Adapting Copycat to context-dependent visual object recognition