Three-cornered coevolution learning classifier systems for classification

<p>This thesis introduces a Three-Cornered Coevolution System that is capable of addressing classification tasks through coevolution (coadaptive evolution) where three different agents (i.e. a generation agent and two classification agents) learn and adapt to the changes of the problems without human involvement. In existing pattern classification systems, humans usually play a major role in creating and controlling the problem domain. In particular, humans set up and tune the problem’s difficulty. A motivation of the work for this thesis is to design and develop an automatic pattern generation and classification system that can generate various sets of exemplars to be learned from and perform the classification tasks autonomously. The system should be able to automatically adjust the problem’s difficulty based on the learners’ ability to learn (e.g. determining features in the problem that affect the learners’ performance in order to generate various problems for classification at different levels of difficulty). Further, the system should be capable of addressing the classification tasks through coevolution (coadaptive evolution), where the participating agents learn and adapt to the changes of the problems without human participation. Ultimately, Learning Classifier System (LCS) is chosen to be implemented in the participating agents. LCS has several potential characteristics, such as interpretability, generalisation capability and variations in representation, that are suitable for the system. The work can be broken down into three main phases. Phase 1 is to develop an automated evolvable problem generator to autonomously generate various problems for classification, Phase 2 is to develop the Two-Cornered Coevolution System for classification, and Phase 3 is to develop the Three-Cornered Coevolution System for classification. Phase 1 is necessary in order to create a set of problem domains for classification (i.e. image-based data or artificial data) that can be generated automatically, where the difficulty levels of the problem can be adjusted and tuned. Phase 2 is needed to investigate the generation agent’s ability to autonomously tune and adjust the problem’s difficulty based on the classification agent’s performance. Phase 2 is a standard coevolution system, where two different agents evolve to adapt to the changes of the problem. The classification agent evolves to learn various classification problems, while the generation agent evolves to tune and adjust the problem’s difficulty based on the learner’s ability to learn. Phase 3 is the final research goal. This phase develops a new coevolution system where three different agents evolve to adapt to the changes of the problem. Both of the classification agents evolve to learn various classification problems, while the generation agent evolves to tune and adjust the problem’s difficulty based on the classification agents’ ability to learn. The classification agents use different styles of learning techniques (i.e. supervised or reinforcement learning techniques) to learn the problems. Based on the classification agents’ ability (i.e. the difference in performance between the classification agents) the generation agent adjusts and creates various problems for classification at different levels of difficulty (i.e. various ‘hard’ problems). The Three-Cornered Coevolution System offers a great potential for autonomous learning and provides useful insight into coevolution learning over the standard studies of pattern recognition. The system is capable of autonomously generating various problems, learning and providing insight into each learning system’s ability by determining the problem domains where they perform relatively well. This is in contrast to humans having to determine the problem domains.</p>

Three-cornered coevolution learning classifier systems for classification

<p>This thesis introduces a Three-Cornered Coevolution System that is capable of addressing classification tasks through coevolution (coadaptive evolution) where three different agents (i.e. a generation agent and two classification agents) learn and adapt to the changes of the problems without human involvement. In existing pattern classification systems, humans usually play a major role in creating and controlling the problem domain. In particular, humans set up and tune the problem’s difficulty. A motivation of the work for this thesis is to design and develop an automatic pattern generation and classification system that can generate various sets of exemplars to be learned from and perform the classification tasks autonomously. The system should be able to automatically adjust the problem’s difficulty based on the learners’ ability to learn (e.g. determining features in the problem that affect the learners’ performance in order to generate various problems for classification at different levels of difficulty). Further, the system should be capable of addressing the classification tasks through coevolution (coadaptive evolution), where the participating agents learn and adapt to the changes of the problems without human participation. Ultimately, Learning Classifier System (LCS) is chosen to be implemented in the participating agents. LCS has several potential characteristics, such as interpretability, generalisation capability and variations in representation, that are suitable for the system. The work can be broken down into three main phases. Phase 1 is to develop an automated evolvable problem generator to autonomously generate various problems for classification, Phase 2 is to develop the Two-Cornered Coevolution System for classification, and Phase 3 is to develop the Three-Cornered Coevolution System for classification. Phase 1 is necessary in order to create a set of problem domains for classification (i.e. image-based data or artificial data) that can be generated automatically, where the difficulty levels of the problem can be adjusted and tuned. Phase 2 is needed to investigate the generation agent’s ability to autonomously tune and adjust the problem’s difficulty based on the classification agent’s performance. Phase 2 is a standard coevolution system, where two different agents evolve to adapt to the changes of the problem. The classification agent evolves to learn various classification problems, while the generation agent evolves to tune and adjust the problem’s difficulty based on the learner’s ability to learn. Phase 3 is the final research goal. This phase develops a new coevolution system where three different agents evolve to adapt to the changes of the problem. Both of the classification agents evolve to learn various classification problems, while the generation agent evolves to tune and adjust the problem’s difficulty based on the classification agents’ ability to learn. The classification agents use different styles of learning techniques (i.e. supervised or reinforcement learning techniques) to learn the problems. Based on the classification agents’ ability (i.e. the difference in performance between the classification agents) the generation agent adjusts and creates various problems for classification at different levels of difficulty (i.e. various ‘hard’ problems). The Three-Cornered Coevolution System offers a great potential for autonomous learning and provides useful insight into coevolution learning over the standard studies of pattern recognition. The system is capable of autonomously generating various problems, learning and providing insight into each learning system’s ability by determining the problem domains where they perform relatively well. This is in contrast to humans having to determine the problem domains.</p>

Molecular Design of a “Two-in-One” Orthosteric-Allosteric Chimeric Mutant Selective EGFR Inhibitor

Inhibitors developed to target the epidermal growth factor receptor (EGFR) are an effective therapy for patients with non-small cell lung cancer harbouring drug-sensitive activating mutations in the EGFR kinase domain. Drug resistance due to treatment-acquired mutations within the receptor itself has motivated development of successive generations of inhibitors that bind in the ATP-site, and third-generation agent osimertinib is now a first-line treatment for this disease. More recently, allosteric inhibitors have been developed to overcome the C797S mutation that confers resistance to osimertinib. In this study, we present the rational structure-guided design and synthesis of a mutant-selective EGFR inhibitor that spans the ATPand allosteric sites. The lead compound consists of a pyridinyl imidazole scaffold that binds irreversibly in the orthosteric site fused with a benzylisoindolinedione occupying the allosteric site. The compound potently inhibits enzymatic activity in L858R/T790M/C797S mutant EGFR (4.9 nM), with relative sparing of wild-type EGFR (47 nM). Additionally, this compound achieves cetuximab-independent, mutant-selective cellular efficacy on the L858R and L858R/T790M variants

Molecular Design of a “Two-in-One” Orthosteric-Allosteric Chimeric Mutant Selective EGFR Inhibitor

Inhibitors developed to target the epidermal growth factor receptor (EGFR) are an effective therapy for patients with non-small cell lung cancer harbouring drug-sensitive activating mutations in the EGFR kinase domain. Drug resistance due to treatment-acquired mutations within the receptor itself has motivated development of successive generations of inhibitors that bind in the ATP-site, and third-generation agent osimertinib is now a first-line treatment for this disease. More recently, allosteric inhibitors have been developed to overcome the C797S mutation that confers resistance to osimertinib. In this study, we present the rational structure-guided design and synthesis of a mutant-selective EGFR inhibitor that spans the ATPand allosteric sites. The lead compound consists of a pyridinyl imidazole scaffold that binds irreversibly in the orthosteric site fused with a benzylisoindolinedione occupying the allosteric site. The compound potently inhibits enzymatic activity in L858R/T790M/C797S mutant EGFR (4.9 nM), with relative sparing of wild-type EGFR (47 nM). Additionally, this compound achieves cetuximab-independent, mutant-selective cellular efficacy on the L858R and L858R/T790M variants

RL-Duet: Online Music Accompaniment Generation Using Deep Reinforcement Learning

This paper presents a deep reinforcement learning algorithm for online accompaniment generation, with potential for real-time interactive human-machine duet improvisation. Different from offline music generation and harmonization, online music accompaniment requires the algorithm to respond to human input and generate the machine counterpart in a sequential order. We cast this as a reinforcement learning problem, where the generation agent learns a policy to generate a musical note (action) based on previously generated context (state). The key of this algorithm is the well-functioning reward model. Instead of defining it using music composition rules, we learn this model from monophonic and polyphonic training data. This model considers the compatibility of the machine-generated note with both the machine-generated context and the human-generated context. Experiments show that this algorithm is able to respond to the human part and generate a melodic, harmonic and diverse machine part. Subjective evaluations on preferences show that the proposed algorithm generates music pieces of higher quality than the baseline method.

Survival Outcome of Salvage Therapy with Second-Generation Novel Agents in Patients with Relapsed and Refractory Multiple Myeloma in Routine Clinical Practice

Survival outcome of patients with multiple myeloma (MM) has been markedly improved by the introduction of novel agents such as bortezomib, thalidomide, and lenalidomide over the past decade. Recent clinical studies have shown further improvement by second-generation agents, namely carfilzomib, ixazomib, pomalidomide, elotuzumab, and daratumumab. However, the prognosis of patients who became refractory to both bortezomib and lenalidomide was extremely poor, and the efficacy of the second-generation agents in such patient remains unclear in routine clinical practice. In this study, we analyzed the outcome of MM patients in terms of overall survival (OS) and survival time after salvage therapy. A total of 174 patients (89 male and 85 female) treated between 2003 and 2017 were enrolled. The median age was 69 years old (range 35-89). The type of M protein was IgG in 93, IgA in 35, IgD in 10, Bence Jones type in 30, and non-secretory type in 5 patients, respectively. The international staging system (ISS) stages I, II, and III were 32, 66, and 63 patients, respectively. There were 100 patients with normal serum LDH and 39 with high LDH. As for initial therapy, 62 patients were treated with conventional chemotherapy, 83 with bortezomib-based regimens, 7 with lenalidomide-based regimens, and 22 with bortezomib + lenalidomide + dexamethasone. Forty-seven patients received autologous stem cell transplantation in the upfront setting. During the observation period 120 patients relapsed and 6 were refractory to initial therapy, and 38 patients were treated with conventional chemotherapy and 58 with the first-generation novel agents such as bortezomib, lenalidomide, or thalidomide as salvage therapy. The remaining 30 patients were further treated with the second-generation novel agents such as carfilzomib in 21, ixazomib in 5, pomalidomide in 6, elotuzumab in 4, or daratumumab in 8 patients after refractory to bortezomib and lenalidomide. The median progression-free survival was 17.7 months and the median OS was 51.0 months from initial therapy. Regarding the outcome according to salvage therapy, the median OS was 22.0 months for conventional chemotherapy group, 46.4 months for first-generation agent group, and 98.7 months for second-generation agent group (p<0.0001). The survival of the second-generation group was longer compared with the first-generation group in both transplanted (median OS, 80.5 vs 118.3 months) and non-transplanted patients (39.2 vs 56.6 months). In patients who received the first-generation agents as initial therapy, the median OS was 45.1 months for the first-generation group and 71.5 months for the second-generation group. In the survival time after salvage therapy, the median survival was 18.6 months, and there was no significant difference by initial therapy (16.8 months for conventional chemotherapy group and 20.1 months for novel agent group). According to salvage therapy, the median survival after salvage therapy was 7.5 months for conventional chemotherapy group, 20.4 months for first-generation agent group, and 29.5 months for second-generation agent group (p<0.001, Figure). The survival benefit after salvage therapy with the second-generation agents was also observed in both transplanted (median survival after salvage, 29.5 vs 69.8 months) and non-transplanted patients (18.6 vs 23.6 months). In patients who received the first-generation agents as initial therapy, the median survival after salvage was 25.0 months for the first-generation group and 29.5 months for the second-generation group, and again survival time was longer in the second-generation group. Regarding the risk factors, the median survivals after second-generation novel agent therapy were not reached, 69.8, and 14.8 months in the ISS I, II, and III stages, respectively, and were 69.8 and 14.8 months in normal and high LDH groups, respectively. Thus, our results have demonstrated that the second-generation novel agents significantly prolonged survival in relapsed and refractory patients regardless of the use of the first-generation agents. However, the survival benefit was mostly observed in standard risk and transplanted patients but not in high-risk patients. Therefore, further studies are needed to establish more effective strategies using the next generation agents in routine clinical practice. Figure. Figure. Disclosures No relevant conflicts of interest to declare.