Showing 1–6 of 6 results for author: Choudhuri, C

FPIRM: Floating-point Processing in Racetrack Memories

Authors: Sébastien Ollivier, Xinyi Zhang, Yue Tang, Chayanika Choudhuri, **gtong Hu, Alex K. Jones

Abstract: Convolutional neural networks (CNN) have become a ubiquitous algorithm with growing applications in mobile and edge settings. We describe a compute-in-memory (CIM) technique called FPIRM using Racetrack Memory (RM) to accelerate CNNs for edge systems. Using transverse read, a technique that can determine the number of '1's multiple adjacent domains, FPIRM can efficiently implement multi-operand bu… ▽ More Convolutional neural networks (CNN) have become a ubiquitous algorithm with growing applications in mobile and edge settings. We describe a compute-in-memory (CIM) technique called FPIRM using Racetrack Memory (RM) to accelerate CNNs for edge systems. Using transverse read, a technique that can determine the number of '1's multiple adjacent domains, FPIRM can efficiently implement multi-operand bulk-bitwise and addition computations, and two-operand multiplication. We discuss how FPIRM can implement both variable precision integer and floating point arithmetic. This allows both CNN inference and on-device training without expensive data movement to the cloud. Based on these functions we demonstrate implementation of several CNNs with back propagation using RM CIM and compare these to state-of-the-art implementations of CIM inference and training in Field-Programmable Gate Arrays. During training FPIRM improves by 2$\times$ the efficiency, by reducing the energy consumption by at least 27% and increasing the throughput by at least 18% against FPGA. △ Less

Submitted 1 August, 2022; v1 submitted 28 April, 2022; originally announced April 2022.

Comments: This paper is accepted to the IEEE Micro Magazine with the title "POD-RACING: Bulk-Bitwise to Floating-point Compute In Racetrack Memory for Machine Learning at the Edge"

Cascade Source Coding with a Side Information "Vending Machine"

Authors: Behzad Ahmadi, Chiranjib Choudhuri, Osvaldo Simeone, Urbashi Mitra

Abstract: The model of a side information "vending machine" (VM) accounts for scenarios in which the measurement of side information sequences can be controlled via the selection of cost-constrained actions. In this paper, the three-node cascade source coding problem is studied under the assumption that a side information VM is available and the intermediate and/or at the end node of the cascade. A single-l… ▽ More The model of a side information "vending machine" (VM) accounts for scenarios in which the measurement of side information sequences can be controlled via the selection of cost-constrained actions. In this paper, the three-node cascade source coding problem is studied under the assumption that a side information VM is available and the intermediate and/or at the end node of the cascade. A single-letter characterization of the achievable trade-off among the transmission rates, the distortions in the reconstructions at the intermediate and at the end node, and the cost for acquiring the side information is derived for a number of relevant special cases. It is shown that a joint design of the description of the source and of the control signals used to guide the selection of the actions at downstream nodes is generally necessary for an efficient use of the available communication links. In particular, for all the considered models, layered coding strategies prove to be optimal, whereby the base layer fulfills two network objectives: determining the actions of downstream nodes and simultaneously providing a coarse description of the source. Design of the optimal coding strategy is shown via examples to depend on both the network topology and the action costs. Examples also illustrate the involved performance trade-offs across the network. △ Less

Submitted 11 July, 2012; originally announced July 2012.

Comments: A shorter version has been submitted to ITW 2012 and is available online at arXiv:1204.1548

On Cascade Source Coding with A Side Information "Vending Machine"

Authors: Behzad Ahmadi, Osvaldo Simeone, Chiranjib Choudhuri, Urbashi Mitra

Abstract: The model of a side information "vending machine" accounts for scenarios in which acquiring side information is costly and thus should be done efficiently. In this paper, the three-node cascade source coding problem is studied under the assumption that a side information vending machine is available either at the intermediate or at the end node. In both cases, a single-letter characterization of t… ▽ More The model of a side information "vending machine" accounts for scenarios in which acquiring side information is costly and thus should be done efficiently. In this paper, the three-node cascade source coding problem is studied under the assumption that a side information vending machine is available either at the intermediate or at the end node. In both cases, a single-letter characterization of the available trade-offs among the rate, the distortions in the reconstructions at the intermediate and at the end node, and the cost in acquiring the side information are derived under given conditions. △ Less

Submitted 6 April, 2012; originally announced April 2012.

Comments: A shorter version has been submitted to ITW 2012

Causal State Communication

Authors: Chiranjib Choudhuri, Young-Han Kim, Urbashi Mitra

Abstract: The problem of state communication over a discrete memoryless channel with discrete memoryless state is studied when the state information is available strictly causally at the encoder. It is shown that block Markov encoding, in which the encoder communicates a description of the state sequence in the previous block by incorporating side information about the state sequence at the decoder, yields… ▽ More The problem of state communication over a discrete memoryless channel with discrete memoryless state is studied when the state information is available strictly causally at the encoder. It is shown that block Markov encoding, in which the encoder communicates a description of the state sequence in the previous block by incorporating side information about the state sequence at the decoder, yields the minimum state estimation error. When the same channel is used to send additional independent information at the expense of a higher channel state estimation error, the optimal tradeoff between the rate of the independent information and the state estimation error is characterized via the capacity- distortion function. It is shown that any optimal tradeoff pair can be achieved via rate-splitting. These coding theorems are then extended optimally to the case of causal channel state information at the encoder using the Shannon strategy. △ Less

Submitted 27 March, 2012; originally announced March 2012.

Comments: 25 pages, 1 figure

Action Dependent Strictly Causal State Communication

Authors: Chiranjib Choudhuri, Urbashi Mitra

Abstract: The problem of communication and state estimation is considered in the context of channels with actiondependent states. Given the message to be communicated, the transmitter chooses an action sequence that affects the formation of the channel states, and then creates the channel input sequence based on the state sequence. The decoder estimates the channel to some distortion as well as decodes the… ▽ More The problem of communication and state estimation is considered in the context of channels with actiondependent states. Given the message to be communicated, the transmitter chooses an action sequence that affects the formation of the channel states, and then creates the channel input sequence based on the state sequence. The decoder estimates the channel to some distortion as well as decodes the message. The capacity-distortion tradeoff of such a channel is characterized for the case when the state information is available strictly causally at the channel encoder. The problem setting extends the action dependent framework of [1] and as a special case recovers the results of few previously considered joint communication and estimation scenarios in [2], [3], [4]. The scenario when the action is also allowed to depend on the past observed states (adaptive action) is also considered. It is shown that such adaptive action yields an improved capacity-distortion function. △ Less

Submitted 4 February, 2012; originally announced February 2012.

Capacity Bounds for Relay Channels with Inter-symbol Interference and Colored Gaussian Noise

Authors: Chiranjib Choudhuri, Urbashi Mitra

Abstract: The capacity of a relay channel with inter-symbol interference (ISI) and additive colored Gaussian noise is examined under an input power constraint. Prior results are used to show that the capacity of this channel can be computed by examining the circular degraded relay channel in the limit of infinite block length. The current work provides single letter expressions for the achievable rates with… ▽ More The capacity of a relay channel with inter-symbol interference (ISI) and additive colored Gaussian noise is examined under an input power constraint. Prior results are used to show that the capacity of this channel can be computed by examining the circular degraded relay channel in the limit of infinite block length. The current work provides single letter expressions for the achievable rates with decodeand- forward (DF) and compress-and-forward (CF) processing employed at the relay. Additionally, the cut-set bound for the relay channel is generalized for the ISI/colored Gaussian noise scenario. All results hinge on showing the optimality of the decomposition of the relay channel with ISI/colored Gaussian noise into an equivalent collection of coupled parallel, scalar, memoryless relay channels. The region of optimality of the DF and CF achievable rates are also discussed. Optimal power allocation strategies are also discussed for the two lower bounds and the cut-set upper bound. As the maximizing power allocations for DF and CF appear to be intractable, the desired cost functions are modified and then optimized. The resulting rates are illustrated through the computation of numerical examples. △ Less

Submitted 1 March, 2011; originally announced March 2011.

Comments: 42 pages, 9 figures