Threshold logic (TL) circuits gain increasing attention due to their feasible realization with emerging technologies and strong bind to neural network applications. In this work, we devise techniques ...for automatic synthesis and verification of TL circuits based on constraint solving. For synthesis, we formulate a fundamental operation to collapse TL functions, and derive a necessary and sufficient condition of collapsibility for linear combination of two TL functions. An approach based on solving the subset sum problem is proposed for fast circuit transformation. For verification, we propose a procedure to convert a TL function to a multiplexer (MUX) tree and to pseudo-Boolean (PB) constraints for formal Boolean and PB reasoning, respectively. Experiments on synthesis show that the collapse operation further reduces gate counts of synthesized TL circuits by an average of 18%. Experiments on verification demonstrate good scalability of the MUX-based method for equivalence checking of synthesized TL circuits, and efficiency of PB constraint conversion in cases where the conjunctive normal form (CNF) formula conversion and MUX tree conversion suffer from memory explosion.
Pembrolizumab previously demonstrated robust antitumor activity and manageable safety in a phase Ib study of patients with heavily pretreated, programmed death ligand 1 (PD-L1)-positive, recurrent or ...metastatic nasopharyngeal carcinoma (NPC). The phase III KEYNOTE-122 study was conducted to further evaluate pembrolizumab versus chemotherapy in patients with platinum-pretreated, recurrent and/or metastatic NPC. Final analysis results are presented.
KEYNOTE-122 was an open-label, randomized study conducted at 29 sites, globally. Participants with platinum-pretreated recurrent and/or metastatic NPC were randomly assigned (1 : 1) to pembrolizumab or chemotherapy with capecitabine, gemcitabine, or docetaxel. Randomization was stratified by liver metastasis (present versus absent). The primary endpoint was overall survival (OS), analyzed in the intention-to-treat population using the stratified log-rank test (superiority threshold, one-sided P = 0.0187). Safety was assessed in the as-treated population.
Between 5 May 2016 and 28 May 2018, 233 participants were randomly assigned to treatment (pembrolizumab, n = 117; chemotherapy, n = 116); Most participants (86.7%) received study treatment in the second-line or later setting. Median time from randomization to data cut-off (30 November 2020) was 45.1 months (interquartile range, 39.0-48.8 months). Median OS was 17.2 months 95% confidence interval (CI) 11.7-22.9 months with pembrolizumab and 15.3 months (95% CI 10.9-18.1 months) with chemotherapy hazard ratio, 0.90 (95% CI 0.67-1.19; P = 0.2262). Grade 3-5 treatment-related adverse events occurred in 12 of 116 participants (10.3%) with pembrolizumab and 49 of 112 participants (43.8%) with chemotherapy. Three treatment-related deaths occurred: 1 participant (0.9%) with pembrolizumab (pneumonitis) and 2 (1.8%) with chemotherapy (pneumonia, intracranial hemorrhage).
Pembrolizumab did not significantly improve OS compared with chemotherapy in participants with platinum-pretreated recurrent and/or metastatic NPC but did have manageable safety and a lower incidence of treatment-related adverse events.
•No difference was observed in efficacy between pembrolizumab and chemotherapy in advanced platinum-pretreated NPC.•Median OS was 17.2 months with pembrolizumab versus 15.3 with chemotherapy (median PFS, 4.1 versus 5.5 months.•Pembrolizumab had manageable safety and a lower incidence of treatment-related adverse events (AEs) than chemotherapy.•Grade 3-5 treatment-related AEs occurred in 10.3% of participants treated with pembrolizumab versus 43.8% with chemotherapy.
Dependency stochastic Boolean satisfiability (DSSAT), which generalizes stochastic Boolean satisfiability (SSAT) and dependency quantified Boolean formula (DQBF), is a new logical formalism that ...allows compact encoding of NEXPTIME decision problems under uncertainty. Despite potentially broad applications, a decision procedure for DSSAT remains lacking. In this work, we present the first sound and complete resolution calculus for DSSAT. The resolution system deduces the maximum satisfying probability of a DSSAT formula and provides a witnessing certificate. We also show that when the special case of SSAT formulas is considered, the DSSAT resolution calculus p-simulates a known SSAT resolution scheme. Our result may pave a theoretical foundation for further development and certification of DSSAT solvers.
Hardware acceleration enables neural network (NN) inferencing on edge devices and for high throughput applications. Most approaches use neural processing elements for computation while storing ...weights in memory blocks. To avoid costly memory access, recent efforts seek direct logic implementation with weights hardwired into the circuit. However, special training strategies are often needed, and they could not maintain accuracy. In contrast, we take a trained and quantized NN as input and synthesize it by Booth encoding and logic sharing, resulting in a hardware accelerator without degrading accuracy. Experiments demonstrate that our method outperforms existing work in area reduction and/or throughput and power efficiency.
A thermal ion driven bursting instability with rapid frequency chirping, considered as an Alfvénic ion temperature gradient mode, has been observed in plasmas having reactor-relevant temperature in ...the DIII-D tokamak. The modes are excited over a wide spatial range from macroscopic device size to microturbulence size and the perturbation energy propagates across multiple spatial scales. The radial mode structure is able to expand from local to global in ∼0.1 ms and it causes magnetic topology changes in the plasma edge, which can lead to a minor disruption event. Since the mode is typically observed in the high ion temperature ≳ 10 keV and high-β plasma regime, the manifestation of the mode in future reactors should be studied with development of mitigation strategies, if needed. This is the first observation of destabilization of the Alfvén continuum caused by the compressibility of ions with reactor-relevant ion temperature.
Circuit learning has gained significant attention due to machine learning advancements and approximate synthesis applications. The task is to learn a circuit to model an unknown Boolean function ...subject to different design constraints. When circuit size is hard-constrained, decision-tree-based learning plays a crucial role in state-of-the-art methods. However, it can be ineffective due to its structural restriction. This work proposes graph learning to overcome the limitation, provide trade-offs between circuit size and accuracy, and enrich the portfolio of circuit learning tools. Experimental results show the superiority of our approach to prior work in accuracy, training time, and circuit size.
Neural networks (NNs) are key to deep learning systems. Their efficient hardware implementation is crucial to applications at the edge. Binarized NNs (BNNs), where the weights and output of a neuron ...are of binary values {−1,+1} (or encoded in {0, 1}), have been proposed. As no multiplier required, BNNs are particularly attractive and suitable for hardware realization. Most prior NN synthesis methods target on hardware architectures with neural processing elements (NPEs), where the weights of a neuron are loaded and the output of the neuron is computed. The load-and-compute method, though area efficient, requires expensive memory access, which deteriorates energy and performance efficiency. In this work we aim at synthesizing BNN layers into dedicated logic circuits. We formulate the corresponding model pruning problem and matrix covering problem to reduce the area and routing cost of BNNs. For model pruning, we propose and compare three strategies at the BNN training stage. For matrix covering, we propose a scalable logic-sharing algorithm. By combining these two methods, experimental results justify the effectiveness of the method in terms of area and net savings on FPGA implementation. Our method provides an alternative implementation of BNNs, and can be applied in combination with NPE-based implementation for area, speed, and power tradeoffs.
TMPRSS4 is a novel type II transmembrane serine protease found at the cell surface that is highly expressed in pancreatic, colon and gastric cancer tissues. However, the biological functions of ...TMPRSS4 in cancer are unknown. Here we show, using reverse transcription-PCR, that TMPRSS4 is highly elevated in lung cancer tissues compared with normal tissues and is also broadly expressed in a variety of human cancer cell lines. Knockdown of TMPRSS4 by small interfering RNA treatment in lung and colon cancer cell lines was associated with reduction of cell invasion and cell-matrix adhesion as well as modulation of cell proliferation. Conversely, the invasiveness, motility and adhesiveness of SW480 colon carcinoma cells were significantly enhanced by TMPRSS4 overexpression. Furthermore, overexpression of TMPRSS4 induced loss of E-cadherin-mediated cell-cell adhesion, concomitant with the induction of SIP1/ZEB2, an E-cadherin transcriptional repressor, and led to epithelial-mesenchymal transition events, including morphological changes, actin reorganization and upregulation of mesenchymal markers. TMPRSS4-overexpressing cells also displayed markedly increased metastasis to the liver in nude mice upon intrasplenic injection. Taken together, these studies suggest that TMPRSS4 controls the invasive and metastatic potential of human cancer cells by facilitating an epithelial-mesenchymal transition; TMPRSS4 may be a potential therapeutic target for cancer treatment.
A long-standing question in the field of tumor immunotherapy is how Th2 cytokines block tumor growth. Their antitumor effects are particularly prominent when they are secreted continuously in tumors, ...suggesting that Th2 cytokines may create a tumor microenvironment unfavorable for tumor growth independently of adaptive immunity. In this study, we show that local production of IL-33 establishes a high number of type 2 innate lymphoid cells (ILC2s) with potent antitumor activity. IL-33 promotes secretion of a massive amount of CXCR2 ligands from ILC2s but creates a tumor microenvironment where tumor cells express CXCR2 through a dysfunctional angiogenesis/hypoxia/reactive oxygen species axis. These two signaling events converge to reinforce tumor cell-specific apoptosis through CXCR2. Our results identify a previously unrecognized antitumor therapeutic pathway wherein ILC2s play a central role.
Quantified Boolean formulae (QBF) allow compact encoding of many decision problems. Their importance motivated the development of fast QBF solvers. Certifying the results of a QBF solver not only ...ensures correctness, but also enables certain synthesis and verification tasks. To date the certificate of a true formula can be in the form of either a syntactic cube-resolution proof or a semantic Skolem-function model whereas that of a false formula is only in the form of a syntactic clause-resolution proof. The semantic certificate for a false QBF is missing, and the syntactic and semantic certificates are somewhat unrelated. This paper identifies the missing Herbrand-function countermodel for false QBF, and strengthens the connection between syntactic and semantic certificates by showing that, given a true QBF, its Skolem-function model is derivable from its cube-resolution proof of satisfiability as well as from its clause-resolution proof of unsatisfiability under formula negation. Consequently Skolem-function derivation can be decoupled from special Skolemization-based solvers and computed from standard search-based ones. Experimental results show strong benefits of the new method.