Parallelized Linear Classification with Volumetric Chemical Perceptrons

Abstract

In this work, we introduce a new type of linear classifier that is implemented in a chemical form. We propose a novel encoding technique which simultaneously represents multiple datasets in an array of microliter-scale chemical mixtures. Parallel computations on these datasets are performed as robotic liquid handling sequences, whose outputs are analyzed by highperformance liquid chromatography. As a proof of concept, we chemically encode several MNIST images of handwritten digits and demonstrate successful chemical-domain classification of the digits using volumetric perceptrons. We additionally quantify the performance of our method with a larger dataset of binary vectors and compare the experimental measurements against predicted results. Paired with appropriate chemical analysis tools, our approach can work on increasingly parallel datasets. We anticipate that related approaches will be scalable to multilayer neural networks and other more complex algorithms. Much like recent demonstrations of archival data storage in DNA, this work blurs the line between chemical and electrical information systems, and offers early insight into the computational efficiency and massive parallelism which may come with computing in chemical domains.

Multicomponent molecular memory

“…The composition of a chemical sample can represent abstract information, whether the sample consists of a single compound selected from a defined chemical space 17 , a pool of sequence-controlled polymers 4,13 , or a mixture of unique compounds 15 . With small molecule libraries, the most direct way to encode information is to use the presence or absence of each library element in a sample to represent one bit of data 14,39 . Thus, our 1500 compound Ugi libraries could encode up to 1500 bits of information per mixture.…”

Principles of Information Storage in Small-Molecule Mixtures

“…Regardless of the types of molecules, the identities and concentrations of molecules within a mixture can serve as atomic-scale representations of abstract digital data. We have demonstrated these ideas experimentally using several families of small molecules, including the demonstrations in Figure 6 and Figure 7, as well as other datasets using phenols [2], metabolites [28], and multi-component reaction products [3]. These experiments have significant room for growth, using error correcting codes and expanded chemical libraries.…”

“…Modern information technology is moving towards a more unified vision of computation and memory, and fluid molecular mixtures offer an intriguing space for future generations of computing systems that take advantage of the natural complexity and intrinsic statistics of chemical systems [2,10,26,27,39,46]. More precisely quantifying the information capacity of chemical mixtures represents an early step in this direction, and we anticipate that valuable scientific advances may come from using this lens to consider pathways within mixtures of reactive chemical libraries.…”

“…Although few chemistries are as mature as those available for DNA, we show that diversified small-molecule approaches have intrinsic capacities for gigabyte-scale data storage. In addition to new experimental [2,3,28] and theoretical tools for interrogating heterogeneous mixtures of molecules, this new perspective may also contribute to new ways of quantifying the information contained in the chemical states of living systems. Figure 1.…”

Encoding Information in Synthetic Metabolomes

“…The presence or absence of one metabolite in one spot encodes one bit of information. Therefore, the total number of bits stored in one spot is equal to the number of available library elements [27].…”