arXiv subjects

Forrest Sheldon

Publications and source records attributed to Forrest Sheldon.

6 recordsLinked to original sources

arXiv2023

Characterizing contaminant noise in barcoded perturbation experiments

Bursting cells lead to ambient RNA that contaminates sequencing data. This process is especially problematic in perturbation experiments where transcription factors are implanted into cells to determine their effects. The presence of contaminants makes it difficult to determine whether a factor is truly expressed in the cell. This paper studies the properties of contaminant noise from an analytical perspective, showing that the cell bursting process constrains the form of the noise distribution across factors. These constraints can be leveraged to improve decontamination by removing counts that are more likely the result of noise than expression. In two biological replicates of a perturbation experiment, run across two sequencing protocols, decontaminated counts agree with bulk genomic measurements of the transduction rate and are automatically corrected for differences in sequencing.

q-bio.QM↗

arXiv2021

The Backpropagation Algorithm Implemented on Spiking Neuromorphic Hardware

The capabilities of natural neural systems have inspired new generations of machine learning algorithms as well as neuromorphic very large-scale integrated (VLSI) circuits capable of fast, low-power information processing. However, it has been argued that most modern machine learning algorithms are not neurophysiologically plausible. In particular, the workhorse of modern deep learning, the backpropagation algorithm, has proven difficult to translate to neuromorphic hardware. In this study, we present a neuromorphic, spiking backpropagation algorithm based on synfire-gated dynamical information coordination and processing, implemented on Intel's Loihi neuromorphic research processor. We demonstrate a proof-of-principle three-layer circuit that learns to classify digits from the MNIST dataset. To our knowledge, this is the first work to show a Spiking Neural Network (SNN) implementation of the backpropagation algorithm that is fully on-chip, without a computer in the loop. It is competitive in accuracy with off-chip trained SNNs and achieves an energy-delay product suitable for edge computing. This implementation shows a path for using in-memory, massively parallel neuromorphic processors for low-power, low-latency implementation of modern deep learning applications.

cs.NE↗

arXiv2018

Taming a non-convex landscape with dynamical long-range order: memcomputing Ising benchmarks

Recent work on quantum annealing has emphasized the role of collective behavior in solving optimization problems. By enabling transitions of clusters of variables, such solvers are able to navigate their state space and locate solutions more efficiently despite having only local connections between elements. However, collective behavior is not exclusive to quantum annealers, and classical solvers that display collective dynamics should also possess an advantage in navigating a non-convex landscape. Here, we give evidence that a benchmark derived from quantum annealing studies is solvable in polynomial time using digital memcomputing machines, which utilize a collection of dynamical components with memory to represent the structure of the underlying optimization problem. To illustrate the role of memory and clarify the structure of these solvers we propose a simple model of these machines that demonstrates the emergence of long-range order. This model, when applied to finding the ground state of the Ising frustrated-loop benchmarks, undergoes a transient phase of avalanches which can span the entire lattice and demonstrates a connection between long-range behavior and their probability of success. These results establish the advantages of computational approaches based on collective dynamics of continuous dynamical systems.

cond-mat.dis-nn↗

arXiv2018

Stress-testing memcomputing on hard combinatorial optimization problems

Memcomputing is a novel paradigm of computation that utilizes dynamical elements with memory to both store and process information on the same physical location. Its building blocks can be fabricated in hardware with standard electronic circuits, thus offering a path to its practical realization. In addition, since memcomputing is based on non-quantum elements, the equations of motion describing these machines can be simulated efficiently on standard computers. In fact, it was recently realized that memcomputing, and in particular its digital (hence scalable) version, when simulated on a classical machine provides a significant speed-up over state-of-the-art algorithms on a variety of non-convex problems. Here, we stress-test the capabilities of this approach on finding approximate solutions to hard combinatorial optimization problems. These fall into a class which is known to require exponentially growing resources in the worst cases, even to generate approximations. We recently showed that in a region where state of the art algorithms demonstrate this exponential growth, simulations of digital memcomputing machines performed using the Falcon$^\copyright$ simulator of MemComputing, Inc. only require time and memory resources that scale linearly. These results are extended in a stress-test up to $64\times10^6$ variables (corresponding to about 1 billion literals), namely the largest case that we could fit on a single node with 128 GB of DRAM. Since memcomputing can be applied to a wide variety of optimization problems, this stress test shows the considerable advantage of non-combinatorial, physics-inspired approaches over standard combinatorial ones.

cs.ET↗

arXiv2017

Evidence of an exponential speed-up in the solution of hard optimization problems

Optimization problems pervade essentially every scientific discipline and industry. Many such problems require finding a solution that maximizes the number of constraints satisfied. Often, these problems are particularly difficult to solve because they belong to the NP-hard class, namely algorithms that always find a solution in polynomial time are not known. Over the past decades, research has focused on developing heuristic approaches that attempt to find an approximation to the solution. However, despite numerous research efforts, in many cases even approximations to the optimal solution are hard to find, as the computational time for further refining a candidate solution grows exponentially with input size. Here, we show a non-combinatorial approach to hard optimization problems that achieves an exponential speed-up and finds better approximations than the current state-of-the-art. First, we map the optimization problem into a boolean circuit made of specially designed, self-organizing logic gates, which can be built with (non-quantum) electronic components; the equilibrium points of the circuit represent the approximation to the problem at hand. Then, we solve its associated non-linear ordinary differential equations numerically, towards the equilibrium points. We demonstrate this exponential gain by comparing a sequential MatLab implementation of our solver with the winners of the 2016 Max-SAT competition on a variety of hard optimization instances. We show empirical evidence that our solver scales linearly with the size of the problem, both in time and memory, and argue that this property derives from the collective behavior of the simulated physical circuit. Our approach can be applied to other types of optimization problems and the results presented here have far-reaching consequences in many fields.

cs.AI↗

arXiv2017

Phase-dependent noise in Josephson junctions

In addition to the usual superconducting current, Josephson junctions (JJs) support a phase-dependent conductance related to the retardation effect of tunneling quasi-particles. This introduces a dissipative current with a memory-resistive (memristive) character and thus should also affect the current noise. By means of the microscopic theory of tunnel junctions we compute the complete current autocorrelation function of a Josephson tunnel junction and show that this memristive component gives rise to a non-stationary, phase-dependent noise. As a consequence, dynamic and thermal noise necessarily show a phase dependence otherwise absent in nondissipative JJ models. This phase dependence may be realized experimentally as a hysteresis effect if the unavoidable time averaging of the experimental probe is shorter than the period of the Josephson phase.

cond-mat.mes-hall↗