Publications

Journal Papers

Roman Kaplan, Leonid Yavits and Ran Ginosar, “RASSA: Resistive Pre-Alignment Accelerator for Approximate DNA Long Read Mapping”

IEEE Micro, 2019 (preprint).

In IEEExplore      PDF

Roman Kaplan, Leonid Yavits and Ran Ginosar, “PRINS: Processing-in-Storage Acceleration of Machine Learning.”

IEEE Transactions on Nanotechnology, vol. 17, no. 5, pp. 889-896, 2018.

In IEEExplore      PDF

Leonid Yavits, Roman Kaplan and Ran Ginosar, "Enabling Full Associativity with Memristive Address Decoder

IEEE Micro Special Issue on Memristor-Based Computing, vol. 38, no. 5, pp. 32-40, 2018.

In IEEExplore      PDF

Roman Kaplan, Leonid Yavits, Ran Ginosar and Uri Weiser, “A Resistive CAM Processing-in-Storage Architecture for DNA Sequence Alignment.”

IEEE Micro Special Issue on Architectures for the Post Moore Era, vol. 37, no. 4, pp. 20-28, 2017.

In IEEExplore

Roman Kaplan, Leonid Yavits, and Ran Ginosar, "From Processing-in-Memory to Processing-in-Storage."

Journal of Supercomputing Frontiers and Innovations, vol. 4, no. 3, pp. 99-116, 2017.

In journal's website      PDF

Refereed Papers in Conference Proceedings

Roman Kaplan, Leonid Yavits, Amir Morad, and Ran Ginosar, "Deduplication in Resistive Content Addressable Memory Based Solid State Drive."

26th Power and Timing Modeling, Optimization and Simulation, pp. 100-106, 2016.

In IEEExplore

Posters

Refereed Posters

Roman Kaplan, Leonid Yavits, and Ran Ginosar, "From Processing-in-Memory to Processing-in-Storage."

International Supercomputing Conference for High Performance Computing, Jun 2017, Frankfurt.

Research Poster Award Winner at the ‘Architectures and Networks’ category

Poster PDF

Roman Kaplan, Leonid Yavits, and Ran Ginosar, "In-Storage Implementation of K-Means in Resistive CAM."

International Conference on Memristive Materials, Devices & Systems, Apr 2017, Athens.

Poster absract

Roman Kaplan, Leonid Yavits, and Ran Ginosar, "From Processing-in-Memory to Processing-in-Storage."

In PACT, Sep 2016, Haifa.

Winner of PACT ACM Student Research Competition

Poster PDF

Cool Personal Projects

Asynchronous In-Browser Parallel Training of Convolutional Neural Networks

This is a fullstack project, it has both a client (training node) and a server (training manager). Both sides are written in Javascript (Node.js at the server).
A CNN training code is fetched by accessing the client page (link to client page). Pressing on "Train" starts the training process.
The server then receives a training request and sends a the client the current CNN model and a batch number to train. The client downloads the training batch and starts training.
Monitoring the training is possible via an admin page.
The training dataset is CIFAR10. Batch size is 1000 images. Training client code is based on ConvNetJS, written by Andrej Karpathy.
Link to project's GitHub page

Generate PowerPoint Presentations with Python

In one sentence, this project generates PowerPoint slides with Python. Accessing the Windows API is possible via pyWin32, a set of extension modules that provides access to many of the Windows API functions.
The project generates a K-means demo slides. The dataset contains 100 samples, divided to 3 clusters.
The script generates 100 random samples, initializes 3 centers (means) then executes the K-means algorithm for several iterations.
Every iteration is then translated to a PowerPoint slide, with an animation for the center and classification updates.
Link to project's GitHub page
The inspiration for project came from a YouTube talk by Anand S. Link to Anand's GitHub page