Sunny Raj

A compact 8-bit adder design using in-memory memristive computing: Towards solving the Feynman Grand Prize challenge

We introduce a new compact in-memory computing design for implementing 8-bit addition using eight vertically-stacked nanoscale crossbars of one-diode one-memristor 1D1M switches. Each crossbar in our design only has 5 rows and 4 columns. Hence, the design may be used to fabricate a compact 8-bit adder that meets the size constraint of 50nm χ 50nm χ 50nm imposed by the electrical component of the Feynman Grand Prize. The potential availability of sub-5nm nanoscale memristors and single-molecule diode devices coupled with the ability to fabricate high-density nanoscale memristor crossbars suggests that our design may eventually be fabricated to meet the size constraints of the Feynman Grand Prize.

Duality-Based Nested Controller Synthesis from STL Specifications for Stochastic Linear Systems

We propose an automatic synthesis technique to generate provably correct controllers of stochastic linear dynamical systems for Signal Temporal Logic (STL) specifications. While formal synthesis problems can be directly formulated as exists-forall constraints, the quantifier alternation restricts the scalability of such an approach. We use the duality between a system and its proof of correctness to partially alleviate this challenge. We decompose the controller synthesis into two subproblems, each addressing orthogonal concerns - stabilization with respect to the noise, and meeting the STL specification. The overall controller is a nested controller comprising of the feedback controller for noise cancellation and an open loop controller for STL satisfaction. The correct-by-construction compositional synthesis of this nested controller relies on using the guarantees of the feedback controller instead of the controller itself. We use a linear feedback controller as the stabilizing controller for linear systems with bounded additive noise and over-approximate its ellipsoid stability guarantee with a polytope. We then use this over-approximation to formulate a mixed-integer linear programming (MILP) problem to synthesize an open-loop controller that satisfies STL specifications.

In-memory flow-based stochastic computing on memristor crossbars using bit-vector stochastic streams

Nanoscale memristor crossbars provide a natural fabric for in-memory computing and have recently been shown to efficiently perform exact logical operations by exploiting the flow of current through crossbar interconnects. In this paper, we extend the flow-based crossbar computing approach to approximate stochastic computing. First, we show that the natural flow of current through probabilistically-switching memristive nano-switches in crossbars can be used to perform approximate stochastic computing. Second, we demonstrate that optimizing the approximate stochastic computations in terms of the number of required random bits leads to stochastic computing using bit-vector stochastic streams of varying bit-widths — a hybrid of the traditional full-width bit-vector computing approach and the traditional bit-stream stochastic computing methodology. This hybrid approach based on bit-vector stochastic streams of different bit-widths can be efficiently implemented using an in-memory nanoscale memristive crossbar computing framework.

SATYA: Defending against Adversarial Attacks using Statistical Hypothesis Testing

The paper presents a new defense against adversarial attacks for deep neural networks. We demonstrate the effectiveness of our approach against the popular adversarial image generation method DeepFool. Our approach uses Wald’s Sequential Probability Ratio Test to sufficiently sample a carefully chosen neighborhood around an input image to determine the correct label of the image. On a benchmark of 50,000 randomly chosen adversarial images generated by DeepFool we demonstrate that our method \(\mathcal {SATYA}\) is able to recover the correct labels for 95.76% of the images for CaffeNet and 97.43% of the correct label for GoogLeNet.

Testing autonomous cyber-physical systems using fuzzing features from convolutional neural networks: work-in-progress

Autonomous cyber-physical systems rely on modern machine learning methods such as deep neural networks to control their interactions with the physical world. Testing of such intelligent cyber-physical systems is a challenge due to the huge state space associated with high-resolution visual sensory inputs. We demonstrate how fuzzing the input using patterns obtained from the convolutional filters of an unrelated convolutional neural network can be used to test computer vision algorithms implemented in intelligent cyber-physical systems. Our method discovers interesting counterexamples to a pedestrian detection algorithm implemented in the popular OpenCV library. Our approach also unearths counterexamples to the correct behavior of an autonomous car similar to NVIDIAs end-to-end self-driving deep neural net running on the Udacity open-source simulator.

A theorem proving approach for automatically synthesizing visualizations of flow cytometry data

BackgroundPolychromatic flow cytometry is a popular technique that has wide usage in the medical sciences, especially for studying phenotypic properties of cells. The high-dimensionality of data generated by flow cytometry usually makes it difficult to visualize. The naive solution of simply plotting two-dimensional graphs for every combination of observables becomes impractical as the number of dimensions increases. A natural solution is to project the data from the original high dimensional space to a lower dimensional space while approximately preserving the overall relationship between the data points. The expert can then easily visualize and analyze this low-dimensional embedding of the original dataset.ResultsThis paper describes a new method, SANJAY, for visualizing high-dimensional flow cytometry datasets. This technique uses a decision procedure to automatically synthesize two-dimensional and three-dimensional projections of the original high-dimensional data while trying to minimize distortion. We compare SANJAY to the popular multidimensional scaling (MDS) approach for visualization of small data sets drawn from a representative set of benchmarks, and our experiments show that SANJAY produces distortions that are 1.44 to 4.15 times smaller than those caused due to MDS. Our experimental results show that SANJAY also outperforms the Random Projections technique in terms of the distortions in the projections.ConclusionsWe describe a new algorithmic technique that uses a symbolic decision procedure to automatically synthesize low-dimensional projections of flow cytometry data that typically have a high number of dimensions. Our algorithm is the first application, to our knowledge, of using automated theorem proving for automatically generating highly-accurate, low-dimensional visualizations of high-dimensional data.