| 1904 | | |
| | 1904 | def pairwise_crosscorrelate(self, nb_pairs, pairs_generator=None, time_bin=1., average=True, lag=None, n_pred=1, predictor=None, display=False, kwargs={}): |
| | 1905 | """ |
| | 1906 | Function to generate an array of cross correlations computed |
| | 1907 | between pairs of cells within the SpikeTrains. Based on pairwise_cc but |
| | 1908 | using function analysis.crosscorrelate instead. |
| | 1909 | |
| | 1910 | Inputs: |
| | 1911 | nb_pairs - int specifying the number of pairs |
| | 1912 | pairs_generator - The generator that will be used to draw the pairs. If None, a default one is |
| | 1913 | created as RandomPairs(spk, spk, no_silent=False, no_auto=True) |
| | 1914 | time_bin - The time bin used to gather the spikes |
| | 1915 | average - If true, only the averaged CC among all the pairs is returned (less memory needed) |
| | 1916 | display - if True, a new figure is created. Could also be a subplot. The averaged |
| | 1917 | spike_histogram over the whole population is then plotted |
| | 1918 | kwargs - dictionary contening extra parameters that will be sent to the plot |
| | 1919 | function |
| | 1920 | |
| | 1921 | Examples |
| | 1922 | >> a.pairwise_cc(500, time_bin=1, averaged=True) |
| | 1923 | >> a.pairwise_cc(500, time_bin=1, averaged=True, display=subplot(221), kwargs={'color':'r'}) |
| | 1924 | >> a.pairwise_cc(100, CustomPairs(a,a,[(i,i+1) for i in xrange(100)]), time_bin=5) |
| | 1925 | |
| | 1926 | See also |
| | 1927 | pairwise_pearson_corrcoeff, pairwise_cc_zero, RandomPairs, AutoPairs, CustomPairs |
| | 1928 | """ |
| | 1929 | subplot = get_display(display) |
| | 1930 | |
| | 1931 | ## We have to extract only the non silent cells, to avoid problems |
| | 1932 | if pairs_generator is None: |
| | 1933 | pairs_generator = RandomPairs(self, self, False, True) |
| | 1934 | |
| | 1935 | # Then we select the pairs of cells |
| | 1936 | pairs = pairs_generator.get_pairs(nb_pairs) |
| | 1937 | N = len(pairs) |
| | 1938 | if newnum: |
| | 1939 | length = 2*(len(pairs_generator.spk1.time_axis(time_bin))-1) |
| | 1940 | else: |
| | 1941 | length = 2*len(pairs_generator.spk1.time_axis(time_bin)) |
| | 1942 | if not average: |
| | 1943 | results = numpy.zeros((N,length), float) |
| | 1944 | else: |
| | 1945 | results = numpy.zeros(length, float) |
| | 1946 | for idx in xrange(N): |
| | 1947 | # We need to avoid empty spike histogram, otherwise the ccf function |
| | 1948 | # will give a nan vector |
| | 1949 | hist_1 = pairs_generator.spk1[pairs[idx,0]].time_histogram(time_bin) |
| | 1950 | hist_2 = pairs_generator.spk2[pairs[idx,1]].time_histogram(time_bin) |
| | 1951 | if not average: |
| | 1952 | results[idx,:] = analysis.ccf(hist_1,hist_2) |
| | 1953 | else: |
| | 1954 | results += analysis.ccf(hist_1,hist_2) |
| | 1955 | if not subplot or not HAVE_PYLAB: |
| | 1956 | if not average: |
| | 1957 | return results |
| | 1958 | else: |
| | 1959 | return results/N |
| | 1960 | else: |
| | 1961 | if average: |
| | 1962 | results = results/N |
| | 1963 | else: |
| | 1964 | results = numpy.sum(results, axis=0)/N |
| | 1965 | xaxis = time_bin*numpy.arange(-len(results)/2, len(results)/2) |
| | 1966 | xlabel = "Time (ms)" |
| | 1967 | ylabel = "Cross Correlation" |
| | 1968 | subplot.plot(xaxis, results, **kwargs) |
| | 1969 | set_labels(subplot, xlabel, ylabel) |
| | 1970 | pylab.draw() |
| | 1971 | |
