WorksComputer Science
Hotplug Coded Caching: Uncoded versus Row-Wise Coded Placements
Abstract
In the hotplug coded caching model, users fill their caches of size as a function of files, but only an unknown subset of users is active in the delivery phase. In all regimes where the optimal hotplug tradeoff is known, it depends only on , not on ; whether this holds in general is open. We show that: (i) Under uncoded placement, YMA delivery with MAN placement, where offline users are assigned the demands of active users, is optimal. Hence, offline users cause a strict load penalty at intermediate memory sizes, while the MAN placement is optimal for every number of active users; (ii) Under row-wise placement (RWP), where each cache is a function of the user's own row of each file, every classical scheme can be embedded in a Reed-Solomon code. The resulting hotplug scheme has the tradeoff of the classical model with users, for every : there is no load penalty, but the placement depends on . We then derive a converse for RWP, which shows that: (a) for , the scheme of Gomez-Vilardebo (GV) is RWP optimal for , which settles the RWP tradeoff for ; (b) YMA is RWP optimal at ; and (c) for , GV is RWP optimal for . Hence, the Reed-Solomon embeddings of these schemes are optimal among embedded RWP schemes.
Introduction
Classical coded caching trades off network load for storage at the end users. Maddah-Ali and Niesen [1] introduced a shared-link model with cache-aided users, and proposed a scheme with an uncoded combinatorial placement and a coded delivery (referred to as MAN), which is to within a constant factor of a cut-set bound. Let be the number of users, the number of files, and the cache size. Under the constraint of uncoded placement, the exact worst-case memory-load tradeoff was characterized by Wan et al. [2] for , through an acyclic index coding converse, and by Yu et al. [3] for all and ; the YMA scheme of [3] is also optimal to within a factor of two without restrictions on the placement [4]. Coded placement strictly improves on uncoded placement at small memory sizes: the scheme of Chen, Fan and Letaief (CFL) [5] is optimal for when , and the scheme of Gómez-Vilardebó (GV) [6] is optimal for when [7]. Linear coded placement schemes based on rank-metric codes were developed in [8, 9], which outperform YMA and GV in some regimes; [10, 19] derived computer-aided converse bounds.
The classical model requires every user that took part in the placement phase to also take part in the delivery phase. In [11] we introduced the hotplug model, in which only of the users are active in the delivery phase, and the identity of the active users is unknown at the time of placement. In [11, 12] we proposed schemes based on Maximum Distance Separable (MDS) coded placement that are exactly optimal in the small and large memory regimes, and optimal to within a factor of two otherwise. In all cases where we could establish exact optimality, the optimal hotplug tradeoff depends only on the number of active users , and not on the total number of users ; whether this holds in general was left open. Hotplug placement delivery arrays were proposed in [13, 14]; their algorithms provide examples of tradeoffs that outperform the baseline scheme in [11], but no closed-forms are available.
Models of random users’ inactivity have also been proposed [15, 16], which cannot be compared to hotplug. Decentralized placement [17] is compatible with hotplug, since the placement does not depend on which users, or how many, will be active; the decentralized load is the limit of the centralized load when the number of users tends to infinity.
Main Contributions
(1) Uncoded placement. We show that the extension of YMA in [11] (referred to as YMA+), in which the offline users are assigned the demands of the active users, is optimal under uncoded placement (Theorem 1). Hence, under uncoded placement and when , a system with active users pays a strict load penalty at intermediate memory sizes with respect to a classical system with users (Corollary 1). Moreover, the MAN placement is universal: it is optimal for every number of active users (Corollary 2). The penalty is thus in the load only: the active users do not see the symmetric MAN placement structure of users, but the restriction of a -user structure to an unknown subset. Hence, matching the classical load with users requires coded placement.
(2) Reed–Solomon embedding of RWP placement (RWP). In an RWP for a classical system (i.e., ), each file is a matrix with rows, and the cache of a user is a function of its row of every file; the caches are thus independent and . We show that every classical RWP scheme can be embedded in a Reed–Solomon code, which yields a hotplug scheme for every whose memory-load tradeoff is the same as that of the classical model (Theorem 2), i.e., there is no load penalty for offline users. This greatly generalizes some schemes of [11] and gives a tradeoff complementary to that of uncoded placement: the placement depends on both and , but the load depends only on $A. (3) Converse for RWP placement. We then ask how good RWP is. We derive a converse bound for RWP in the classical model (Theorem 3), in the spirit of the acyclic index coding bound for uncoded placement [2]. It shows that, for every , the GV scheme is optimal among RWP schemes for ; thus the RWP tradeoff is fully characterized for . It also shows that the YMA point with one cached subfile per file, at , is RWP optimal for every and (Corollary 4). For , a separate converse shows that the GV scheme is optimal among RWP schemes for (Theorem 4). Consequently, the Reed–Solomon embeddings of these schemes are optimal among embedded RWP schemes for every (Corollary 5).
Notation
Calligraphic symbols denote sets, bold symbols vectors and matrices, and sans-serif symbols system parameters. unless . For integers , and . For a set , . For a set , , and is the sub-vector of restricted by .
Problem Setting
An hotplug model consists of a server with files and cache-aided users connected to the server via a shared link. File , , contains i.i.d. uniformly distributed symbols over , for a prime power . In the placement phase, the server fills , the cache of user , as
In the delivery phase, a set of users is active, unknown at the time of placement; active user requests file . The server sends as
Every active user decodes its demanded file as
The optimal worst-case load is the infimum of the achievable over all , and schemes that satisfy (1), (2) and (3) for every and . The case is the classical MAN model with users. We let .
Uncoded placement: If every cache is a subset of the symbols of the files, the placement is said to be uncoded, and the optimal worst-case load is .
RWP placement: For a classical system with users, a placement is RWP if each file is split into rows, , , and the cache of user is a function of the -th row of every file, i.e., , where using the same function at every user is without loss of generality [10]. Since a row has entropy , RWP is restricted to . The optimal RWP load is denoted by . RWP may be uncoded, and uncoded placement can be RWP, thus, no general ordering between and exists.
Known Results
YMA+ (uncoded) scheme MT2022offline for hotplug
For users and , the MAN placement splits , , and user stores for all and ; hence, . For , YMA+ [11] assigns to the offline users the demand of an active user and runs the YMA delivery [3] for all users, with at most distinct demands. YMA+ thus achieves the lower convex envelope of
denoted by . For this is YMA, which is optimal under uncoded placement [2, 3]. For every set of active users, the caches of the active users and the delivery form a classical scheme for users, hence
where the lower bounds depend only on , and the inequalities marked as (i) and (ii) are from [4], with
being the load with decentralized placement [3, 17]. The chain of inequalities in (5) shows that YMA+ is optimal to within a factor of two [12].
GV (row-wise) scheme gomez2018fundamental for classical coded caching
The GV scheme [6] for users is an RWP scheme that further splits each row into pieces indexed by the -subsets of and caches sums of pieces of distinct files of the same row; it achieves the lower convex envelope of and the points
For , this is the CFL point [5], which caches . For , this is the YMA point with , , which caches the row itself. The line through and the CFL point is optimal for [5]. The line through the CFL point and the GV point with is optimal for [6] and for [7].
MDS-coded placement MT2022offline,MT2024arxiv for hotplug
Our past work showed that the following points of a classical system with users are achievable in the hotplug model for every : the CFL point for [11], Scheme new2 (HT2 in [12]), the YMA point with [11], Scheme new1 (HT1 with in [12]), and the YMA point with for [12], Scheme HT3. Together with (5), these results imply whenever the latter is attained by one of these points. In [11], Scheme new1 we also added an MDS layer to the MAN placement with , which attains the YMA load of a classical system with users, but with a memory size that depends on both and .
Main Results
We state next our main results, with proof sketches; the full proofs can be found in Appendix.
Uncoded placement for hotplug
Theorem 1 (Optimality of YMA+). For an hotplug model, for all .
Proof Sketch: For the converse, we apply the acyclic index coding bound of [2] to the active users only and average it over the set of active users, their ordering and their distinct demands, which gives , where is the fraction of the library cached by exactly of the users; minimizing over the distribution subject to , as in [2, 3], gives the YMA+ load.
Corollary 1 (Penalty of offline users). For an hotplug model under uncoded placement, we have:
(C1.I) if and , then for every
(C1.II) for , i.e., the dependance on is in the mid-memory regime.
Corollary 2 (Universality of the MAN placement). For an hotplug model, the MAN placement followed by YMA+ for the users that turn out to be active, attains simultaneously for every .
In words, for hotplug under uncoded placement, not knowing which users will be active has a load cost (Corollary 1), whereas not knowing how many has none (Corollary 2). Since the inequality in (7) is strict, an optimality result for the hotplug model that matches the classical lower bound in (5) necessarily requires coded placement.
Reed–Solomon embedding of RWP for hotplug
Theorem 2 (Reed–Solomon embedding). Let a classical scheme for users with RWP, over a field , achieve , with possibly nonlinear cache placement and delivery functions. Then is also achievable in the hotplug model for every , over any extension field of with .
Proof Sketch: The main idea is to view each file as a matrix , . Let be the standard basis vector with a one in position . In the classical model with RWP, the cache contents are , . In hotplug, users are assigned the “MDS vectors”
where are distinct, which requires ; users cache , , thus same cache size as in the classical model. Any of the vectors in (8) are linearly independent (Vandermonde). Once the set of active users is known, let be the invertible matrix with rows , and define the library , . Since , the cache is the cache of classical user for the library . The server then uses the classical transmission for the library and demands ; user decodes and recovers .
Remark 1. Under uncoded placement, the optimal placement does not depend on the number of active users (Corollary 2), but the load depends on (Corollary 1). The Reed–Solomon (RS) embedding has the opposite behavior: the load depends only on , but the placement function depends on (the number of rows, i.e., the dimension of the RS code) and on (the length of the RS code). If only users turn out to be active, the server can still deliver to any -subset of users that contains them, with load , but this is in general not optimal.
Remark 2. Theorem 2 applied to the GV scheme in (6) gives the achievability of the GV points of a classical system with users, for every . It recovers HT2 and HT1 with of [12]; HT1 with and HT3 are not RWP. Theorem 2 is more general than [12]: (a) it applies to every RWP scheme, with a linear or nonlinear placement and delivery; and (b) any RWP scheme that improves on GV in the open range of Corollary 4 immediately yields an improved hotplug scheme in the same regime. For example, in the scheme of Amiri and Gündüz [18], user caches , , at ; this is an RWP, hence it yields a hotplug scheme with the same load for every . In this case the embedding gives no new point, since the scheme of [18] is strictly dominated by GV for , and it coincides with CFL for .
Converse for RWP in the classical model
Motivated by the ability to embed any classical RWP scheme in the hotplug model for any , we next study the fundamental limits of RWP for uncoded caching with users. For such a classical system, let for ; by the symmetry argument in [10], without loss of generality does not depend on the user or on the subset . For every placement, is non-increasing in (conditioning reduces entropy) and concave in (submodularity of entropy), and
where, without loss of generality, : otherwise the unused memory can be filled with uncoded symbols of the user’s row, which preserves the RWP structure and does not increase the load.
Theorem 3 (RWP converse). For an classical coded caching model with ,
where is the set of concave that satisfy (9) and
Proof Sketch: Fix , distinct users requesting distinct files ; let and . Then, for every
The bound in (11) follows from the RWP structure. □
Theorem 3 plays the role of the acyclic index coding bound for uncoded placement in [2]: the conditional cache entropies replace the amounts of data cached by a given number of users, and the minimization over replaces the optimization over the placement distribution. For , the optimization in (10) has a closed form.
Corollary 3 (Theorem 3 for ). For an classical coded caching model with ,
i.e., piecewise linear among , .
Remark 3. For , (10) is a linear program in that does not appear to have a simple closed-form solution, but it is easy to solve numerically.
Corollary 4 (RWP optimality). For an classical model we have: (C4.I) If ,
i.e., the GV scheme is RWP optimal on the segment between the points with and , for every .
(C4.II) *For and ,
(C4.III) *For , and every and ,
with equality at , attained by with .
Corollary 4 has the following consequences:
(C4.I) extends the optimality of the GV point with , known among all placements for [7], to every among RWP schemes. Its optimality without placement restrictions for remains open.
For , the tradeoff without restrictions on the placement is known for [1], [10], while for it is open on , except that for it is known on [8], [10]; in all cases the open range contains . (C4.II) characterizes the optimal RWP tradeoff; for its optimality without placement restrictions, see § 5.
(C4.III) holds for every and , i.e., also for .
The corner points of (13) are at , , while the GV points are at , ; hence a GV point has the same memory size as a corner point of (12) if and only if . In this case, a direct computation gives they have the same load if and only if (the CFL point, ) or (the YMA point with , ). For example, for and , the GV point lies above the converse. The GV point with is RWP optimal even though, for , its memory size is not a corner point of (12): it lies on the segment of (12) with .
For , the right-hand side of (12) lies strictly below the lower convex envelope of the GV points for , where the RWP optimal load remains open in general; the case is settled by Theorem 4 next.
Optimality of RWP in the hotplug model
Theorem 4 (RWP optimality for ). For and ,
i.e., the GV scheme is optimal among RWP schemes, linear or not; its corner points are , , and . In particular, no RWP scheme attains .
Proof Sketch: The first two terms in (14) follow from Corollary 3, which on coincides with the outer bound of Tian [10] for , valid for every placement. The third term, , is a separate RWP converse, strictly tighter than Tian’s bound on ; it follows by summing Shannon inequalities that involve only , one cache, whole files, and one row of at most two files, together with decoding, the RWP identities, and user–file symmetry. Since , the linear scheme of Cao and Xu [19] at , which is optimal among linear schemes, is not RWP, and no RWP scheme attains the same point¹. □
Corollary 5 (Optimal Reed–Solomon embedded schemes). Let be the minimum load over the hotplug schemes obtained from a classical RWP scheme for users by the embedding of Theorem 2. Then for every and every ; in particular, these loads do not depend on , and they are given by Corollary 4 and Theorem 4 wherever these characterize .
Examples
We provide two examples to further clarify the relationship between various achievable schemes.
a) Case : Figure 1 shows the various tradeoffs in the small memory regime for . Under uncoded placement (Theorem 1), the optimal load at is for , for , and for . The Reed–Solomon embedding of the GV scheme for 3 users (Theorem 2) attains the points , and for every . By Theorem 4, memory sharing among these points is RWP optimal on , for every . On , Corollary 3 coincides with the converse of Tian [10], which holds for every placement, and it is tight only on and at .
Figure 1. Hotplug model with : optimal load under uncoded placement for (Theorem 1) and its limit for , i.e., the load of decentralized placement; load of the Reed–Solomon embedding of the GV scheme for 3 users, which does not depend on (Theorem 2) and is RWP optimal (Theorem 4); the point of Cao and Xu [19] for the classical system, which is optimal under linear coding, with memory sharing with and ; and Corollary 3, which on coincides with the converse of [10] for every placement.
The optimal tradeoff under linear coding for the classical system is known [19]: its corner points are at , where is attained by the non-RWP scheme of Cao and Xu [19]. With the schemes known to us, the hotplug model matches it for , by Theorem 2 and HT3 [12]. Whether the optimal load under linear coding of the hotplug model at equals for is thus open; it hinges on whether this point can be attained in the hotplug model, for example by embedding the scheme of Cao and Xu [19]. By Theorem 4, such an embedding cannot be obtained from a RWP scheme.
b) Case and : For the classical model with files, the optimal tradeoff is known for [1, 10], and all its corner points are achievable in the hotplug model for every , by Theorem 2 and HT3 [12]. For , which is not covered by [7], the classical tradeoff is known only for [5] and [10 Theorem 4.12], and it is matched in the hotplug model in both ranges: for by Theorem 2, and for by the YMA points of a classical system with users with [12 Scheme HT3] and .
For (see Figure 2), the scheme of Tian and Chen [8] attains , which meets the converse of [10]; hence the classical tradeoff is known for , whereas, with the schemes known to us, the hotplug model matches it for . In particular, in the hotplug model at with , , YMA+ attains , which increases to as ; memory sharing between the embedded YMA points with and attains ; and memory sharing between the CFL point (Theorem 2) and the embedded YMA point with attains , for every . Whether for is thus open; it hinges on whether this point can be attained in the hotplug model, for example by embedding the scheme of [8].
Figure 2. Hotplug model with and : optimal load under uncoded placement for (Theorem 1) and its limit for ; load of the Reed–Solomon embedding of the YMA points for users with , which equals the RWP optimum (Corollary 4) and does not depend on (Theorem 2); load of the non-RWP scheme of Tian and Chen [8] for the classical system with 4 users; and the converse of [10] for the classical system, which holds for every placement.
Conclusions
Under uncoded placement, YMA+ is optimal for the hotplug model: the MAN placement is optimal for every number of active users, and offline users strictly increase the load at intermediate memory sizes. The Reed–Solomon embedding of RWP schemes gives the complementary tradeoff, with a load that depends only on the number of active users. Our converse for RWP establishes the RWP optimality of the GV scheme at its two smallest memory points for every , characterizes the RWP tradeoff for and for , and establishes the RWP optimality of the YMA point with for all parameters. Open questions are the RWP tradeoff for when (except ), and whether the optimal load of the hotplug model, without any restrictions on the placement, depends only on and not on . Our results show that, if it does, the optimal schemes must use coded placement.
Acknowledgment
This work was funded in part by NSF under award 2312229.
Use of AI Disclosure
The Author used Anthropic’s Claude (Opus 5.5) during the development and preparation of this manuscript. It assisted in: numerically solving linear programs to search for row-wise schemes and converse bounds, including the extraction of the sparsest Shannon-inequality certificate underlying the proof of Theorem 4; numerically checking the closed-form expressions in the examples; generating the figures; and polishing the text. The Author verified every result, proof, statement and reference in the final manuscript and assumes full responsibility for its contents.
References
Appendix
Proof of Theorem 1
Achievability is YMA+ in (4). By Pascal’s rule, , and is non-increasing and convex; hence is the piecewise linear interpolation of . For the converse, fix a scheme with uncoded placement and partition each file as , where is cached exactly by the users in . Let be the fraction of the library cached by exactly users; then and .
Fix , an ordered list of distinct users in , and demands with distinct . Let , and let be the rest of the library. Then is a function of , and by (3) and induction on , . Hence
the acyclic index coding bound [2] applied to the active users only. Average (15) over uniform in , uniform given , and the distinct demands uniform. Then is uniform in and avoids a given with probability , so that by Jensen’s inequality and monotonicity.
Proof of Corollary 1
(C1.I): For real , let ; it is convex, non-increasing, and equals at integers, so . At , for , when ; hence , strictly for .
(C1.II): On the load is , and on it is , for every .
Proof of Corollary 2
The MAN placement does not depend on , YMA+ attains with , and the abscissas of the corner points of do not depend on .
Proof of Theorem 2
Since the classical scheme is correct for every realization of the library, running instances in parallel we may assume that every row is a vector over , . The vectors in (8) are the columns of the generator matrix of a (extended) Reed–Solomon code, so that any of them are linearly independent: for vectors with the determinant is a Vandermonde determinant, and if is included, expanding along it leaves an Vandermonde determinant.
Placement. The coded row of file is , and user stores , of the same size as the classical cache.
Delivery. Given , the library , where acts on the rows of , consists of independent and uniformly distributed files, since is invertible. Then is the cache of classical user for the library . The server sends and the classical transmission for the library and the classical demands , of load .
Decoding. User runs the decoder of classical user , obtains , and computes . The map acts on each file separately, which is why the demanded file can be recovered.
Properties of RWP
Lemma 1 (Symmetrization). Every RWP scheme for users, with possibly different cache functions , can be converted into a RWP scheme with the same cache function at every user, and the same memory and load.
Proof. Split every row subfile into pieces, , and let every user store of its own row. For , run an instance of the original scheme in which user plays the role of user (mod $A), on the pieces ; within each instance the roles are distinct, so every instance is a copy of the original scheme, with $1/A$ of the file size.
Lemma 2. For the RWP, (11) holds, and
Proof. Fix and with , and let . Since is a function of and all row subfiles are independent, is independent of ; hence
Upper bound in (11). Conditioning reduces entropy, so . Moreover, given , is a function of , which is independent of ; hence .
Lower bound in (11). Clearly, , and by (17), .
Equality (16). Given , the caches are functions of distinct rows of the files not in , which are independent.
Proof of Theorem 3
Fix , distinct users requesting distinct files , and let , , , . Then , with by (16), and , where the equality uses the chain rule and , and the last step uses (16). The resulting bound holds for every , and the entropies of the symmetrized scheme belong to by (9) and Lemma 2.
Proof of Corollary 3
With and , the bracket in (10) is , and by (9) and (11); substituting gives (13). With , the sum equals , which gives the corner points. The bound is the exact value of (10): belongs to , and, writing the bracket as with , each term is non-negative for , so that the maximum over is attained at .
Proof of Corollary 4
(C4.I): For and , (13) gives the two lines. They are attained by memory sharing among , the CFL point , and the GV point with , , which lies on the second line, with .
(C4.II): For , , so item (C4.I) covers the whole range.
(C4.III): Take in (10) and let . Telescoping gives . By concavity, is non-decreasing in ; with and by (11), for every , . Hence , i.e., (14), for every . The YMA point with attains it at .
Proof of Corollary 5
By Theorem 2, every RWP load is achievable in the hotplug model for every . Conversely, in a hotplug scheme obtained by embedding an RWP, the active users form a classical RWP scheme for the library , whose files are again independent and uniformly distributed; hence its load is at least , for every .
Proof of Theorem 4
We use the following notation. Files , , have unit size; row of file is , of size , and denotes . User caches , a function of , and , so . Consider the demand in which users , , request , , , with transmission , . By time sharing over the joint permutations of users/rows and files, we may assume that every entropy below is invariant under these permutations. The facts used are:
decoding: is a function of , of , of ;
RWP identities (caches depend only on their own row, rows are independent):
;
Only , one cache, whole files and one row of at most two files appear. Let
We bound in three ways, each obtained by completing differently.
(i) Through user 2. Since and is decoded from ,
where the last step swaps users (and files ). Next, gives
and . Therefore
(ii) Through row 2. gives . Moreover, is decoded from , so
where the equality applies the cyclic shift of users/rows with , . Therefore
(iii) Through row 1. , hence, swapping users ,
Summing. The terms and cancel, and
Swapping users (and files ), and using that determines and conversely determines ,
Hence . Finally , so , and .
The proof is complete, with four multipliers, of 14 Shannon inequalities (a sparse dual solution of the genie LP restricted to single-row genies); the three bounds (i)–(iii) are the three ways the LP completes to the whole library. No pair of caches and no two-row genie is needed.
References
- [1]M. A. Maddah-Ali and U. Niesen, “Fundamental limits of caching,” IEEE Trans. Inf. Theory, vol. 60, no. 5, pp. 2856–2867, 2014. doi: https://doi.org/10.1109/TIT.2014.2306938
- [2]K. Wan, D. Tuninetti, and P. Piantanida, “An index coding approach to caching with uncoded cache placement,” IEEE Trans. Inf. Theory, vol. 66, no. 3, pp. 1318–1332, 2020. doi: https://doi.org/10.1109/TIT.2020.2967753.
- [3]Q. Yu, M. A. Maddah-Ali, and A. S. Avestimehr, “The exact rate-memory tradeoff for caching with uncoded prefetching,” IEEE Trans. Inf. Theory, vol. 64, no. 2, pp. 1281–1296, 2018. doi: https://doi.org/10.1109/TIT.2017.2785237.
- [4]Q. Yu, M. A. Maddah-Ali, and A. S. Avestimehr, “Characterizing the rate-memory tradeoff in cache networks within a factor of 2,” IEEE Trans. Inf. Theory, vol. 65, no. 1, pp. 647–663, 2019. doi: https://doi.org/10.1109/TIT.2018.2870566.
- [5]Z. Chen, P. Fan, and K. B. Letaief, “Fundamental limits of caching: Improved bounds for users with small buffers,” IET Commun., vol. 10, no. 17, pp. 2315–2318, 2016. doi: https://doi.org/10.1049/iet-com.2015.1205.
- [6]J. Gómez-Vilardebó, “Fundamental limits of caching: Improved rate-memory tradeoff with coded prefetching,” IEEE Trans. Commun., vol. 66, no. 10, pp. 4488–4497, 2018. doi: https://doi.org/10.1109/TCOMM.2018.2834364.
- [7]V. Kumar K. P., B. K. Rai, and T. Jacob, “The exact rate memory tradeoff for small caches with coded placement,” arXiv:2102.04797, 2021.
- [8]C. Tian and J. Chen, “Caching and delivery via interference elimination,” IEEE Trans. Inf. Theory, vol. 64, no. 3, pp. 1548–1560, 2018. doi: https://doi.org/10.1109/TIT.2018.2794543
- [9]K. Zhang, “Fundamental limits of caching: Symmetry structure and coded placement schemes,” Ph.D. dissertation, Dept. Elect. Comput. Eng., Texas A&M Univ., College Station, TX, USA, 2020.
- [10]C. Tian, “Symmetry, outer bounds, and code constructions: A computer-aided investigation on the fundamental limits of caching,” Entropy, vol. 20, no. 8, p. 603, 2018. doi: https://doi.org/10.3390/e20080603.
- [11]Y. Ma and D. Tuninetti, “On coded caching systems with offline users,” in Proc. IEEE ISIT, pp. 1133–1138, 2022. doi: https://doi.org/10.1109/ISIT50566.2022.9834866
- [12]Y. Ma and D. Tuninetti, “On coded caching systems with offline users, with and without demand privacy against colluding users,” arXiv:2401.06894, 2024.arxiv.org/abs/2401.06894
- [13]C. Rajput and B. S. Rajan, “Improved hotplug caching scheme using PDAs,” in Proc. IEEE ISIT, pp. 1574–1579, 2024. doi: https://doi.org/10.1109/ISIT57864.2024.10619483.
- [14]M. Chinnapadamala, C. Rajput, and B. S. Rajan, “A new hotplug coded caching scheme using PDAs,” in Proc. IEEE ITW, pp. 478–483, 2024. doi: https://doi.org/10.1109/ITW616385.2024.10806946
- [15]J. Liao and O. Tirkkonen, “Coded caching in presence of user inactivity,” in Proc. IEEE WCNC, pp. 1437–1442, 2022. doi: https://doi.org/10.1109/WCNC51071.2022.9771827.
- [16]Ç. Yapar, K. Wan, R. F. Schaefer, and G. Caire, “On the optimality of D2D coded caching with uncoded cache placement and one-shot delivery,” IEEE Trans. Commun., vol. 67, no. 12, pp. 8179–8192, 2019. doi: https://doi.org/10.1109/TCOMM.2019.2940671.
- [17]M. A. Maddah-Ali and U. Niesen, “Decentralized coded caching attains order-optimal memory-rate tradeoff,” IEEE/ACM Trans. Netw., vol. 23, no. 4, pp. 1029–1040, 2015. doi: https://doi.org/10.1109/TNET.2014.2317316.
- [18]M. Mohammadi Amiri and D. Gündüz, “Fundamental limits of coded caching: Improved delivery rate-cache capacity tradeoff,” IEEE Trans. Commun., vol. 65, no. 2, pp. 806–815, Feb. 2017. doi: https://doi.org/10.1109/TCOMM.2016.2638841.
- [19]D. Cao and Y. Xu, “Characterizing linear memory-rate tradeoff of coded caching: The (N, K) = (3, 3) case,” arXiv:2005.05548, 2020.arxiv.org/abs/2005.05548

