This article addresses the challenge of reproducing machine learning (ML) experiments by integrating processes based on MLOps and semantic technologies. The inherent complexity of experimentation in scientific research hinders reproducibility through conventional methods, which has led to the need to automate processes. In this work, a solution has been developed allowing the execution of ML experiments of other researchers and their reproducibility. The use of semantic technologies allows the complete description of the experiment, including the data and resources necessary for its execution. The approach proposed in this work contributes to the automation of the experimentation phases based on MLOps, demonstrating how it can be used to reproduce experiments and offer a solution to the complexity of experimentation in scientific research. The effectiveness of the solution proposed in this work is evaluated by means of a survey-based analysis carried out among researchers who currently use manual processes to perform machine learning experiments. The results indicate that manual processing is prone to errors and not scalable regarding the size and complexity of most experiments. Moreover, the solution proposed in this work, which combines MLOps-based processes and semantic technologies, has been well received by researchers and considered to significantly improve the efficiency, reproducibility, and scalability of machine learning experimentation.

We present achievable error exponent regions for the Two-Way AWGN channel under an expected block power constraint and variable-length coding (VLC). We propose an achievability scheme that allows terminals to cooperate via interaction to detect decoding errors and request re-transmissions. Under this scheme, in certain rate-pair regimes both directions are able to simultaneously attain error exponent pairs larger than the feedback-free point-to-point random coding error exponents1.

Achievable error exponents for the one-way with noisy feedback and two-way AWGN channels are derived for the transmission of a finite number of messages M under almost sure (AS) and expected block (EXP) transmit power constraints. In the one-way setting under noisy AWGN feedback, under an AS power constraint, known linear and non-linear passive schemes are modified to incorporate AS constraints in the feedback link as well. In addition, a new active feedback scheme is presented in which the receiver feeds back the most likely pair of codewords, and the transmitter re-transmits which of these two was originally sent. This active feedback scheme outperforms one of the passive feedback schemes for all channel parameters; the linear scheme outperforms the others for low feedback noise variance. Under the EXP constraint, a known achievable error exponent for the transmission of two messages is generalized to any arbitrary but finite number of messages M through the use of simplex codes and erasure decoding. In the two-way AWGN setting, each user has its own message to send in addition to (possibly) aiding in the transmission of feedback for the opposite direction. Two-way error exponent regions are defined and achievable error exponent regions are derived for the first time under both AS and EXP power constraints. For the presented achievability schemes, feedback or interaction leads to error exponent gains in one direction, possibly at the expense of a decrease in the error exponents attained in the other direction. The relationship between M and n supported by our achievable strategies is explored.

We investigate error exponent regions for the parallel two-way DMC in which each terminal sends its own message and provides feedback to the other terminal. Various error exponents are presented in different rate-region regimes based on the relative rates and zero-error capacities of both directions. The schemes employed are extensions of error exponents for one-way DMCs with noiseless, rate-limited and noisy feedback1.

Achievable error exponents for two-way parallel discrete memoryless channels (DMC) using variable block length coding (VLC) are presented. First, Forney's erasure decoding error exponent is shown to be achievable for both directions simultaneously. Next, for some rate-pairs, it is shown that the error exponent of the direction with a smaller capacity may be further increased by allocating feedback resources to it in the other direction, at the price of a decreased error exponent for the other terminal. The presented two-way communication scheme builds upon Draper-Sahai's one-way DMC achievability scheme with noisy feedback under VLC. Both achievable error exponent regions demonstrate that the use of VLC and interaction between the terminals may benefit both directions' error exponents over fixed block length and feedback free transmission.1

A one-way additive white Gaussian noise (AWGN) channel with active feedback sent over another AWGN feedback channel is considered. Achievable error exponents are presented in the finite message / zero-rate regime for a variable length coding (VLC) scheme. This coding scheme uses a form of round-robin scheduling of messages, and a simplex-based feedback code to obtain reliable feedback and remain synchronized, despite the noise in the feedback link. Our results show that this new VLC scheme under an almost-sure power constraint achieves an error exponent similar to an achievable exponent attained using a fixed block length scheme under a much more relaxed expected block power constraint, and is larger than that achieved by schemes without feedback.

A primitive relay channel (PRC) has one source (S) communicating a message to one destination (D) with the help of a relay (R). The link between R and D is considered to be noiseless, of finite capacity, and parallel to the link between S and (R,D). Prior work has established, for any fixed number of channel uses, the minimal R-D link rate needed so that the overall S-D message rate equals the zero-error single-input multiple output outer bound (Problem 1). The zero-error relaying scheme was expressed as a coloring of a carefully defined “relaying compression graph”. It is shown here that this relaying compression graph for$n$channel uses is not obtained as a strong product from its$n$= 1 instance. Here we define a new graph, the “primitive relaying graph” and a new “special strong product” such that the n-channel use primitive relaying graph corresponds to the n-fold special strong product of the$n$= 1 graph. We show how the solution to Problem 1 can be obtained from this new primitive relaying graph directly. Further study of this primitive relaying graph has the potential to highlight the structure of optimal codes for zero-error relaying.

Achievable error exponent regions of a two-way additive white Gaussian noise (AWGN) channel, where two terminals exchange a fixed number of messages M, are derived. In particular, error exponent regions for M = 2 messages under expected power and M = 3 messages under almost sure power constraints are considered. For M = 2 messages the use of active feedback is shown to lead to an error exponent gain over that when feedback / interaction is ignored. For M = 3 messages and asymmetric channels, it is shown that the error exponent of the weaker channel may be improved through active feedback, at the expense of a decreased error exponent of the stronger direction. This may, for sufficiently asymmetric channel gains, outperform the error exponent region achieved by having both terminals operate independently of one another (ignoring the possibility of sending feedback for the other).

Interactive advertising based on multiple devices opens new possibilities for mobile applications, where users can search, select, or expand the information provided in advertising commercials by incorporating interactivity-friendly companion devices such as smart phones and tablets. In this paper, we derive the basic requirements for a flexible infrastructure that can support interactive ad applications. The infrastructure comprises a set of components and their externally visible properties, and the relationships between them. The main contributions of this work are first, establishing a desirable set of requirements for a suitable working scenario on which the different interactivity supporting systems people use in a regular basis can be easily integrated, and second, the definition of a platform which considers the different stages and requirements identified from different works in related areas.

In the domain of Digital Television (DTV) broadcasting technology, the enhancement of signals features over classic analog signal transmission allows increasing the amount of content available for TV viewers. Recommender Systems (RS) arose as a suitable choice to assist users in the overwhelming task of selecting audiovisual content, however, the cold-start problem normally associated to the lack of information in early RS stages, causes that user stereotyping approaches are employed meanwhile the lack of information in user profiles is overcome. This paper presents an experimental approach aimed to determine the best conditions for which users who were categorized within a determined stereotype during the cold-start stage, could migrate to a new state in which they receive personalized recommendations. Experimental results show that the best condition under the selected demographic stereotyping scheme for this transition is directly related to the number of TV programs that a user has rated while making use of the system.

Electronic Program Guides (EPGs) describe broadcast programming information provided by TV stations. However, users may obtain more information when these guides have been enriched. The main contribution of this work is to present an automation process for EPG's information enrichment through the use of semantic technologies and external resources. Among the several resources involved in the enrichment process, the following can be mentioned : ontologies, web services, semantic repositories and natural language processing techniques.