), a transparent hyperparameter table (Table I), explicit inference-time measurements for Mamba versus Transformer, and a comparison set spanning several RL backbones and architectures. However, the evidence as presented does not yet fully support the headline claims: the success criteria are mean-based and can be satisfied with low per-seed success rates, the 10x sample-efficiency figure is contradicted in part by Table IV, no significance tests are provided, and the scheduling contribution is not isolated by ablations. The paper also does not release code or raw data, limiting independent verification.

major comments (4)
  1. [Section IV-B, Table III and Eq. (1)] The success criteria 'Mean Episode Reward >= 0' and 'Mean Episode Length <= 25' are averages over episodes and seeds. Since Eq. (1) gives +10 when FoM >= -0.02 and FoM (which is <= -0.02) otherwise, a mean reward of zero can be achieved with only a small fraction of successful episodes (e.g., around 5-10% if typical failures have FoM between -0.1 and -0.5). Thus the claim that M3 'successfully obtains positive mean episode rewards for all circuits before 12,000 steps' does not establish reliable multi-circuit optimization. Please report per-seed success rates at the 10,500-step milestone and justify the threshold criteria against a stricter definition such as a minimum fraction of successful episodes per seed.
  2. [Section IV-B, Table IV] The per-seed success rates in Table IV show M3 after 15,000 steps totals 32/40 (7/10, 5/10, 10/10, 10/10), identical to SAC with Mamba after 150,000 steps (5/10, 7/10, 10/10, 10/10); M3 is actually worse on R2SOA (5/10 versus 7/10). Therefore the sentence 'M3 outperforms SAC-Mamba by about 10x in sample efficiency' overstates the comparison, because the 10x figure comes only from the mean-based criteria in Table III. Please provide a matched comparison, e.g., environment steps required to achieve a fixed per-seed success rate, and include significance tests (bootstrap or Mann-Whitney across the 10 seeds) for the milestone times.
  3. [Section III-C and Q2] The paper argues that the effective scheduling contributes to the improvement, but the only comparison is against MBRL with fixed values (real-synthetic ratio 0.05, rollouts 10, update iterations 20). There is no ablation isolating the effects of alpha(t), R(t), and Ta(t) individually. Without such ablations, the claimed benefit of the scheduling mechanism over the Mamba architecture itself is not established. Please report experiments that vary one schedule component at a time, or otherwise quantify the contribution of each schedule.
  4. [Section III-B] The multi-circuit representation pads variable-length observations and actions into a single token sequence and uses Mamba's final context vector for the actor, critic, and model outputs. This assumes that the padded representation, without masking or explicit topology encoding, carries enough information for a single policy to distinguish circuits and learn appropriate actions for each. No analysis of padding positions, sequence order, or masking is provided. A failure of this representation would invalidate the multi-circuit claim even if the scheduling works. Please include ablations such as randomizing the padding order, using a mask to ignore padding, or adding a one-hot circuit identifier, or provide direct evidence that the context vector encodes circuit identity.
minor comments (7)
  1. [Section III-C] The phrase 'real-synthetic dat ratio' contains a typo; it should read 'real-synthetic data ratio'.
  2. [Section IV-B] The text says 'All other parameters are same with those listed in Table III' but the hyperparameters are listed in Table I, not Table III.
  3. [Table I] The row 'RI, RI 1, 7' appears to have a typo; based on the text, it should be 'RI, RF 1, 7'.
  4. [Section IV-A] The circuit list in the text says 'C2SOA' but the rest of the paper and Table II use 'R2SOA'. Please harmonize the notation.
  5. [Figure 5 caption] The caption says 'learning curves of episode lengths and success rates' but the plots show episode lengths and episode rewards; please correct.
  6. [Table III] The caption says 'The first environment steps satisfying some criteria till 200,000 environment steps' but does not state whether the reported steps are an average over seeds, the median, or from a single representative seed. Clarify this.
  7. [Section II] The references [8] and [9] both relate to model-based RL scheduling; the sentence 'inspired by [8, 9]' is clear, but consider citing [8] and [9] separately in Section III-C where their ideas are used.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: M3's empirical claims rest on its own experiments, and its self-citations appear only as prior-work context, not as load-bearing reductions.

full rationale

The paper does not derive its headline results from its own defining assumptions or from self-citations. The observation, action, and reward definitions are adopted from prior work [1, 23] as standard MDP design choices, not as predictions generated from the model. The effective scheduling formulas (3)-(5) are explicit interpolation schedules referenced to [8, 9], and their effect is measured empirically against fixed-schedule MBRL. The self-citations CRoNuS [23] and Insight [25] are mentioned as related work and as sources of the reward/observation formulation, but the central multi-circuit Mamba-based MBRL claim is validated by direct benchmark comparison in Tables III and IV and Figures 4-5, with no equation reducing to a fitted parameter renamed as a prediction. Any concern that the Table III success criteria are statistically lenient is a correctness or evaluation-strength issue, not a circularity issue. The derivation chain, such as it is, is self-contained: the architecture, scheduling, and baselines are compared empirically rather than inferred from the cited works.

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

M3 does not invent new physical entities; its contributions are a network architecture choice (Mamba), an MBRL training schedule, and the multi-circuit problem setup. The schedule parameters are free choices, and the padded-sequence representation is an untested assumption.

free parameters (4)
  • real-synthetic data ratio schedule (alpha_I, alpha_F, scale) = 0.05 to 0.95, scale=15,000
    The core proposed scheduling mechanism uses hand-chosen endpoints and a hand-chosen scale; no sensitivity analysis or data-driven selection is reported.
  • rollout number schedule (R_I, R_F, scale) = 1 to 7, scale=15,000
    The paper sets initial and final rollout counts by hand; the fixed MBRL baseline uses 10 rollouts, so the comparison mixes scheduling with different initial values.
  • agent update iterations schedule (Ta_I, Ta_F, scale) = 15 to 20, scale=15,000
    Hand-chosen small range around the fixed baseline's 20 updates; the isolated effect is not tested.
  • Mamba architecture dimensions = state/model 64/16, conv 4/2
    The model size is chosen without ablations; performance could depend on this specific capacity.
assumptions (4)
  • domain assumption Mamba is an autoregressive state-space model that can serve as actor, critic, and environment model for RL, with the properties claimed in [5,6].
    The method relies on Mamba's linear-time behavior and ability to produce a useful final context vector; no formal verification or independent reproduction is given.
  • ad hoc to paper Environment resets to a random circuit topology, technology node, and target specifications, and resets when reward is positive or episode length exceeds Tep.
    This reset rule, stated in Section III-C.1, defines the multi-task setup that the central claim depends on; it is constructed for this paper, not established as standard.
  • ad hoc to paper Padding variable-length circuit observations and actions into a single sequence and using the final context vector is a sufficient representation for multi-circuit learning.
    Section III-B Implementation introduces padding with no masking or sequence-order analysis; this is a load-bearing modeling assumption.
  • domain assumption The linear schedules in equations (3), (4), and (5) with scale=15,000 implement the intended exploration-exploitation trade-off.
    The schedules are posited based on prior work [8,9] and not derived; their success is an empirical claim.

how reviews work

0 comments
Cite this review

Pith. "Pith review of M3: Mamba-assisted Multi-Circuit Optimization via MBRL with Effective Scheduling." pith.science (2026). https://pith.science/paper/OVLBPPPO

@misc{pith2026241116019,
  author       = {Pith},
  title        = {Pith review of: M3: Mamba-assisted Multi-Circuit Optimization via MBRL with Effective Scheduling},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OVLBPPPO}},
  note         = {Machine review of arXiv:2411.16019}
}
read the original abstract

Recent advancements in reinforcement learning (RL) for analog circuit optimization have demonstrated significant potential for improving sample efficiency and generalization across diverse circuit topologies and target specifications. However, there are challenges such as high computational overhead, the need for bespoke models for each circuit. To address them, we propose M3, a novel Model-based RL (MBRL) method employing the Mamba architecture and effective scheduling. The Mamba architecture, known as a strong alternative to the transformer architecture, enables multi-circuit optimization with distinct parameters and target specifications. The effective scheduling strategy enhances sample efficiency by adjusting crucial MBRL training parameters. To the best of our knowledge, M3 is the first method for multi-circuit optimization by leveraging both the Mamba architecture and a MBRL with effective scheduling. As a result, it significantly improves sample efficiency compared to existing RL methods.

Figures

Figures reproduced from arXiv: 2411.16019 by the authors.

Figure 1
Figure 1. M3 overview: The Mamba architecture is utilized for both the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Architecture of Neural Networks. o is output size, i.e., the dimension of outputs, so that it depends on the objective of neural networks; for instance, o = 1 for critic (Q-network) since Q-values are real-valued. Algorithm 1 PseudoCode for M3 Require: Set environment E, Ntotal, Nelite, b, αI , αF , NI , Tmodel, Ta,I , Ta,F , Tmax, RI , RF , Tro 1: Initialize actor πθ, critic Qθ, and the ensemble of model networks {… view at source ↗
Figure 3
Figure 3. Benchmark Circuits. (a) Two-stage operational amplifier (2SOA), (b) Two-stage operational amplifier exchanging p-mos and n-mos transistors each other [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Learning curves of episode lengths and episode rewards. Both MT and ST denote multi- and single-task learning, respectively. The solid line and shaded [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: Learning curves of episode lengths and episode rewards. M3 successfully obtains positive mean episode rewards for all circuits before 12,000 steps [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 16 canonical work pages

  1. [1]

    Autockt: Deep reinforcement learning of analog circuit designs,

    K. Settaluri, A. Haj-Ali, Q. Huang, K. Hakhamaneshi, and B. Nikolic, “Autockt: Deep reinforcement learning of analog circuit designs,” in 2020 Design, Automation & Test in Europe Conference & Exhibition (DATE) . IEEE, 2020, pp. 490–495

    work page 2020

  2. [2]

    Gcn-rl circuit designer: Transferable transis- tor sizing with graph neural networks and reinforcement learning,

    H. Wang, K. Wang, J. Yang, L. Shen, N. Sun, H.-S. Lee, and S. Han, “Gcn-rl circuit designer: Transferable transis- tor sizing with graph neural networks and reinforcement learning,” in 2020 57th ACM/IEEE Design Automation Conference (DAC). IEEE, 2020, pp. 1–6

    work page 2020

  3. [3]

    Rose: Robust analog circuit parameter optimization with sampling-efficient re- inforcement learning,

    G. Jian, C. Weidong, and Z. Xuan, “Rose: Robust analog circuit parameter optimization with sampling-efficient re- inforcement learning,” in Design Automation Conference , 2023

    work page 2023

  4. [4]

    Automated design of complex analog circuits with mul- tiagent based reinforcement learning,

    Z. Jinxin, B. Jiarui, H. Zhangcheng, Z. Xuan, and L. Ye, “Automated design of complex analog circuits with mul- tiagent based reinforcement learning,” in Design Automa- tion Conference, 2023

    work page 2023

  5. [5]

    Mamba: Linear-time sequence modeling with selective state spaces,

    A. Gu and T. Dao, “Mamba: Linear-time sequence modeling with selective state spaces,” arXiv preprint arXiv:2312.00752, 2023

    arXiv 2023

  6. [6]

    Transformers are ssms: Generalized models and efficient algorithms through structured state space duality,

    T. Dao and A. Gu, “Transformers are ssms: Generalized models and efficient algorithms through structured state space duality,” arXiv preprint arXiv:2405.21060 , 2024

    arXiv 2024

  7. [7]

    Meta-world: A benchmark and evaluation for multi-task and meta reinforcement learning,

    T. Yu, D. Quillen, Z. He, R. Julian, K. Hausman, C. Finn, and S. Levine, “Meta-world: A benchmark and evaluation for multi-task and meta reinforcement learning,” in Con- ference on robot learning . PMLR, 2020, pp. 1094–1100

    work page 2020

  8. [8]

    On effective scheduling of model-based reinforcement learning,

    H. Lai, J. Shen, W. Zhang, Y . Huang, X. Zhang, R. Tang, Y . Yu, and Z. Li, “On effective scheduling of model-based reinforcement learning,” Advances in Neural Information Processing Systems, vol. 34, pp. 3694–3705, 2021

    work page 2021

Show all 26 references
  1. [10]

    Jamba: A hybrid transformer-mamba language model,

    O. Lieber, B. Lenz, H. Bata, G. Cohen, J. Osin, I. Dalmedigos, E. Safahi, S. Meirom, Y . Belinkov, S. Shalev-Shwartz et al. , “Jamba: A hybrid transformer-mamba language model,” arXiv preprint arXiv:2403.19887, 2024

    2024 arXiv

  2. [11]

    Llama 2: Open foundation and fine-tuned chat models,

    H. Touvron, L. Martin, K. Stone, P. Albert, A. Almahairi, Y . Babaei, N. Bashlykov, S. Batra, P. Bhargava, S. Bhosale et al. , “Llama 2: Open foundation and fine-tuned chat models,” arXiv preprint arXiv:2307.09288 , 2023

    2023 arXiv

  3. [12]

    In- tegrating mamba and transformer for long-short range time series forecasting,

    X. Xu, Y . Liang, B. Huang, Z. Lan, and K. Shu, “In- tegrating mamba and transformer for long-short range time series forecasting,” arXiv preprint arXiv:2404.14757, 2024

    2024

  4. [13]

    Vision mamba: Efficient visual representa- tion learning with bidirectional state space model,

    L. Zhu, B. Liao, Q. Zhang, X. Wang, W. Liu, and X. Wang, “Vision mamba: Efficient visual representa- tion learning with bidirectional state space model,” arXiv preprint arXiv:2401.09417, 2024

    2024 arXiv

  5. [14]

    A survey of mamba,

    H. Qu, L. Ning, R. An, W. Fan, T. Derr, X. Xu, and Q. Li, “A survey of mamba,” arXiv preprint arXiv:2408.01129 , 2024

    2024 arXiv

  6. [15]

    Proximal policy optimization algorithms,

    J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Proximal policy optimization algorithms,” arXiv preprint arXiv:1707.06347 , 2017

    2017 arXiv

  7. [16]

    Soft actor-critic: Off-policy maximum entropy deep reinforce- ment learning with a stochastic actor,

    T. Haarnoja, A. Zhou, P. Abbeel, and S. Levine, “Soft actor-critic: Off-policy maximum entropy deep reinforce- ment learning with a stochastic actor,” in International conference on machine learning. PMLR, 2018, pp. 1861– 1870

    2018

  8. [17]

    When to trust your model: Model-based policy optimization,

    M. Janner, J. Fu, M. Zhang, and S. Levine, “When to trust your model: Model-based policy optimization,” Advances in Neural Information Processing Systems , vol. 32, 2019

    2019

  9. [18]

    Bidirectional model-based policy optimization,

    H. Lai, J. Shen, W. Zhang, and Y . Yu, “Bidirectional model-based policy optimization,” in International Con- ference on Machine Learning . PMLR, 2020, pp. 5618– 5627

    2020

  10. [19]

    Trust the model when it is confident: Masked model-based actor-critic,

    F. Pan, J. He, D. Tu, and Q. He, “Trust the model when it is confident: Masked model-based actor-critic,” Advances in neural information processing systems , vol. 33, pp. 10 537–10 546, 2020

    2020

  11. [20]

    Pvtsizing: A turbo-rl-based batch-sampling optimization framework for pvt-robust analog circuit synthesis,

    Z. Kong, X. Tang, W. Shi, Y . Du, Y . Lin, and Y . Wang, “Pvtsizing: A turbo-rl-based batch-sampling optimization framework for pvt-robust analog circuit synthesis,” in 2024, 61st ACM/IEEE Design Automation Conference (DAC), 2024

    2024

  12. [21]

    Parasitic-aware analog circuit sizing with graph neural networks and bayesian optimization,

    M. Liu, W. J. Turner, G. F. Kokai, B. Khailany, D. Z. Pan, and H. Ren, “Parasitic-aware analog circuit sizing with graph neural networks and bayesian optimization,” in 2021 Design, Automation & Test in Europe Conference & Exhibition (DATE) . IEEE, 2021, pp. 1372–1377

    2021

  13. [22]

    Data is all you need: Finetuning llms for chip design via an au- tomated design-data augmentation framework,

    K. Chang, K. Wang, N. Yang, Y . Wang, D. Jin, W. Zhu, Z. Chen, C. Li, H. Yan, Y . Zhou et al. , “Data is all you need: Finetuning llms for chip design via an au- tomated design-data augmentation framework,” in 2024, 61st ACM/IEEE Design Automation Conference (DAC) , 2024

    2024

  14. [23]

    Cronus: Circuit rapid optimization with neural simulator,

    Y . Oh, D. Kim, Y . H. Lee, and B. Hwang, “Cronus: Circuit rapid optimization with neural simulator,” in 2024 Design, Automation & Test in Europe Conference & Exhibition (DATE). IEEE, 2024, pp. 1–6

    2024

  15. [24]

    Using probabilistic model rollouts to boost the sample efficiency of rein- forcement learning for automated analog circuit sizing,

    M. Ahmadzadeh and G. Gielen, “Using probabilistic model rollouts to boost the sample efficiency of rein- forcement learning for automated analog circuit sizing,” in 2024, 61st ACM/IEEE Design Automation Conference (DAC), 2024

    2024

  16. [25]

    Insight: Universal neural simulator for ana- log circuits harnessing autoregressive transformers,

    S. Poddar, Y . Oh, Y . Lai, H. Zhu, B. Hwang, and D. Z. Pan, “Insight: Universal neural simulator for ana- log circuits harnessing autoregressive transformers,” arXiv preprint arXiv:2407.07346, 2024

    2024 arXiv

  17. [26]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” 2017. [Online]. Available: https: //arxiv.org/abs/1412.6980

    2017 arXiv

  18. [27]

    Spinning Up in Deep Reinforcement Learn- ing,

    J. Achiam, “Spinning Up in Deep Reinforcement Learn- ing,” 2018

    2018

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.