Advanced language models like BERT have great performances in various Natural Language Processing tasks. However, recent researches have shown that language models can learn gender biases from corpus, and lead to discriminatory decisions and unfair allocation of resources. We proposed a measure of gender bias in BERT resume retrieval system, by performing job searches on a group of resumes with different genders but consistent abilities. We proposed to calculate the average ranking and Discounted Cumulative Gain (DCG) scores of male and female resumes, and found that men outperformed women even though the two resumes were identical except for gender. This shows BERT has gender stereotypes, and its resume retrieval systems prefer male candidates. Therefore, we also proposed a regularized debiasing method to promote gender equity. Referring to Densifier method, we can get the subspace vector encoding bias semantics through matrix transformation of word vector difference between occupational and gender words. By defining the correlation between BERT word vector and gender bias subspace as the loss term, we can remove the bias semantics in BERT, and also avoid it learning stereotypes even trained on an unfair data set. After regularized debiased, the gender ranking gap of BERT was reduced by an average of 61.8%, while DCG scores were reduced by 53.9%.

Dominated by delay-sensitive and massive data applications, radio resource management in 5G access networks is expected to satisfy very stringent delay and packet loss requirements. In this context, the packet scheduler plays a central role by allocating user data packets in the frequency domain at each predefined time interval. Standard scheduling rules are known limited in satisfying higher quality of service (QoS) demands when facing unpredictable network conditions and dynamic traffic circumstances. This paper proposes an innovative scheduling framework able to select different scheduling rules according to instantaneous scheduler states in order to minimize the packet delays and packet drop rates for strict QoS requirements applications. To deal with real-time scheduling, the reinforcement learning (RL) principles are used to map the scheduling rules to each state and to learn when to apply each. Additionally, neural networks are used as function approximation to cope with the RL complexity and very large representations of the scheduler state space. Simulation results demonstrate that the proposed framework outperforms the conventional scheduling strategies in terms of delay and packet drop rate requirements.

Maintaining a desired trade-off performance between system throughput maximization and user fairness satisfaction constitutes a problem that is still far from being solved. In LTE systems, different tradeoff levels can be obtained by using a proper parameterization of the Generalized Proportional Fair (GPF) scheduling rule. Our approach is able to find the best parameterization policy that maximizes the system throughput under different fairness constraints imposed by the scheduler state. The proposed method adapts and refines the policy at each Transmission Time Interval (TTI) by using the Multi-Layer Perceptron Neural Network (MLPNN) as a non-linear function approximation between the continuous scheduler state and the optimal GPF parameter(s). The MLPNN function generalization is trained based on Continuous Actor-Critic Learning Automata Reinforcement Learning (CACLA RL). The double GPF parameterization optimization problem is addressed by using CACLA RL with two continuous actions (CACLA-2). Five reinforcement learning algorithms as simple parameterization techniques are compared against the novel technology. Simulation results indicate that CACLA-2 performs much better than any of other candidates that adjust only one scheduling parameter such as CACLA-1. CACLA-2 outperforms CACLA-1 by reducing the percentage of TTIs when the system is considered unfair. Being able to attenuate the fluctuations of the obtained policy, CACLA-2 achieves enhanced throughput gain when severe changes in the scheduling environment occur, maintaining in the same time the fairness optimality condition.

In LTE-A cellular networks there is a fundamental trade-off between the cell throughput and fairness levels for preselected users which are sharing the same amount of resources at one transmission time interval (TTI). The static parameterization of the Generalized Proportional Fair (GPF) scheduling rule is not able to maintain a satisfactory level of fairness at each TTI when a very dynamic radio environment is considered. The novelty of the current paper aims to find the optimal policy of GPF parameters in order to respect the fairness criterion. From sustainability reasons, the multi-layer perceptron neural network (MLPNN) is used to map at each TTI the continuous and multidimensional scheduler state into a desired GPF parameter. The MLPNN non-linear function is trained TTI-by-TTI based on the interaction between LTE scheduler and the proposed intelligent controller. The interaction is modeled by using the reinforcement learning (RL) principle in which the LTE scheduler behavior is modeled based on the Markov Decision Process (MDP) property. The continuous actor-critic learning automata (CACLA) RL algorithm is proposed to select at each TTI the continuous and optimal GPF parameter for a given MDP problem. The results indicate that CACLA enhances the convergence speed to the optimal fairness condition when compared with other existing methods by minimizing in the same time the number of TTIs when the scheduler is declared unfair.

Cognitive Radio (CR) has emerged as the promising technology to address spectrum scarcity issues. One of the challenging tasks in CR networks is to agree on a common control channel to advertise the free channel list (FCL) amongst the participating CR nodes for co-operative communication. In this paper, a CR MAC protocol for searching, scanning, and accessing the control channel is proposed. The protocol consists of two levels of selection: rapid channel accessing and reliable channel accessing. In rapid channel accessing, nodes quickly and efficiently converge to a newly found control channel. In reliable channel accessing, switching to the backup control channel is performed whenever there is a PU claim on licensed channel. Furthermore, our reliable channel accessing allows CR nodes to access more than one control channel simultaneously. We evaluate the performance of the proposed approach through simulation modelling. The results show that our protocol can achieve efficient channel access time and fairness.

The tradeoff concept between system capacity and user fairness attracts a big interest in LTE-Advanced resource allocation strategies. By using static threshold values for throughput or fairness, regardless the network conditions, makes the scheduler to be inflexible when different tradeoff levels are required by the system. This paper proposes a novel dynamic neural Q-learning-based scheduling technique that achieves a flexible throughput-fairness tradeoff by offering optimal solutions according to the Channel Quality Indicator (CQI) for different classes of users. The Q-learning algorithm is used to adopt different policies of scheduling rules, at each Transmission Time Interval (TTI). The novel scheduling technique makes use of neural networks in order to estimate proper scheduling rules for different states which have not been explored yet. Simulation results indicate that the novel proposed method outperforms the existing scheduling techniques by maximizing the system throughput when different levels of fairness are required. Moreover, the system achieves a desired throughput-fairness tradeoff and an overall satisfaction for different classes of users.

Synchronous bandwidth, defined as the maximum time a node is allowed to send its synchronous messages while holding the token, is a sensitive parameter for deadline guarantees of synchronous messages in a timed token network. In order to offer such guarantees, synchronous bandwidth has to be allocated carefully. The allocation of synchronous bandwidths to a general synchronous message set with the minimum message deadline (Dmin) larger than the target token rotation time is studied. A new approach for allocating synchronous bandwidth, which can be easily implemented in practice, is proposed. It is demonstrated, through simulations and numerical examples, that the proposed approach performs better than any of previously proposed local synchronous bandwidth allocation schemes, in terms of its ability in guaranteeing hard real-time traffic.

We investigate timing properties of the timed token protocol that are necessary to guarantee synchronous message deadlines. A tighter upper bound on the elapse time between the token's lth arrival at any node i and its (l + /spl nu/)th arrival at any node k is found. A formal proof of this generalized bound is presented.

Finding the worst-case message response time is important for guaranteeing message deadlines in any hard real-time communication environment. This paper proposes an $O(n)$-time algorithm for exactly capturing the worst-case synchronous message response time in an FDDI network for any given set of synchronous message streams whose message deadlines are no longer than periods. The proposed algorithm can be used to form an optimal test on whether or not a given setting of network parameters can meet the message deadline constraints for a considered synchronous message set with deadlines no longer than periods.

A highly cost-effective adaptive routing algorithm for n-D meshes is designed. It employs only two virtual channels. The planar routing and the 2D VBMAR are used as constructs for it. Routing freedom is greatly enhanced to improve performance. As a result, the algorithm features load balance and high adaptivity. Deadlock freedom of the algorithm is proved. Extensive simulations show that it not only outperforms an algorithm with equal virtual channels, but also excels the fully routing algorithm with much more virtual channels.

Exactly capturing the minimum available time for a network node to transmit its synchronous messages during any given length of time is important for guaranteeing the transmission of synchronous messages before their deadlines in a timed token ring network (such as FDDI) where the timed token medium access control protocol is used. Previous results which only give a lower bound on the minimum transmission time could be too pessimistic for supporting the timely delivery of synchronous traffic in a hard real-time communication environment. This paper presents an O(n)-time algorithm for exactly determining the minimum available transmission time. The algorithm can be used to test whether or not a given allocation of synchronous bandwidths can guarantee a synchronous message set with message deadlines no more than periods; the test so formed is better than any previous testing method whose test is sufficient only, in the sense that it is both sufficient and necessary and therefore optimal.

We study the problem of guaranteeing deadlines of synchronous messages with their deadlines no more than periods in an FDDI network. In order to guarantee the transmission of a synchronous message before its deadline, a precise knowledge of the worst-case situation for such transmission is required. This paper studies the worst-case scenario for transmission of a synchronous message in an FDDI network, in which the time elapsed between arrival of the message and completion of its transmission will reach its maximum (i.e., the longest possible interval). A few conditions under which the worst-case situation can happen are formally derived.

We investigate the inherent timing properties of the timed token medium access control (MAC) protocol that are necessary for guaranteeing synchronous message deadlines in a timed token ring network such as the Fibre Distributed Data Interface (FDDI) network. As a result, the best-so-far result of the upper bound on the time possibly elapsed between any number of successive token arrivals at a particular node was derived. This result, which is particularly important for studies on real-time communications in any timed token ring network, has been published by Zhang and Burns (see IEEE/ACM Trans. on Networking, vol.3, p. 729-41, 1995) with no proof given. In this paper, we complement our early work by presenting a concise formal proof of this upper bound.

One of the key issues related to deadline guarantees of synchronous messages in FDDI networks is the feasibility test of synchronous bandwidth allocation (SBA). Much work has been done on how to allocate synchronous bandwidth properly in order to guarantee synchronous messages transmission before their deadlines, and as a result quite a few SBA schemes have been proposed. In contrast, limited research has been done on how to effectively test the feasibility of a given allocation of synchronous bandwidths (i.e., whether or not a given allocation can guarantee the synchronous message set considered), and the existing approach cannot always give an effective test and may even produce wrong testing results. This paper presents an algorithm of polynomial time worst-case complexity which ensures the correctness and effectiveness of the testing for any given synchronous message set with message deadlines equal to periods.

This paper is concerned with deadline guarantees of synchronous messages with deadlines equal to periods, in a timed token ring network such as FDDI where the timed token medium access control (MAC) protocol is used. The timed token protocol guarantees a bounded access time and an average bandwidth for synchronous traffic. However, this guarantee alone, though necessary, is insufficient for guaranteeing the transmission of synchronous messages before their deadlines. To ensure timely delivery, the synchronous bandwidth must be carefully allocated to individual nodes. We propose and analyse an efficient and practical local synchronous bandwidth allocation (SBA) scheme. The new scheme performs better than any previously published as it calculates the synchronous bandwidth such that during the message period, the total synchronous transmission time definitely available (when judged only by local information) is exactly equal to the transmission time required. Our scheme also differs significantly from previously reported ones by explicitly taking into account the synchronous bandwidth allocation for message sets whose minimum message deadlines (D/sub min/) are less than twice the target token rotation time (TTRT), and consequently can apply to any synchronous message set (with D/sub min/>TTRT). The feasibility of the allocations produced by the proposed scheme and the worst case achievable utilisation of the scheme are also discussed.

We study the problem of guaranteeing synchronous message sets with the minimum message deadline (D/sub min/) less than twice the target token rotation time (TTRT), i.e., D/sub min/<2/spl middot/TTRT, in an FDDI network. While the restriction of D/sub min//spl ges/2/spl middot/TTRT on a synchronous message set is generally accepted as a necessary condition for all message deadlines to be guaranteed, our study shows that this is not always the case. Some synchronous message sets with D/sub min/<2/spl middot/TTRT may still be guaranteed. We show that there is no need to retain the above restriction on a synchronous message set before determining its schedulability. Furthermore an improved version of a previously-published local synchronous bandwidth allocation scheme, with removal of this restriction, is proposed.

This paper investigates the inherent timing properties of the timed-token medium access control (MAC) protocol necessary to guarantee synchronous message deadlines in a timed token ring network such as, fiber distributed data interface (FDDI), where the timed-token MAC protocol is employed. As a result, an exact upper bound, tighter than previously published, on the elapse time between any number of successive token arrivals at a particular node has been derived. Based on the exact protocol timing property, an optimal synchronous bandwidth allocation (SBA) scheme named enhanced MCA (EMCA) for guaranteeing synchronous messages with deadlines equal to periods in length is proposed. This scheme is an enhancement on the previously published MCA scheme.