This thesis makes use of a many-core architecture developed by the Adapteva Company toimplement a parallel version of the Hierarchical Temporal Memory Cortical LearningAlgorithm (HTM CLA). The HTM algorithm is a new machine learning model which ispromising in the aspect of pattern recognition and inference. Due to its complexity,sufficiently large simulations are time-consuming to perform on sequential processor,therefore, in this thesis we have investigated the feasibility of using many-core processors torun HTM simulations.In this thesis, a parallel implementation of the HTM algorithm on the proposed many-coreplatform has been done in C. In order to evaluate the performance of parallel implementation,some metrics such as speedup, efficiency and scalability have been measured throughperforming some simple pattern recognition tasks. Implementing the HTM algorithm on asingle-core computer established the baseline to calculate the speedup and efficiency ofparallel implementation for the purpose of evaluating scalability.In this thesis, three mapping methods which are block-based, column-based and row-based,have been selected to parallelize the HTM from many mapping methods. In the experimentwith small training examples, the row-based mapping method gained the best performancewith a high speedup because of the lesser influence of training example variability, andreflected a good scalability when implemented on different numbers of cores. However, theexperiment with a relatively large amount of training examples gives almost identical resultsfrom all three mapping methods. In contrast with the small experiment, the full set experimentused much more diverse input and the mapping method did not influence the average runningtime for this training set. All three mappings have showed almost perfect scalability and thereis linear speedup increasing with number of cores, for the dataset and HTM size used.