-
Hardware-Friendly Synaptic Orders and Timescales in Liquid State Machines for Speech Classification
Authors:
Vivek Saraswat,
A**kya Gorad,
Anand Naik,
Aakash Patil,
Udayan Ganguly
Abstract:
Liquid State Machines are brain inspired spiking neural networks (SNNs) with random reservoir connectivity and bio-mimetic neuronal and synaptic models. Reservoir computing networks are proposed as an alternative to deep neural networks to solve temporal classification problems. Previous studies suggest 2nd order (double exponential) synaptic waveform to be crucial for achieving high accuracy for…
▽ More
Liquid State Machines are brain inspired spiking neural networks (SNNs) with random reservoir connectivity and bio-mimetic neuronal and synaptic models. Reservoir computing networks are proposed as an alternative to deep neural networks to solve temporal classification problems. Previous studies suggest 2nd order (double exponential) synaptic waveform to be crucial for achieving high accuracy for TI-46 spoken digits recognition. The proposal of long-time range (ms) bio-mimetic synaptic waveforms is a challenge to compact and power efficient neuromorphic hardware. In this work, we analyze the role of synaptic orders namely: δ (high output for single time step), 0th (rectangular with a finite pulse width), 1st (exponential fall) and 2nd order (exponential rise and fall) and synaptic timescales on the reservoir output response and on the TI-46 spoken digits classification accuracy under a more comprehensive parameter sweep. We find the optimal operating point to be correlated to an optimal range of spiking activity in the reservoir. Further, the proposed 0th order synapses perform at par with the biologically plausible 2nd order synapses. This is substantial relaxation for circuit designers as synapses are the most abundant components in an in-memory implementation for SNNs. The circuit benefits for both analog and mixed-signal realizations of 0th order synapse are highlighted demonstrating 2-3 orders of savings in area and power consumptions by eliminating Op-Amps and Digital to Analog Converter circuits. This has major implications on a complete neural network implementation with focus on peripheral limitations and algorithmic simplifications to overcome them.
△ Less
Submitted 29 April, 2021;
originally announced April 2021.
-
Shallow-UWnet : Compressed Model for Underwater Image Enhancement
Authors:
Ankita Naik,
Apurva Swarnakar,
Kartik Mittal
Abstract:
Over the past few decades, underwater image enhancement has attracted increasing amount of research effort due to its significance in underwater robotics and ocean engineering. Research has evolved from implementing physics-based solutions to using very deep CNNs and GANs. However, these state-of-art algorithms are computationally expensive and memory intensive. This hinders their deployment on po…
▽ More
Over the past few decades, underwater image enhancement has attracted increasing amount of research effort due to its significance in underwater robotics and ocean engineering. Research has evolved from implementing physics-based solutions to using very deep CNNs and GANs. However, these state-of-art algorithms are computationally expensive and memory intensive. This hinders their deployment on portable devices for underwater exploration tasks. These models are trained on either synthetic or limited real world datasets making them less practical in real-world scenarios. In this paper we propose a shallow neural network architecture, \textbf{Shallow-UWnet} which maintains performance and has fewer parameters than the state-of-art models. We also demonstrated the generalization of our model by benchmarking its performance on combination of synthetic and real-world datasets.
△ Less
Submitted 6 January, 2021;
originally announced January 2021.
-
Open-Source Matlab-Based Graphical User Interface (GUI) For Computer Control of Microscopes Using Micro-Manager
Authors:
Quang Long Pham,
David Chege,
Timothy Dijamco,
Migle Surblyte,
Akshay Naik,
Kolawole Campbell,
Nhat-Anh-Nguyen Tong,
Roman Voronov
Abstract:
Live time-lapse microscopy is essential for a wide range of biological applications. Software-based automation is the gold standard for the operation of hardware accessories necessary for image acquisition. Given that current software packages are neither affordable nor open to complex structured programming, we have developed a Matlab-based graphical user interface (GUI) that is fundamentally acc…
▽ More
Live time-lapse microscopy is essential for a wide range of biological applications. Software-based automation is the gold standard for the operation of hardware accessories necessary for image acquisition. Given that current software packages are neither affordable nor open to complex structured programming, we have developed a Matlab-based graphical user interface (GUI) that is fundamentally accessible while providing limitless avenues for further customization. The GUI simultaneously communicates with the open-source, cross-functional platform micro-Manager for controlling hardware for time-lapse image acquisition and with other software for image processing. The use of the GUI is demonstrated through an 18-hour cell migration experiment. The results suggest that the GUI can generate high-quality, high-throughput time-lapse images. The core code behind the GUI is open-source so that it can be modified and upgraded. Therefore, it benefits the researchers worldwide by providing them a fundamental yet functional template to add in advanced features specific to their needs, such as additional non-microscopy hardware parts and software packages supported by the Matlab software.
△ Less
Submitted 18 June, 2019; v1 submitted 26 April, 2019;
originally announced April 2019.
