Monday, December 09, 2019

signals processing, artificial intelligence

If you are a brilliant Artificial Intelligence (AI) programmer the U.S. Army Rapid Capabilities Office (RCO) may have a challenge just for you! Between April 30, 2018 and July 30, 2018 participants may go online to register and compete for a total of $150,000 in prize money with the winner awarded $100,000, second place receiving $30,000, and third place receiving $20,000. The registration for the competition will be conducted by the Mitre Corporation online at:https://sites.mitre.org/armychallenge/starting on April 30, 2018.

The potential battlefield of the future is becoming a signals processing nightmare with so many different sources and technologies competing for the electromagnetic spectrum. The Army's Rapid Capabilities Office (RCO) is hoping to assist Electronic Warfare Officers (EWO) in identifying and reacting to these signals by applying artificial intelligence and machine learning capabilities to the environment. RCO will provide the challenge competitors with a training dataset consisting of 4.3 million instances occurring within 24 different modulations including noise class modulations. The 24 modulations to predict will include: BPSK,QPSK, 8PSK, 16PSK, QAM16, QAM64, 2FSK_5KHz, 2FSK_75KHz, GFSK_75KHz, GFSK_5KHz, GMSK, MSK, CPFSK_75KHz, CPFSK_5KHz, APSK16_c34, APSK32_c34, QAM32, OQPSK, PI4QPSK, FM_NB, FM_WB, AM_DSB, AM_SSB, NOISE.

The government is seeking to challenge problem solvers to develop and present advanced algorithms and AI code that can provide a high degree of classification accuracy and performance with low CPU resource requirements. This would enable the government to implement the systems within their existing technology. This type of signals processing would represent the next level of advancement for the battlefield EWO. The government will seek to obtain technical papers(s) describing the methodology of specific implementations that include both the model process and training process. Competitors must also provide essential source code elements that show the implementation process. Each participants' solutions will be scored independently by means of a cross-entropy loss function. Aggregate scores will be based on individual CSV file submissions for each dataset. Test set files submitted must be named "TestSet1Predictions.csv"..."TestSetNPredictions.csv."

Blind signal classification requires little to no prior knowledge. Solutions would automatically classify the modulation scheme or change of RF waveform as the first step in the signal classification process. During the challenge, participants will submit their scores daily and a status leader board will display how well each is performing to help spur competition. The classification challenge will be open for approximately 90 days. Participants will have a minimum of 60 days to develop their models and to work with the training data. There will be only two test data sets released for the competition with a solution submission window of 15 days after publication. The first dataset will be released about 67 days after the start date and the second dataset will be released about 84 days after launch.

Additional details regarding the judging process can be found at:https://www.challenge.gov/challenge/army-signal-classification-challenge/.  Today, Python is generally considered to be the top AI development language but others include R, Lisp, Prolog, and Java.  Happy coding!

GTranslate

en English es Spanish