A Novel Learning Rate Decay Function of Kohonen Self-Organizing Maps Using the Exponential Decay Average Rate of Change for Image Clustering

Peserta didik menghadapi persaingan yang semakin ketat untuk memasuki perguruan tinggi karena angka peminat dan penyeleksi seleksi masuk perguruan tinggi semakin tinggi. Hal itu membuat peserta didik mempersiapkan segalanya dimulai dari memilih perguruan tinggi hingga memilih program studi. Penelitian ini dilakukan untuk memetakan atribut atau alasan-alasan mahasiswa baru Jurusan Matematika Fakultas Matematika dan Ilmu Pengetahuan Alam Universitas Tanjungpura dalam memilih program studi. Pemetaan ini dilakukan dengan menggunakan jaringan syaraf tiruan algoritma Kohonen Self Organizing Maps yang hasilnya divalidasi dengan metode IDB (Indeks Davies-Bouldin). Penelitian dilakukan untuk mengelompokkan alasan-alasan mahasiswa baru menggunakan 3 klaster dan 4 klaster dengan learning rate 0.05, 0.25, 0.5, 0.75 dan 0.95 serta maksimum iterasi 50, 100, 500, 1000, 2000 dan 5000. Berdasarkan jumlah klaster dan learning rate serta maksimum iterasi tersebut diperoleh IDB terkecil sebesar 1.8226 yaitu dengan menggunakan 3 klaster, learning rate 0.05 dan maksimum iterasi 500. Diantara 3 klaster yang terbentuk maka klaster ke-1 yaitu klaster dengan nilai mean terendah sehingga berdasarkan penskoran kuesioner maka masuk dalam kategori sangat penting. Artinya anggota dalam klaster tersebut menjadi pertimbangan para responden dalam memilih program studi. Keanggotaan klaster ke-1 diantaranya yaitu peluang karir, keinginan mencapai cita-cita, tenaga pendidik profesional, akreditasi program studi, instansi terbaik untuk bekerja dan peringkat universitas. Kata Kunci : Klaster, Learning Rate, Indeks Davies-Bouldin

Download Full-text

Antarctic sea ice: a self-organizing map-based perspective

Annals of Glaciology ◽

10.3189/172756407782871549 ◽

2007 ◽

Vol 46 ◽

pp. 391-396 ◽

Cited By ~ 8

Author(s):

David B. Reusch ◽

Richard B. Alley

Keyword(s):

Sea Ice ◽

Rapid Identification ◽

Rate Of Change ◽

Spatial And Temporal Variability ◽

Self Organizing Map ◽

Sea Ice Extent ◽

Self Organizing Maps ◽

Antarctic Sea Ice ◽

Sea Ice Edge ◽

Self Organizing

AbstractSelf-organizing maps (SOMs) provide a powerful, non-linear technique to optimally summarize a complex geophysical dataset using a user-selected number of ‘icons’ or SOM states, allowing rapid identification of preferred patterns, predictability of transitions, rates of transitions, and hysteresis in cycles. The use of SOMs is demonstrated here through application to a 24 year dataset (1973–96) of monthly Antarctic sea-ice edge positions. Variability in sea-ice extent, concentration and other physical characteristics is an important component of the Earth’s dynamic climate system, particularly in the Southern Hemisphere where annual changes in sea-ice extent (temporarily) double the size of the Antarctic cryosphere. SOM-based patterns concisely capture the spatial and temporal variability in these data, including the annual progression of expansion and retreat, a general eastward propagation of anomalies during the winter, and sub-annual variability in the rate of change in extent at different times of the year (e.g. retreat in January is faster than in November). There is also often a general seasonal hysteresis, i.e. monthly anomalies during cooling follow a different spatial path than during warming.

Download Full-text

Learning rate schedules for self-organizing maps

Proceedings of the 12th IAPR International Conference on Pattern Recognition (Cat. No.94CH3440-5) ◽

10.1109/icpr.1994.576908 ◽

2002 ◽

Cited By ~ 5

Author(s):

F. Mulier ◽

V. Cherkassky

Keyword(s):

Learning Rate ◽

Self Organizing Maps ◽

Self Organizing

Download Full-text

ANALYSIS OF THE SELF-ORGANIZING MAP-BASED INVESTMENT STRATEGY

International Journal of Computing ◽

10.47839/ijc.16.1.866 ◽

2017 ◽

pp. 10-17

Author(s):

Piotr Kossakowski ◽

Piotr Bilski

Keyword(s):

Average Rate ◽

Investment Strategy ◽

Correlated Data ◽

Self Organizing Map ◽

Training Set ◽

Self Organizing Maps ◽

Strategy Performance ◽

The Impact ◽

Process Method ◽

Self Organizing

The following paper presents the application of Self-Organizing Maps (SOM) to construct and apply investment strategy on the stock market. Characteristics of this type of neural network and their influence on the investment strategy performance are verified. Considered parameters include the SOM size, here connected to the size of the training set (number of examples). The average number of patterns per neuron was selected as the appropriate measure. Other aspects of the SOM analysis included conscience mechanism, which allows more neurons to be stimulated during the learning process, method of weights updating, determining the number of stimulated neurons. Additionally, the impact of the correlation between features was verified to eliminate redundant ones. Performance of each designed network was verified against the simple investment strategy, generating “buy”, ”sell” and “hold” signals based on the average Rate of Return (RoR). Results show that SOMs with the conscience mechanism outperform their simpler configurations. Elimination of correlated data also improves performance of the SOM-based investment strategy.

Download Full-text