Noisy Channel-Output Feedback Capacity of the Linear Deterministic Interference Channel

Similar documents
Parameter Free Iterative Decoding Metrics for Non-Coherent Orthogonal Modulation

Symmetric Decentralized Interference Channels with Noisy Feedback

Adaptive Modulation for Multiple Antenna Channels

The Stability Region of the Two-User Broadcast Channel

Calculation of the received voltage due to the radiation from multiple co-frequency sources

Rejection of PSK Interference in DS-SS/PSK System Using Adaptive Transversal Filter with Conditional Response Recalculation

Comparative Analysis of Reuse 1 and 3 in Cellular Network Based On SIR Distribution and Rate

Review: Our Approach 2. CSC310 Information Theory

Space Time Equalization-space time codes System Model for STCM

Capacity bounds on multi-pair two-way communication with a base-station aided by a relay

PRACTICAL, COMPUTATION EFFICIENT HIGH-ORDER NEURAL NETWORK FOR ROTATION AND SHIFT INVARIANT PATTERN RECOGNITION. Evgeny Artyomov and Orly Yadid-Pecht

Distributed Resource Allocation and Scheduling in OFDMA Wireless Networks

antenna antenna (4.139)

Performance Analysis of Multi User MIMO System with Block-Diagonalization Precoding Scheme

Dynamic Optimization. Assignment 1. Sasanka Nagavalli January 29, 2013 Robotics Institute Carnegie Mellon University

A study of turbo codes for multilevel modulations in Gaussian and mobile channels

Ergodic Capacity of Block-Fading Gaussian Broadcast and Multi-access Channels for Single-User-Selection and Constant-Power

Multiband Jamming Strategies with Minimum Rate Constraints

Digital Transmission

Define Y = # of mobiles from M total mobiles that have an adequate link. Measure of average portion of mobiles allocated a link of adequate quality.

Direct Sequence Spread Spectrum (DSSS)

A NSGA-II algorithm to solve a bi-objective optimization of the redundancy allocation problem for series-parallel systems

Efficient Large Integers Arithmetic by Adopting Squaring and Complement Recoding Techniques

IMPACT OF LIMITED FEEDBACK ON MIMO- OFDM SYSTEMS USING JOINT BEAMFORMING

Uplink User Selection Scheme for Multiuser MIMO Systems in a Multicell Environment

HUAWEI TECHNOLOGIES CO., LTD. Huawei Proprietary Page 1

Multi-pair bi-directional relay networks part I: protocols which exploit side-information

On Interference Alignment for Multi-hop MIMO Networks

Evaluate the Effective of Annular Aperture on the OTF for Fractal Optical Modulator

Degrees of Freedom of Full-Duplex Multiantenna Cellular Networks

ECE 2133 Electronic Circuits. Dept. of Electrical and Computer Engineering International Islamic University Malaysia

To: Professor Avitabile Date: February 4, 2003 From: Mechanical Student Subject: Experiment #1 Numerical Methods Using Excel

Hierarchical Generalized Cantor Set Modulation

熊本大学学術リポジトリ. Kumamoto University Repositor

An Alternation Diffusion LMS Estimation Strategy over Wireless Sensor Network

A new family of linear dispersion code for fast sphere decoding. Creative Commons: Attribution 3.0 Hong Kong License

AIR FORCE INSTITUTE OF TECHNOLOGY

A Comparison of Two Equivalent Real Formulations for Complex-Valued Linear Systems Part 2: Results

A High-Sensitivity Oversampling Digital Signal Detection Technique for CMOS Image Sensors Using Non-destructive Intermediate High-Speed Readout Mode

The Performance Improvement of BASK System for Giga-Bit MODEM Using the Fuzzy System

IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO. 12, DECEMBER

Impact of Interference Model on Capacity in CDMA Cellular Networks. Robert Akl, D.Sc. Asad Parvez University of North Texas

A New Opportunistic Interference Alignment Scheme and Performance Comparison of MIMO Interference Alignment with Limited Feedback

Throughput Maximization by Adaptive Threshold Adjustment for AMC Systems

Walsh Function Based Synthesis Method of PWM Pattern for Full-Bridge Inverter

NOVEL ITERATIVE TECHNIQUES FOR RADAR TARGET DISCRIMINATION

Approximating User Distributions in WCDMA Networks Using 2-D Gaussian

Approximate Joint MAP Detection of Co-Channel Signals

Revision of Lecture Twenty-One

Bit-interleaved Rectangular Parity-Check Coded Modulation with Iterative Demodulation In a Two-Node Distributed Array

Relevance of Energy Efficiency Gain in Massive MIMO Wireless Network

Information-Theoretic Comparison of Channel Capacity for FDMA and DS-CDMA in a Rayleigh Fading Environment

Figure.1. Basic model of an impedance source converter JCHPS Special Issue 12: August Page 13

Uncertainty in measurements of power and energy on power networks

Performance Analysis of Power Line Communication Using DS-CDMA Technique with Adaptive Laguerre Filters

Chaotic Filter Bank for Computer Cryptography

NATIONAL RADIO ASTRONOMY OBSERVATORY Green Bank, West Virginia SPECTRAL PROCESSOR MEMO NO. 25. MEMORANDUM February 13, 1985

Joint Power Control and Scheduling for Two-Cell Energy Efficient Broadcasting with Network Coding

On Sensor Fusion in the Presence of Packet-dropping Communication Channels

How to Scale Up the Spectral Efficiency of Multi-way Massive MIMO Relaying?

Opportunistic Interference Alignment with 1-Bit Feedback in 3-Cell Interference Channels

High Speed, Low Power And Area Efficient Carry-Select Adder

4492 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 65, NO. 10, OCTOBER 2017

Development of Neural Networks for Noise Reduction

THE GENERATION OF 400 MW RF PULSES AT X-BAND USING RESONANT DELAY LINES *

Two-Phase Cooperative Broadcasting Based on Batched Network Code

RESOURCE CONTROL FOR HYBRID CODE AND TIME DIVISION SCHEDULING

Understanding the Spike Algorithm

A MODIFIED DIFFERENTIAL EVOLUTION ALGORITHM IN SPARSE LINEAR ANTENNA ARRAY SYNTHESIS

POLYTECHNIC UNIVERSITY Electrical Engineering Department. EE SOPHOMORE LABORATORY Experiment 1 Laboratory Energy Sources

LOCAL DECODING OF WALSH CODES TO REDUCE CDMA DESPREADING COMPUTATION

Achievable Rate Region of CSMA Schedulers in Wireless Networks with Primary Interference Constraints

On the Performance of Space-Time MIMO Multiplexing for Free Space Optical Communications

Performance Study of OFDMA vs. OFDM/SDMA

Design Rules for Efficient Scheduling of Packet Data on Multiple Antenna Downlink

Generalized Incomplete Trojan-Type Designs with Unequal Cell Sizes

Side-Match Vector Quantizers Using Neural Network Based Variance Predictor for Image Coding

Bit Error Probability of Cooperative Diversity for M-ary QAM OFDM-based system with Best Relay Selection

Low Switching Frequency Active Harmonic Elimination in Multilevel Converters with Unequal DC Voltages

TECHNICAL NOTE TERMINATION FOR POINT- TO-POINT SYSTEMS TN TERMINATON FOR POINT-TO-POINT SYSTEMS. Zo = L C. ω - angular frequency = 2πf

FUTURE wireless systems will need to provide high data

4.3- Modeling the Diode Forward Characteristic

IEE Electronics Letters, vol 34, no 17, August 1998, pp ESTIMATING STARTING POINT OF CONDUCTION OF CMOS GATES

Distributed Uplink Scheduling in EV-DO Rev. A Networks

The Detection Algorithms Performance in BLAST Enhanced IEEE a WLAN Standard on Measured Channels. University of Bristol

Enhancing Throughput in Wireless Multi-Hop Network with Multiple Packet Reception

Performance Analysis of the Weighted Window CFAR Algorithms

An Attack-Defense Game Theoretic Analysis of Multi-Band Wireless Covert Timing Networks

A MODIFIED DIRECTIONAL FREQUENCY REUSE PLAN BASED ON CHANNEL ALTERNATION AND ROTATION

A Spreading Sequence Allocation Procedure for MC-CDMA Transmission Systems

The Throughput of Hybrid-ARQ in Block Fading under Modulation Constraints

Evaluation of Downlink Performance of a Multiple-Cell, Rake Receiver Assisted CDMA Mobile System

Reduced Cluster Search ML Decoding for QO-STBC Systems

The Spectrum Sharing in Cognitive Radio Networks Based on Competitive Price Game

The Dynamic Utilization of Substation Measurements to Maintain Power System Observability

Index Terms Adaptive modulation, Adaptive FEC, Packet Error Rate, Performance.

THE USE OF CONVOLUTIONAL CODE FOR NARROWBAND INTERFERENCE SUPPRESSION IN OFDM-DVBT SYSTEM

Power Control for Wireless Data

CROSSFIRE. FP7 Contract Number:

A Lower Bound for τ(n) of Any k-perfect Numbers

Transcription:

Nosy Channel-Output Feedback Capacty of the Lnear Determnstc Interference Channel Vctor Quntero, Samr M. Perlaza, Jean-Mare Gorce arxv:.4649v6 [cs.it] Jan 6 Abstract In ths paper, the capacty regon of the two-user lnear determnstc (LD) nterference channel wth nosy output feedback (IC-NOF) s fully characterzed. Ths result allows the dentfcaton of several asymmetrc scenaros n whch mplementng channel-output feedback n only one of the transmtterrecever pars s as benefcal as mplementng t n both lnks, n terms of achevable ndvdual rate and sum-rate mprovements w.r.t. the case wthout feedback. In other scenaros, the use of channel-output feedback n any of the transmtter-recever pars benefts only one of the two pars n terms of achevable ndvdual rate mprovements or smply, t turns out to be useless,.e., the capacty regons wth and wthout feedback turn out to be dentcal even n the full absence of nose n the feedback lnks. Index Terms Capacty, Lnear Determnstc Interference Channel, Nosy Channel-Output Feedback. I. INTRODUCTION Perfect channel-output feedback (POF) has been shown to dramatcally enlarge the capacty regon of the two-user nterference channel (IC) [], [], [], [4], []. More recently, the same observaton has been made wth a larger number of transmtter-recever pars n the IC [6]. In general, when a transmtter observes the channel-output at ts ntended recever, t obtans a nosy verson of the sum of ts own transmtted sgnal and the nterferng sgnals from other transmtters. Ths mples that, subject to a fnte feedback delay, transmtters know at least partally the nformaton transmtted by other transmtters n the network. Ths nduces an mplct cooperaton between transmtters that allows them to use nterference as sde-nformaton [], [], [], [8], [9]. A more explct cooperaton s also observed n the case n whch one of the transmtter-recever pars acts as a relay for the other transmtter-recever par by provdng an alternatve path: transmtter recever j transmtter j recever []. These types of cooperaton, even when t s not explctly desred by both transmtter-recever pars, play a fundamental role n enlargng the capacty regon. Interestngly, ths holds also n the case of fully decentralzed networks n whch each transmtter-recever par seeks exclusvely to ncrease ts The authors are wth the CITI Laboratory of the Insttut Natonal de Recherche en Informatque et en Automatque (INRIA), Unversté de Lyon and Insttut Natonal de Scences Aplquées (INSA) de Lyon. 6 Av. des Arts 696 Vlleurbanne, France. ({Vctor.Quntero-Florez, Samr. Perlaza, Jean- Mare.Gorce}@nra.fr). Vctor Quntero s also wth Unversdad del Cauca, Popayán, Colomba. Samr M. Perlaza s also wth the Department of Electrcal Engneerng at Prnceton Unversty, Prnceton, NJ. Ths research was supported n part by the European Commsson under Mare Sklodowska-Cure Indvdual Fellowshp No. 696 (CYBERNETS); the Doctoral School n Electroncs, Electrcal Engneerng and Automaton (EEA) of Unversté de Lyon; Unversdad del Cauca, Popayán, Colomba; and the Admnstratve Department of Scence, Technology and Innovaton of Colomba (Colcencas), fellowshp No. 6-. ndvdual rate. That s, channel-output feedback ncreases both the capacty regon and the Nash equlbrum (NE) regon []. Despte the vast exstng lterature, the benefts of feedback are unfortunately less well understood when the channeloutput feedback lnks are mpared by addtve nose. The capacty regon of the LD-IC wth nosy channel-output feedback (NOF) s known only n the two-user symmetrc case, see []. The converse regon n [] nherts exstng outer bounds from the case of POF, the cut-set outer bounds and ncludes two new outer bounds. The outer-bounds nherted from the POF are those of the ndvdual rates and the sumrate n []. The new outer-bounds are of the form R + R and R + R j. The achevablty regon n [] s obtaned usng a partcularzaton of the achevablty scheme presented n [], whch holds for a more general model,.e., nterference channel wth generalzed feedback. In ths paper, the results presented n [] are generalzed for the asymmetrc case and the correspondng capacty regon of the two-user LD-IC-NOF s fully characterzed. Ths generalzaton s acheved by usng the same tools as n [], however, t s far from trval due to the number of parameters that descrbe ths channel model: two forward sgnal to nose ratos (SNRs) n, n, two feedback SNRs n, n and two forward nterference to nose ratos (INRs) n, n. The new converse regon also nherts exstng outer bounds from the case of POF, the cut-set outer bounds and ncludes two new outer bounds on R + R and R + R j. These new bounds generalze those presented n []. The achevablty regon s obtaned by usng a codng scheme that combnes a three-part message splttng, superposton codng and backward decodng. Despte the fact that ths codng scheme s bult usng the exact number of requred message-splttng parts for the IC-NOF, t can stll be consdered as a specal case of the general scheme presented n []. Fnally, ths paper s concluded by a dscusson n whch numercal examples are presented to hghlght the benefts of NOF. At the same tme, examples n whch NOF s absolutely useless n terms of capacty regon mprovement are also presented. Due to space constrants, the achevablty and the converse parts of the proof are presented n []. II. LINEAR DETERMINISTIC INTERFERENCE CHANNEL WITH NOISY-CHANNEL OUTPUT FEEDBACK Consder the two-user LD-IC-NOF, wth parameters n, n, n, n, n and n descrbed n Fg.. n, {, }, s a non-negatve nteger used to represent the sgnal to nose rato (SNR) at recever ; n j, {, } and j {, }\ {}, s a non-negatve nteger used to represent the nterference

n n! n! n X, X, X, X,4 X, X, X, X, X,4 TX TX RX RX X, X, X, L X, X,4 L X, X, L X, X, X, L X, X, L X, X, L X,4 X,4 L X, Sgnal Interference Feedback Fg.. Two-user lnear determnstc nterference channel wth nosy channeloutput feedback (LD-IC-NOF). n n calculated as follows p = M M {ˆb,l b,l}. () l= A rate par (R, R ) R + s sad to be achevable f t satsfes the followng defnton. Defnton (Achevable Rate Pars): The rate par (R, R ) R + s achevable f there exsts at least one par of codebooks X and X wth codewords of length N, wth the correspondng encodng functons f (),..., f (N) and f (),..., f (N) such that the average bt error probablty can be made arbtrarly small by lettng the block length N grows to nfnty. The followng secton determnes the set of all the rate pars (R, R ) that are achevable n the LD-IC-NOF wth parameters n, n, n, n, n and n. to nose rato (INR) at recever from transmtter j; and n, {, }, s a non-negatve nteger used to represent the sgnal to nose rato (SNR) at transmttter n the feedback lnk from recever. At transmtter, wth {, }, the channel-nput X (n) at channel use n, wth n {,..., N}, ä T, s a q-dmensonal bnary vector X (n) = Ä X (n),,..., X(n),q wth q = max ( n, n, n, n ) and N the block-length. At recever, the channel-output Y (n) a q-dmensonal bnary vector Y (n) at channel use n s also = Ä (n) Y,,..., ä T. Y (n),q The nput-output relaton durng channel use n s gven as follows (n) Y =S q n X (n) + S q nj X (n) j, () and the feedback sgnal avalable at transmtter at the end of channel use n s: (n) Y =S (q n ) + Y (n d), () where d s a fnte feedback delay, addtons and multplcatons are defned over the bnary feld, and S s a q q lower shft matrx. Ä n and n j correspond to äù SNR and log (INR j ) res- The parameters ö n, log Ä äù ö SNR, log pectvely, where SNR, SNR and INR j are parameters of the Gaussan nterference channel (G-IC). Transmtter sends M nformaton bts b,,..., b,m by sendng the codeword Ä ä X (),..., X (N). The encoder of transmtter can be modeled as a set of determnstc mappngs f (),..., f (N), wth f () : {, } M {, } q and n {,..., N}, f (n) : {, } M {, } q(n ) {, } q, such that ( ) b,,..., b,m and () X () =f () X (n) =f (n) ( b,,..., b,m, Y (),..., Y (n ) ). (4) At the end of the block, recever uses the sequence Y (),..., Y (N) to generate the estmates ˆb,,..., ˆb,M. The average bt error probablty at recever, denoted by p, s III. MAIN RESULTS Denote by C( n, n, n, n, n, n ) the capacty regon of the LD-IC-NOF wth parameters n, n, n, n, n and n. Theorem (on top of next page) fully characterzes the capacty regon C( n, n, n, n, n, n ). A. Proofs Theorem fully characterzes the capacty regon of the LD-IC-NOF. That s, the converse and achevable regons are dentcal. In the converse regon, the nequaltes (6) and (8) are nherted from the converse regon of the LD-IC-POF n []. The nequalty n () s a smple cut-set bound whose proof s presented n []. The nequaltes (9) and () are new and generalze those presented n []. The correspondng proofs are presented n []. The achevablty regon s obtaned usng a codng scheme that combnes a three-part message splttng, superposton codng and backward decodng, as frst suggested n [], [], []. Ths codng scheme s fully descrbed n [] and t s specally desgned for the IC-NOF. However, t can also be obtaned as a specal case of the more general scheme,.e., nterference channel wth generalzed feedback, presented n []. The relevance of ths new achevablty scheme s that t plays a key role n the achevablty of the NE regon, subject to the ncluson of random messages, as suggested n [], []. Nonetheless, the analyss of the achevablty of the NE regon [4] s out of the scope of ths paper. B. Connectons to Exstng Results In the case n whch channel-output feedback s not avalable,.e., n = n =, n = n, for all {, }, Theorem reduces to the capacty regon of the LD-IC wthout feedback,.e., Lemma 4 n []. Conversely, when perfect channel-output feedback lnks are avalable.e., n = max ( n, n ) and n = max ( n, n ), Theorem reduces to the capacty of the LD-IC-POF, that s, Corollary n [] n the correspondng notaton.

Theorem : The capacty regon C( n, n, n, n, n, n ) of the two-user LD-IC-NOF s the set of non-negatve rate pars (R, R ) that satsfy {, } and j {, } \ {}: R mn (max ( n, n j ), max ( n, n j )), (6) R mn Ä max ( n, n j ), max Ä n, n jj ( n jj n j ) +ää, () R +R mn Ä max ( n, n ) + ( n n ) +, max ( n, n ) + ( n n ) +ä, (8) R +R max Ä ( ä Ä ä n n ) +, n + max ( n n ) +, n (9) ( Ä + mn ( n, max ( n, n )) ( n n ) +ä + (n n ) + mn ( n, n ) + mn Ä ( ä ) + n n ) +, n Ä +( mn ( n, max ( n, n )) ( n n ) +ä + (n n ) + mn ( n, n ) + mn Ä ( ä ) +, n n ) +, n R +R j max ( n jj, n j ) + max ( n, n j ) + ( n n j ) + mn Ä ( ä n jj n j ) +, n j () Ä +( mn ( n jj, max ( n jj, n j )) ( n jj n j ) +ä +! (nj n jj ) + mn ( n jj, n j ) + mn Ä ( ä ) +, n n jj n j ) + =,! n =,n =,n =, n, = n j, n = 4 () In the symmetrc case n whch channel-output feedback lnks are mpared by nose,.e., n = n = n, n = n = m and n = n = l, Theorem reduces to the capacty of the symmetrc LD-IC-NOF,.e., Theorem n []. Note also that n the case n whch the nterference channel s symmetrc, Theorem partally generalzes Theorem 4. n [9]. For nstance, note that when there exsts only one perfect channel-output feedback,.e., n = n = n, n = n = m, n = max ( n, n ), and n =, Theorem reduces to Theorem 4. (model ) n [9]. In the two-user IC-NOF, a transmtter sees a nosy verson of the sum of ts own transmtted sgnal and the nterferng sgnal from the other transmtter. Hence, subject to a fnte delay, one transmtter knows at least partally the nformaton transmtted by the other transmtter n the network. Ths observaton hghlghts the connectons between the IC wth feedback and the IC wth source cooperaton studed n []. C. Dscusson Ths secton provdes a set of numercal examples n whch partcular scenaros are hghlghted to show that channeloutput feedback can be strongly benefcal for enlargng the capacty regon of the two-user LD-IC even when channeloutput feedback lnks are mpared by nose. At the same tme, t also hghlghts other examples n whch channel-output feedback does not brng any beneft n terms of the capacty regon. These benefts are gven n terms of the followng metrcs: (a) ndvdual rate mprovements and ; and (b) sum-rate mprovement Σ. In order to formally defne, and Σ, consder an LD-IC-NOF wth parameters n, n, n, n, n and n. The maxmum mprovement ( n, n, n, n, n, n ) of the ndvdual rate R due to the effect of channel-output feedback wth respect to R = R +R = R +R = = R + R = 4 R + R = Fg.. Capacty regon C(,,,,, ) wthout feedback (thck red lne) and C(,,,,, 4) wth nosy channel-output feedback (thn blue lne) of the example n Sec. III-C. Note that (,,,,, 4) = bts/ch.use, (,,,,, 4) = bts/ch.use and Σ(,,,,, 4) = bts/ch.use. the case wthout feedback s: = ( n, n, n, n, n, n ) = () max R j> sup (R, R j ) C (R, R j) C R R, and the maxmum sum rate mprovement Σ( n, n, n, n, n, n ) wth respect to the case wthout feedback s: Σ( n, n,n, n, n, n ) = () sup (R, R ) C (R, R ) C R + R (R + R ), where C = C( n, n, n, n, n, n ) and C = C( n, n, n, n,, ) are the capacty regon wth nosy channel-output feedback and wthout feedback, respectvely. The followng descrbes partcular scenaros that hghlght some nterestng observatons. ) Example : Only one channel-output feedback lnk allows smultaneous maxmum mprovement of both ndvdual

!! n = ;! n n ; n = ; n =. = ;! n = ; n = ; n =.! n = ;! n = ; n = ; n =.! n = ;! n = ; n = ; n =8.! n = ;! n = ; n = ; n =8.! n = ;! n = ; n = ; n =8. 4. 4 n n! n n 4 n =,! n =,n =,n =8, n =9, n =4. n 4! n = ;! n = ; n = ; n =8. 8 n n n 8 9 Fg.. Maxmum mprovement of ndvdual rates of the example n Sec. III-C = R +R = R +R = = R + R = X = R + R = 8 9 n! n n =,! n =,n =6,n 4 =, n =, n = Fg.. Maxmum mprovement of one ndvdual rate and the sum rate of the example n Sec. III-C R + R = R + R = R + R = Fg. 4. Capacty regon C(,,, 8,, ) wthout feedback (thck red lne) and C(,,, 8, 9, 4) wth nosy channel-output feedback (thn blue lne) of the example n Sec. III-C. Note that (,,, 8, 9, 4) = bt/ch.use, (,,, 8, 9, 4) = bt/ch.use and Σ(,,, 8, 9, 4) = bt/ch.use. R + R = 4 =. R + R = = rates: Consder the case n whch transmtter-recever pars and are n weak and moderate nterference regmes, wth n =, n =, n =, n =. In Fg. the capacty regon s plotted wthout channel-output feedback and wth nosy channel-output feedback ( n =, n = 4). In Fg., (,,,, n, n ) s plotted for both = and = as a functon of n and n. Theren, t s shown that: (a) Increasng parameter n beyond threshold n = allows smultaneous mprovement of both ndvdual rates ndependently of the value of n. Note that n the case of perfect channel-output feedback,.e., n = max ( n, n ), the maxmum mprovement of both ndvdual rates s smultaneously acheved even when n =. (b) Increasng parameter n beyond threshold n = provdes smultaneous mprovement of both ndvdual rates. However, the mprovement on the ndvdual rate R strongly depends on the value of n. (c) Fnally, the sum rate does not ncrease by usng channel-output feedback n ths case. ) Example : Only one channel-output feedback lnk allows maxmum mprovement of one ndvdual rate and the sum-rate: Consder the case n whch transmtter-recever pars and are n very weak and moderate nterference regmes, wth n =, n =, n =, n = 8. In Fg. 4 the capacty regon s plotted wthout channel-output feedback and wth nosy channel-output feedback( n = 9, n = 4). In Fg., (,,, 8, n, n ) s plotted for both = and = as a functon of n and n. Theren, t s shown that: (a) Increasng n beyond threshold n = 8 or ncreasng n beyond threshold n = allows smultaneous mprovement of both ndvdual rates. Fg. 6. Capacty regon C(,, 6,,, ) wthout feedback (thck red lne) and C(,, 6,,, ) wth nosy channel-output feedback (thn blue lne) of the example n Sec. III-C. Note that (,, 6,,, ) =. bts/ch.use, (,, 6,,, ) = bts/ch.use and Σ(,, 6,,, ) = bts/ch.use. Nonetheless, maxmum mprovement on R s acheved by ncreasng n. (b) Increasng ether n or n beyond thresholds n and n! n, allows maxmum mprovement of = ;! n = ; n = 6; n =.! the sum rate (see Fg. ). n = ;! n = ; n = 6; n =. 4 n 8 n Fg.. Maxmum mprovement of one ndvdual rate of the example n Sec. III-C ) Example : At least one channel-output feedback lnk does not have any effect over the capacty regon: Consder the case n whch transmtter-recever pars and are n the weak nterference regme, wth n =, n =, n = 6, n =. In Fg. 6 the capacty regon s plotted wthout channel-output feedback and wth nosy channeloutput feedback ( n =, n = ). In Fg., (,, 6,, n, n ) s plotted for both = and = as a functon of n and n. Theren, t s shown that: 6 4 n 8 n

! n =,! n =8,n =,n =, n =, n =9 R = = R + R = IC wth and wthout channel-output feedback are dentcal,.e., nether n nor n enlarges the capacty regon. In ths specfc example, t s not possble to take advantage of the feedback lnks for usng nterference as sde nformaton or to generate an alternatve path. Fg. 8. Capacty regon C(, 8,,,, ) wthout feedback (thck red lne) and C(, 8,,,, 9) wth nosy channel-output feedback (thn blue lne) of the example n Sec. III-C4. Note that (, 8,,,, 9) = bts/ch.use, (, 8,,,, 9) = bts/ch.use and Σ(, 8,,,, 9) = bts/ch.use. (a) Increasng parameter n does not enlarge the capacty regon, ndependently of the value of n. (b) Increasng parameter n beyond threshold n = 8 allows smultaneous mprovement! of both ndvdual rates. (c) Fnally, none of the parameters n = ;!! n = ;! n = 8; n = ; n =. n or n = 8; n = ; n =. n ncreases the sum-rate n ths case. 6 n n = n 8 n Fg. 9. Maxmum mprovement of one ndvdual rate of the example n Sec. III-C4 4) Example 4: The channel-output feedback of lnk exclusvely mproves R j : Consder the case n whch transmtter-recever pars and are n the very strong and strong nterference regmes, wth n =, n = 8, n =, n =. In Fg. 8 the capacty regon s plotted wthout channel-output feedback and wth nosy channel-output feedback ( n =, n = 9). In Fg. 9, (, 8,,, n, n ) s plotted for both = and = as a functon of n and n. Theren, t s shown that: (a) Increasng parameter n beyond threshold n = 8 exclusvely mproves R. (b) Increasng parameter n beyond threshold n = exclusvely mproves R. (c) None of the parameters n or n has an mpact over the sum rate n ths case. Note that these observatons are n lne wth the nterpretaton of channel-output feedback as an altrustc technque, as n [], [4]. Ths s bascally because the lnk mplementng channel-output feedback provdes an alternatve path to the nformaton sent by the other lnk, as frst suggested n []. ) Example : None of the channel-output feedback lnks has any effect over the capacty regon: Consder the case n whch transmtter-recever pars and are n the very weak and strong nterference regmes, wth n =, n = 9, n =, n =. Note that the capacty regon of the LD- IV. CONCLUSIONS In ths paper, the nosy channel-output feedback capacty of the lnear determnstc nterference channel has been fully characterzed by generalzng exstng results. Based on specfc asymmetrc examples, t s hghlghted that even n the presence of nose, the benefts of channel-output feedback can be sgnfcantly relevant n terms of achevable ndvdual rate and sum-rate mprovements wth respect to the case wthout feedback. Nonetheless, there also exst scenaros n whch these benefts are totally nexstent. REFERENCES [] D. Tunnett, On nterference channel wth generalzed feedback (IFC- GF), n Proc. of Internatonal Symposum on Informaton Theory (ISIT), Nce, France, Jun., pp. 66 66. [], An outer bound regon for nterference channels wth generalzed feedback, n IEEE Informaton Theory and Applcatons Workshop (ITA), Feb., pp.. [] C. Suh and D. N. C. Tse, Feedback capacty of the Gaussan nterference channel to wthn bts, IEEE Transactons on Informaton Theory, vol., no., pp. 66 68, May.. [4] A. Vahd, C. Suh, and A. S. Avestmehr, Interference channels wth ratelmted feedback, IEEE Transactons on Informaton Theory, vol. 8, no., pp. 88 8, May.. [] S. Yang and D. Tunnett, Interference channel wth generalzed feedback (a.k.a. wth source cooperaton): Part I: Achevable regon, IEEE Transactons on Informaton Theory, vol., no., pp. 686, May.. [6] S. Mohajer, R. Tandon, and H. V. Poor, On the feedback capacty of the fully connected-user nterference channel, IEEE Transactons on Informaton Theory, vol. 9, no., pp. 86 88, May.. [] V. M. Pabhakaran and P. Vswanath, Interference channel wth source cooperaton, IEEE Transactons on Informaton Theory, vol., no., pp. 6 86, Jan.. [8] D. Tunnett, An outer bound for the memoryless two-user nterference channel wth general cooperaton, n IEEE Informaton Theory Workshop (ITW), Sep., pp.. [9] A. Saha, V. Aggarwal, M. Yuksel, and A. Sabharwal, Capacty of all nne models of channel output feedback for the two-user nterference channel, IEEE Transactons on Informaton Theory, vol. 9, no., pp. 69 699,. [] S. M. Perlaza, R. Tandon, H. V. Poor, and Z. Han, Perfect output feedback n the two-user decentralzed nterference channel, arxv:6.88,. [] S.-Q. Le, R. Tandon, M. Motan, and H. V. Poor, Approxmate capacty regon for the symmetrc Gaussan nterference channel wth nosy feedback, IEEE Transactons on Informaton Theory, vol. 6, no., pp. 6, Jul.. [] V. Quntero, S. M. Perlaza, and J.-M. Gorce, Nosy channel-output feedback capacty of the lnear determnstc nterference channel, INRIA Grenoble - Rhône-Alpes, Tech. Rep. 46, Jan.. [] R. A. Berry and D. N. C. Tse, Shannon meets Nash on the nterference channel, IEEE Transactons on Informaton Theory, vol., no., pp. 8 86, May.. [4] S. M. Perlaza, R. Tandon, and H. V. Poor, Symmetrc decentralzed nterference channels wth nosy feedback, n Proc. IEEE Intl. Symposum on Informaton Theory (ISIT), Honolulu, HI, USA, Jun. 4. [] G. Bresler and D. N. C. Tse, The two user Gaussan nterference channel: A determnstc vew, European Transactons on Telecommuncatons, vol. 9, no. 4, pp. 4, Apr. 8.