M. Mahmudul Hasan Sajeeb
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
M. Mahmudul Hasan Sajeeb has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Corentin Delacour
University of California, Santa Barbara
2 shared publications - Kerem Yunus Camsari
Department of Electrical and Computer Engineering University of California 93106 Santa Barbara Santa Barbara CA USA
2 shared publications - Atharva Raut
Department of Electrical and Computer Engineering, Carnegie Mellon University, Pittsburgh, USA
1 shared publication - Cesely Smith
Carnegie Mellon University
1 shared publication - Dhruv Chinmay
Carnegie Mellon University
1 shared publication - Joao Pedro Hespanha
Department of Electrical and Computer Engineering, University of California, Santa Barbara, Santa Barbara, CA 93106, USA
1 shared publication
Latest Publications
- 26 August 2025
- Published by American Physical Society (APS) in Physical Review E
- Vol. 112 (2) , L023301
Two-dimensional parallel tempering for constrained optimization
Sampling Boltzmann probability distributions plays a key role in machine learning and optimization, motivating the design of hardware accelerators such as Ising machines. While the Ising model can, in principle, encode arbitrary optimization problems, practical implementations are often hindered by soft constraints that either slow down mixing when too strong or fail to enforce feasibility when too weak. We introduce a two-dimensional extension of the powerful parallel tempering algorithm (PT) that addresses this challenge by adding a second dimension of replicas interpolating the penalty strengths. This scheme ensures constraint satisfaction in the final replicas, analogous to low-energy states at low temperature. The resulting two-dimensional parallel tempering algorithm (2D-PT) improves mixing in heavily constrained replicas and eliminates the need to explicitly tune the penalty strength. In a representative example of graph sparsification with copy constraints, 2D-PT achieves near-ideal mixing, with Kullback-Leibler divergence decaying as O(1/t). When applied to sparsified Wishart instances, 2D-PT yields orders-of-magnitude speedup over conventional PT with the same number of replicas. The method applies broadly to constrained Ising problems and can be deployed on existing Ising machines.
- 26 August 2025
- Published by American Physical Society (APS) in Physical Review E
- Vol. 112 (2) , L023301
Two-dimensional parallel tempering for constrained optimization
Sampling Boltzmann probability distributions plays a key role in machine learning and optimization, motivating the design of hardware accelerators such as Ising machines. While the Ising model can, in principle, encode arbitrary optimization problems, practical implementations are often hindered by soft constraints that either slow down mixing when too strong or fail to enforce feasibility when too weak. We introduce a two-dimensional extension of the powerful parallel tempering algorithm (PT) that addresses this challenge by adding a second dimension of replicas interpolating the penalty strengths. This scheme ensures constraint satisfaction in the final replicas, analogous to low-energy states at low temperature. The resulting two-dimensional parallel tempering algorithm (2D-PT) improves mixing in heavily constrained replicas and eliminates the need to explicitly tune the penalty strength. In a representative example of graph sparsification with copy constraints, 2D-PT achieves near-ideal mixing, with Kullback-Leibler divergence decaying as O(1/t). When applied to sparsified Wishart instances, 2D-PT yields orders-of-magnitude speedup over conventional PT with the same number of replicas. The method applies broadly to constrained Ising problems and can be deployed on existing Ising machines.
- 2 July 2025
- Published by American Physical Society (APS) in Physical Review Applied
- Vol. 24 (1)
Scalable connectivity for Ising machines: Dense to sparse
- 2 July 2025
- Published by American Physical Society (APS) in Physical Review Applied
- Vol. 24 (1)
Scalable connectivity for Ising machines: Dense to sparse
- 12 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Microwave and Wireless Technology Letters
- Vol. 34 (2) , 191-194
A Widely Tunable Switched-Capacitor Balun Based on N-Path Filters
This letter introduces an inductor-less eight-path balanced-to-unbalanced (balun) bandpass filter that performs single-to-differential conversion by switching the bottom plate of the $N$ -path filter capacitors between ground and the negative terminal of the differential output. Measurement results of a 0.32-mm $^2$ 65-nm CMOS chip demonstrate an insertion loss below 3 dB, power consumption below 12.2 mW, a tunable range of 0.1–1 GHz, a high-selectivity filtering performance with a bandwidth of $\sim$ 30 MHz, and input impedance matching capability. The amplitude (phase) imbalance is less than 0.3 dB ( $\pm{3}^\circ$ ) over the tuning range. The measured in-band IIP $_{3}$ , IIP $_{2}$ , OP $_{1}$ dB, and simulated noise figure are $+$ 13.8 dBm, $+$ 38 dBm, $-$ 4 dBm, and 2.45 dB, respectively, at a center frequency of 500 MHz.
- 12 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Microwave and Wireless Technology Letters
- Vol. 34 (2) , 191-194
A Widely Tunable Switched-Capacitor Balun Based on N-Path Filters
This letter introduces an inductor-less eight-path balanced-to-unbalanced (balun) bandpass filter that performs single-to-differential conversion by switching the bottom plate of the $N$ -path filter capacitors between ground and the negative terminal of the differential output. Measurement results of a 0.32-mm $^2$ 65-nm CMOS chip demonstrate an insertion loss below 3 dB, power consumption below 12.2 mW, a tunable range of 0.1–1 GHz, a high-selectivity filtering performance with a bandwidth of $\sim$ 30 MHz, and input impedance matching capability. The amplitude (phase) imbalance is less than 0.3 dB ( $\pm{3}^\circ$ ) over the tuning range. The measured in-band IIP $_{3}$ , IIP $_{2}$ , OP $_{1}$ dB, and simulated noise figure are $+$ 13.8 dBm, $+$ 38 dBm, $-$ 4 dBm, and 2.45 dB, respectively, at a center frequency of 500 MHz.