Utilizing light propagation and optical nonlinearities is one of the strategies to accelerate computational tasks in tandem with electrical circuits in an energy-efficient manner. Computing with multimode optical fibers has been demonstrated to be energy efficient due to the high light confinement and multidimensionality. However, these optical nonlinearities have not been programmed for a specific computational task and thus the performance is not optimal. In this study, we demonstrate that the nonlinear transformation in the fiber can be programmed to obtain improved performances on several different machine learning tasks by shaping the wavefront of the information encoding beam.

The emerging fields of on-chip photonic data processing, neuromorphic computing, and quantum technology are enabled by mature integrated photonics platforms. Silicon is considered the material of choice and silicon photonics is rapidly becoming a mainstream industrial platform. One of the key elements lacking is the availability of non-volatile programmable materials compatible with silicon photonics which could be used to reconfigure and program circuits without requiring continuous power to maintain its state. Recently, a new family of phase change materials Sb2S3 and Sb2Se3 have gained interest for their properties including a refractive index close to silicon, large switching of refractive index at telecommunications wavelength and ultralow optical losses. I will present here our results in developing the materials and their integration into new types of reconfigurable silicon photonics devices.

Decision-making by artificial neural networks with minimal latency is paramount for numerous applications such as navigation, tracking, and real-time machine action systems. This requires the machine learning hardware to handle multidimensional data with a high throughput. Processing convolution operations being the major computational tool for data classification tasks, unfortunately, follows a challenging run-time complexity scaling law. However, implementing the convolution theorem homomorphically in a Fourier-optic display-light-processor enables a non-iterative O(1) runtime complexity for data inputs beyond 1,000 × 1,000 large matrices. Following this approach, here we demonstrate data streaming multi-kernel image batch-processing with a Fourier Convolutional Neural Network (FCNN) accelerator. We show image batch processing of large-scale matrices as passive 2-million dot-product multiplications performed by digital light-processing modules in the Fourier domain. In addition, we parallelize this optical FCNN system further by utilizing multiple spatio-parallel diffraction orders, thus achieving a 98-times throughput improvement over state-of-art FCNN accelerators. The comprehensive discussion of the practical challenges related to working on the edge of the system’s capabilities highlights issues of crosstalk in Fourier domain and resolution scaling laws. Accelerating convolutions by utilizing the massive parallelism in display technology brings forth a non-van Neuman-based machine learning acceleration.

The ability of metamaterials to manipulate optical waves in the spatial and spectral domain has provided opportunities for image encoding. This combined with recent advances in hyperspectral imaging suggest exciting new opportunities for secure encryption. In this work, we propose a multi-channel scheme for secure image transmission across multiple wavelength channels. In contrast to conventional encryption schemes that perform a 1-to-1 transformation on a given plain image, we propose a 1-to-n transformation. We show that our scheme provides security against attacks of varying complexity provided a reasonable number of spectral channels is used.

In recent years, heterogeneous machine learning accelerators have become of significant interest to science, engineering, and industry. At the same time, demand for data security has increased significantly, especially in the looming post- quantum encryption era. From a hardware processing point of view, both are challenged by electronic capacitive interconnect delay and energy, and, in the case of heterogeneous systems such as electronic-photonic accelerators, by parasitic domain crossings. With analog optical AI accelerators having demonstrated high throughout potential (TOPS to even POPS) and high operation efficiency (TOPS/W), they have not demonstrated the ability to perform AI classification task on encrypted data. Here, we present an optical hashing and compression scheme that is based on SWIFFT - a post-quantum hashing family of algorithms. High degree optical hardware-to-algorithm homomorphism allows to optimally harvest well- understood potential of free-space processing: innate parallelism, low latency tensor by-element multiplication and Fourier transform. The algorithm can provide several orders of magnitude increase in processing speed by replacing slow high-resolution CMOS cameras with ultra-fast and signal-triggered CMOS detector arrays. Additionally, the information acquired in this way will require much lower transmission throughput, less in silico processing power, storage, and will be pre-hashed facilitating cheap optical information security. This technology has the potential to allow heterogeneous convolutional 4f classifiers to get closer in performance to their fully electronic counterparts

Colorectal cancer is a very common cancer and is currently the second most common cause of death from cancer. It is a very serious cancer, and receiving proper treatment is critical. Testing for microsatellite instability (MSI), which is present in 15% of colorectal cancer cases, is currently an expensive and very time-consuming process. However, testing is necessary for these individuals, to help determine how their treatment should progress. This paper presents a deep learning algorithm to distinguish between MSI and MSS scans. It shows that by heavily compressing these types of algorithms, they can run on embedded computing systems such as a raspberry pi or a cell phone. These computing systems can be cheap and use little power. The algorithms can still retain relatively high accuracy, in this case around 80%. Colorectal deep learning algorithms have not been implemented on low power devices in prior publications.