<p>Mitigation of channel impairments and detection of transmitted symbols remain challenging for both wireless and optical environments. Multi input multi output orthogonal frequency division multiplexing (MIMO OFDM) modeling can be used for both wireless and free space optical channels to analyze the receiver performance. But the receiver performance degrades due to fading effects in wireless channels and scintillation effects in free space optical channels. As equalizer role is crucial to compensate the channel effects, development of adaptive equalization models help to enhance the receiver throughput. Low complexity neural network architecture can be effectively used to develop equalization models that can minimize bit errors even in the practical channel conditions. Single layer neural network based equalization models are proposed in the paper to reconstruct data stream with minimum bit error. Gamma-gamma model is used to analyze free space optical (FSO) medium. Similarly the fading statistics of wireless channel can be realized using Rayleigh and Nakagami-m models which characterize outdoor wireless environment. Hyperbolic secant cost Least Mean Square (LMS) algorithm is used to train the models and the performance is verified as well as evaluated in terms of BER, MSE and eye diagram.</p>