Hasil Pencarian  ::  Simpan CSV :: Kembali

"Dalam makalah ini, penulis membahas dan memperkenalkan sebuah metodologi pencaharian struktur arsitektur Jaringan Neural Buatan propagasi balik berbasis fuzzy (JNB-Fuzzy) yang optimal dengan menggunakan algoritma genetika. Optimasi struktur jaringan neural dapat dilakukan dengan memperkecil jumlah neuron dalam lapis tersembunyi atau jumlah bobot dalam jaringan neural. Dalam makalah ini penulis membuat optimasi struktur jaringan dengan memperkecil jumlah bobot dalam jaringan, karena jumlah bobot ini jauh lebih besar daripada jumlah neuron yang ada. Jaringan neural yang telah dioptimasi ini kemudian digunakan sebagai subsistem pengenal pola pada Sistem Penciuman Elektronik yang dikembangkan oleh penulis. Hasil eksperimen dengan menggunakan jaringan ini menunjukkan peningkatan derajat pengenalan sistem, dari 70,4% pada struktur jaringan tidak dioptimasi, menjadi 85,2% bila menggunakan struktur jaringan yang telah dioptimasi.

---

In this article we proposed a method for optimizing the structure of a fuzzy artifi cial neural networks (FANN) through genetic algorithms. This genetic algorithm (GA) is used to optimize the number of weight connections in a neural network structure, by evolutionary calculating the fi tness function of those structures as individuals in a population. This fuzzy neural is then applied as the pattern recognition in our developed odor recognition system. Experimental results show that the optimized neural system provides higher recognition capability compare with that of unoptimized neural system. Recognition rate of the unoptimized neural structure is 70.4% and could be increased up to 85.2% in the optimized neural system. It is also shown that the computational cost of the optimized structure of neural system is also lower than the unoptimized structure."

"Beginning with the Airline Deregulation Act of 1978 in US, followed by the European Union in 1997, airlines have been constructing route networks of their own choosing rather than operating ones implicitly chosen for them by civil aviation authority. These changes have had profound effects on many aspects of airline operation, particularly fares, service, quality, and safety. But, most importantly, airlines have altered their route structures by developing hub-and-spoke networks, and this has affected all of these aspects. This structure is likely to flourish around the world as a consequence of airline liberalization and the growing trend toward privatization of this industry.In a hub-and-spoke network, centrally located service facilities serve as the hubs. Flows from a set of outlying nonhub nodes arrive at hubs and, after regrouping, all leave the hub facilities bound either to other hubs or to their ultimate destinations. Thus, the flows from the same origin with different destinations are consolidated on the route to a hub facility and the flows with different origins but the same destination on the route out of a hub facility. The centralization and broader scope of operations let the system take advantage of economies of scale.This paper proposes a framework to optimize the flight network using hub-and-spoke system. This problem consists of the determination of hub number, hub location and route assignment in order to minimize the overall transportation cost. The model is solved using genetic algorithm approach. Two networking strategies are considered:1. Strict hubbing, in which a spoke is assigned to exactly one hub and all flows to/from spoke are channeled trough the same hub and 2. Nonstrict hubbing, in which a spoke can be assigned to more than one hub under certain condition. Different values of airport fixed costs are also implemented. Variations of these strategies are evaluated along with various parameters of air transport production using data on air passenger flows between top 30 Indonesian airports in 2000.The result shows that the adoption of hub-and-spoke network increase the overall system performance with increasing load factor, frequency, coverage area, revenue passenger kilometer, available seat kilometer and more efficient utilization of aircraft. Moreover, Nonstrict hubbing strategy offers smaller total system cost, more routes and more nonstop flights."

"Makalah ini membahas pengembangan Sistem Penciuman Elektronik menggunakan 16 buah sensor kuarsa terlapis membran sensitif. Penulis telah mengembangkan Sistem Penciuman Elektronik dengan jumlah sensor sebanyak 4 buah, akan tetapi sistem ini hanya mampu membuat klasifikasi aroma campuran dengan tingkat pengenalan dibawah 40%. Pengembangan sistem dilakukan dengan meningkatkan jumlah sensor untuk memperbesar dimensi ruang pengamatan dan peningkatan frekuensi dasar sensor untuk mendapatkan akurasi yang lebih tinggi.Hasil penelitian menunjukkan bahwa sistem 16 sensor mempunyai kapabilitas yang tinggi untuk klasifikasi aroma campuran. Tingkat pengenalan sistem dengan 16 sensor untuk aroma campuran dengan 6 tingkat konsentrasi alkohol berkisar 89.9%, bila diproses secara terpisah, sedangkan apabila dilaksanakan secara ?batch? akan menghasilkan tingkat pengenalan sekitar 82.4%.

---

An artificial odor recognition system is developed for discriminating odors. This artificial system consisted of 16 quartz resonator crystals as the sensor array, a frequency modulator and a frequency counter for each sensor that are connected directly to a microcomputer. We have already shown that the artificial odor recognition system with 4 sensors is high enough to discriminate simple odor correctly, however, when it was used to discriminate compound odors, the recognition capability of this system is dropped significantly to be about 40%.Results of experiments show that the developed artificial system with 16 sensors could discriminate compound aroma based on 6 gradient of alcohol concentrations with high recognition rate of 89.9% for non batch processing system, and 82.4% for batch processing of the classes of odors."

"Sistem penciuman elektronik telah dikembangkan dengan menggunakan kuarsa terlapis membran sebagai sensornya dan jaringan neural buatan Propagasi Balik (JNB-BP) sebagai sub-sistem pengenal polanya. Beberapa kelemahan penggunaan JNB-BP pada sistem penciuman elektronik adalah lamanya waktu pembelajaran dan adanya keterbatasan dalam mengenal pola aroma campuran. Untuk mengatasi masalah tersebut maka digunakan implementasi algoritma jaringan neural buatan berbasis Probabilistic Neural Network (JNB-PNN). JNB-PNN mempunyai 2 proses utama dalam tahap pembelajarannya yaitu menggunakan data pelatihan untuk membangun topologi JNB-PNN dan mencari parameter pemulus/smoothing parameter.Pengujian yang dilakukan dengan mengklasifikasikan aroma campuran secara bertahan yaitu 6, 8, 12 dan 18 aroma. Tujuan daritahapan pengklasifikasian tersebut adalah untuk melihat kemampuan dari sistem dalam mengenai pola dari aroma campuran dengan membandingkan penggunaan JNB-BP dan JNB-PNN. Hasil kedua eksperimen menunjukkan bahwa semakin banyak pola aroma uamg diklasifikasin, tingkat pengenalan sistem semakin menurun. Kemampuan dari sistem penciuman elektronik yang menggunakan JNB-BP dalam mengenal 18 pola aroma menghasilkan tingkat pengenalan di bawah 70%. Sedangkan untuk JNB-PNN, walaupun terjadi penurunan terhadap pengenalan 18 pola yang diujikan, hasil pengenalannya masih di atas 90%."

"Dalam beberapa tahun ini, telah banyak penelitian mengenai pengenalan pola yang dilakukan dengan jarigan syaraf tiruan. Skripsi ini membahas sistem pengenalan pola berbasis Jaringan Saraf Tunggal (JST). Penelitian ini membahas metode pembelajaran Levenberg Marquardt dalam melakukan pengenalan pola. Terdapat 9 dataset pola, 8 dataset dari "UCI Repository of Machine Learning Database" dan satu set dari data uranium dioxide pellet. Prosedur kerja sistem terdiri dari tahap pra-pemrosesan, pelatihan, dan pengujian. Hasil pengujian yang ditinjau dari computational cost dan recognition rate menunjukkan JSE berbasis metode Levenberg Marquardt memberikan performa yang lebih baik dibandingkan JST berbasis metode Levenberg Marquardt atau Backpropagation.

---

In recent years, many people have been working on pattern recognition using artificial neural network. This bachelor pra-thesis discuss about pattern recognition system based on Single Neural Network (SNN). This research discuss about Levenberg Marquardt learning algorithm in pattern recognition.There are 9 datasheets used in this experiment, which 8 of them are obtained from "UCI Repository of Machine Learning Database" and and one dataset of uranium dioxide pellet. The working procedures of the systems consists of pre-processing, training, and testing stages.

The testing result, which is measured from computational computational cost and recognition rate, shows that ENN based on Levenberg Marquardt learning algorithm has a better performance than SNN based on Levenberg Marquardt or Backpropagation."