Noisy computation and reversible computation have been studied separately, and it is known that they are as powerful as unrestricted computation. We study the case where both noise and reversibility are combined and show that the combined model is weaker than unrestricted computation. In our noisy reversible circuits, each wire is flipped with probability p each time step, and all the inputs to the circuit are present in time 0. We prove that any noisy reversible circuit must have size exponential in its depth in order to compute a function with high probability. This is tight as we show that any circuit can be converted into a noise-resistant reversible one with a blow up in size which is exponential in the depth. This establishes that noisy reversible computation has the power of the complexity class NC^1. We extend this to quantum circuits(QC). We prove that any noisy QC which is not worthless, and for which all inputs are present at time 0, must have size exponential in its depth. (This high-lights the fact that fault tolerant QC must use a constant supply of inputs all the time.) For the lower bound, we show that quasi-polynomial noisy QC are at least powerful as logarithmic depth QC, (or QNC^1). Making these bounds tight is left open in the quantum case.

We construct a polynomial-time classical algorithm that samples from the output distribution of noisy geometrically local Clifford circuits with any product-state input and single-qubit measurements in any basis. Our results apply to circuits with nearest-neighbor gates on an \(O(1)\text{-D}\) architecture with depolarizing noise after each gate. Importantly, we assume that the circuit does not contain qubit resets or mid-circuit measurements. This class of circuits includes Clifford-magic circuits and Conjugated-Clifford circuits, which are important candidates for demonstrating quantum advantage using non-universal gates. Additionally, our results can be extended to the case of IQP circuits augmented with CNOT gates, which is another class of non-universal circuits that are relevant to current experiments. Importantly, these results do not require randomness assumptions over the circuit families considered (such as anticoncentration properties) and instead hold for every circuit in each class as long as the depth is above a constant threshold. This allows us to rule out the possibility of fault-tolerance in these circuit models. As a key technical step, we prove that interspersed noise causes a decay of long-range entanglement at depths beyond a critical threshold. To prove our results, we merge techniques from percolation theory and Pauli path analysis.

Wearable devices have strict power and memory limitations. As a result, there is a need to optimize the power consumption on those devices without sacrificing the accuracy. This paper presents AdaSense: a sensing, feature extraction and classification co-optimized framework for Human Activity Recognition. The proposed techniques reduce the power consumption by dynamically switching among different sensor configurations as a function of the user activity. The framework selects configurations that represent the pareto-frontier of the accuracy and energy trade-off. AdaSense also uses low-overhead processing and classification methodologies. The introduced approach achieves 69% reduction in the power consumption of the sensor with less than 1.5% decrease in the activity recognition accuracy.