Generated recordings with same template and config, different from previously generated with older package version

[LuighiV]

Hi @alejoe91,

I am trying to generate recordings with other noise levels, using the templates and configuration that I used to generate recordings with older package version. However, I am noticing that the pattern generated is different from the previous ones. I am using the same seeds for both cases.

This is the configuration information:

{'spiketrains': {'n_exc': 7, 'n_inh': 3, 'f_exc': 5, 'f_inh': 15, 'st_exc': 1, 'st_inh': 3, 'min_rate': 0.5, 'ref_per': 2, 'process': 'poisson', 'gamma_shape': 2, 't_start': 0, 'duration': 10}, 'cell_types': {'excitatory': ['PC', 'SS', 'SP'], 'inhibitory': ['AC', 'BP', 'BC', 'BTC', 'ChC', 'DBC', 'MC', 'NGC']}, 'seeds': {'spiketrains': 100, 'templates': 100, 'convolution': 100, 'noise': 100}, 'templates': {'min_dist': 25, 'min_amp': 50, 'max_amp': 300, 'n_overlap_pairs': None, 'xlim': None, 'ylim': None, 'zlim': None, 'overlap_threshold': 0.9, 'n_jitters': 10, 'upsample': 8, 'pad_len': [3, 3]}, 'recordings': {'fs': 24000, 'dtype': 'float32', 'overlap': False, 'extract_waveforms': False, 'sync_rate': None, 'sync_jitt': 1, 'chunk_duration': 0, 'modulation': 'electrode', 'sdrand': 0.05, 'bursting': False, 'bursting_units': None, 'exp_decay': 0.2, 'n_burst_spikes': 10, 'max_burst_duration': 100, 'shape_mod': False, 'shape_stretch': 30.0, 'n_bursting': None, 'noise_level': 0, 'noise_mode': 'uncorrelated', 'noise_color': False, 'noise_half_distance': 30, 'far_neurons_n': 300, 'far_neurons_max_amp': 10, 'far_neurons_noise_floor': 0.5, 'far_neurons_exc_inh_ratio': 0.8, 'color_peak': 300, 'color_q': 2, 'color_noise_floor': 1, 'filter': True, 'filter_cutoff': [300, 6000], 'filter_order': 3, 'drifting': False, 'drift_mode': 'slow', 'n_drifting': None, 'preferred_dir': [0, 0, 1], 'angle_tol': 15, 'slow_drift_velocity': 5, 'fast_drift_period': 10, 'fast_drift_max_jump': 20, 'fast_drift_min_jump': 5, 't_start_drift': 0}}

Do you know why it has this behavior? Perhaps there is something that I am missing in the configuration for the new package?

Thanks,
Luighi

[alejoe91]

Hi Luighi,

What versions are you comparing? If it's a major version release (1.8 vs 1.9) then there might have been major changes that don't make the results exactly the same. In that case you would have to use the previous version

[LuighiV]

Hi @alejoe91,

Thanks for your prompt answer. Yes, I was working with version 1.8, and when I updated to the newest I had obtained that behavior. I will try using the previous one, although I will also have to use an outdated version of numpy as the latest has issues with 1.8 version (np.float not supported as alias in new version anymore).

Thanks!
Luighi